Hey friends,

So today I want to explore what I think to be a more robust approach to design a model to any sort of effect you want to capture.

Imagine that you’re a fisherman and your objective is to capture as much fish as you can from a single location. You already vetted that location, you know there’s fish there, now all you need to do is harvest it.

You have a few options:

• Use 1 small net

• Use multiple nets with different properties and sizes

• Use multiple boats to cover a wider area

What would you rather choose to achieve your objective in an optimal way?

Well, I am not a fisherman but I’d be willing to bet that if there’s enough fish to cover the costs, I want to run that at capacity and at least have the largest nets and multiple boats to capture all I can from that one location.

Well I see the sort of trading I do from the same lens. You have a certain effect (fish) and you want to capture as much as you can from it.

You can be the fisherman with 1 small net, or you can be the fisherman running things more optimally.

So for the past few weeks I’ve been working on turning one of my trend models into a multi-signal continuous model. You might wonder why not just leave it to run as is, since it’s been performing just fine (performance below).

In this article I will not disclose the signals themselves as thats part of my proprietary research, so ill keep it broad. The purpose is to show the effectiveness of this method and make the case for its long-term added robustness.

Read more

If you’re not a subscriber, here’s what you missed this past month so far:

If you’re not yet a part of our community, subscribe to stay updated with these more of these posts, and to access all our content.

Hey friends,

Over the past few articles, we have been working on this “dirty carry crypto trade (part 1 and part 2)”. The first model was inefficient because we didn’t care that much about the signal itself which made it turnover 100x+ of capital a year, which is A LOT for these sorts of models, and my intuition tells me that our signal isn’t that strong to be turned over so often.

That poses an interesting question, how really does the signal behave, and what other variations tell us about the signal?

Well, let’s look at it.

How should we think about a signal? To me, certainly not from the traditional online backtestoors perspective, where a signal is used merely as a rule to buy above or below, and hope for the equity curve to be heading towards the top right corner of the screen.

Backtests alone are prone to problems like path dependency and pure luck. But looking at what a signal on average should yield, well, that’s a better way of looking at it in my opinion.

This is a recent podcast I listened to, where liquiditygoblin touches on that topic for a bit, which I completely subscribe to – having a defined forecast horizon and sticking to it. It’s not just about did this backtest made money or not. It is about what does this signal yield in a forward looking period.

If you do a backtest of a random buy of the S&P500 and hold it for the next 10 days, it’d probably made money over the past decade, and no one would say that strategy has an edge right?

The idea of starting from a forward return perspective is having something that is expected to make money over a future horizon on average, and then decomposing it from known market returns (a topic for another day).

How can we define that horizon for ourselves?

Well in the last piece of these series we noticed how funding is so sticky in that first 5 day window, and so that should be our initial threshold.

Let’s begin there before we think about anything else. What’s the expected price movement in the first N days after the signal?

I want to reinforce the idea that this is PRICE MOVEMENT that we’re measuring. Why am I making that a point? Well, this is a carry trade, so we expect to get paid that carry, but there’s a whole directionality thing we have to solve for. If we go long BTC and short DOGE, these things don’t move together ALL the time, so we need to think carefully about that later on. Also we want to be measuring apples to apples right? So how do we define our forward return given that DOGE has a way differently volatility profile than say BTC?

Read more

If you’re not a subscriber, here’s what you missed this past month so far:

If you’re not yet a part of our community, subscribe to stay updated with these more of these posts, and to access all our content.

Hi friends,

We have to take a break from our series on the dirty carry trade (part 1 and part 2) to look at an interesting concept that I found when running my systems though different exchanges.

Below there’s 4 simulated equity curves, by exchange (binance, extended, hyperliquid, kraken) for a standard crypto momentum model.

As we can see, there’s a considerable disparity between DEXE’s (decentralized exchanges) and CEXE’s (centralized exchanges). The distinction between a DEX and CEX is probably not the root cause here but rather the second order effects that newer exchanges might create. I have a few other ideas about why this might be the case, but let’s prove it with some data and figure it out.

This is an exploratory article, as that’s what’s most useful to me when there’s no solid answers, like right now as I am writing this… just a thesis and a dream (just kidding sounded nice).

Let’s begin by taking a look at the individual day’s returns for each exchange.

Most days the exchanges are pretty aligned in returns. This makes sense as the underlying model is the same. But something rather interesting happened in early November-25 (mostly on the 7th and 10th)…

You can see that only Binance lost money on that day, and pretty heavy, over 15% cut from yesterday’s equity.

If we go into the excel of the open positions on those days, we can clearly see what I am saying. During November, most exchanges start virtually the same, but as it approaches the 7th, Binance accumulates major losses, where the others don’t.

Lets take a look at the individual losses on binance that differed from the other exchanges.

Read more

If you’re not yet a part of our community, subscribe to stay updated with these more of these posts, and to access all our content.

Hey friends,

For the past few days the WTIOIL-USDC contract on hyperliquid has been paying between 200%-400% (the annualizoor enters the room) to everyone that is willing to long this thing.

That is about 1.1% each day, just to be long.

The question is , why?

Also is it that easy to make money off of it?

I am going to be straightforward and say that these are not the markets I trade or think about often, so if you see something wrong, please tell me, so I can move forward with better information. This will be my attempt to understand these insane funding readings over the last few days on the WTIOIL-USDC contract on hyperliquid.

So let’s start going down the line of reasoning of why this is happening.

The WTIOIL-USDC contract on Hyperliquid tracks the price of the front-month WTI futures contract which is the nearest-expiry oil delivery contract on the CME. Think of the oil futures market as a chain of monthly contracts, each representing a barrel delivered at a specific date:

• May delivery

• June delivery

• July delivery

And the relationship between these contracts is described as either in contango or backwardation.

What is contango if not a different breed of an orangotango? (sry about the dad joke)

Well imagine you’re a Saudi producer and I want to buy oil from you.

I say I want 1,000 barrels delivered and you tell me the price is $60 per barrel.

But I want delivery in 6 months.

So you tell me, well, that’s $62 a barrel.

Why are you charging me more though? Because you have costs if you’re committing to that future delivery, like storage, insurance, and financing the inventory. Those added costs you pass to me, the consumer. The further out the delivery, the more it costs.

This is the contango, the futures curve slopes upward. It’s the normal state for most commodities.

Now, what is backwardation?

Well, lets say I want those 1,000 barrels delivered today.

You tell me that the price is $62 per barrel.

What if I want the delivery in June?

Well you tell me that for that it costs $60.

But why is it cheaper for a later delivery? We just said that this operation has added costs so it goes against our idea on the contango scenario. Because right now the supply is tight and everyone needs oil IMMEDIATELY. The premium to have it is now, since there’s not much going around. Nobody’s paying that storage premium, and on top of that, the market is discounting future delivery because it expects the supply crunch to ease. This is backwardation, the curve slopes downward.

So now back to our hyperliquid contract. Why is it paying so much negative funding?

Well, first things first, we have this little thing that apparently no one looks at in the crypto world, that is called documentation.

If we go to the deployer’s (tradexyz) documentation of this contract, we find that the WTIOIL Oracle is currently the CLK6 futures contract, which is the Crude Oil front-month futures contract.

But we won’t be trading the May contract forever right? It has to roll to the next contract which is supposed to happen between April 8-14. After the roll, it will track the June contract which is the CLM6 contract that is currently trading around $86.56.

Now, how is the rolling of the contract done? This neat little thing called the documentation once again has the answer.

Now you can see that the rolling dates are presented in the Nth business day, which varies from month to month.

So for our current situation it translates to:

Since we are in backwardation (June’s contract is trading at lower price than front-month price) we have baked into the perpetual contract an automatic price drop each day from today until April 14th. The market has a CERTAIN expectation at what price our contract should be trading.

Given that expectation, the market is already discounting the current front-month contract price for that of the June contracts. Obviously the market isn’t waiting for the exact second to trade at new prices. This means that we’re trading at a sticky discount to the front-month (CLK6) price.

It’s sticky because there’s no longer an incentive to push price towards the current front-month contract as that will be replaced in a matter of days. So that stickiness is making the funding really really negative.

Now can you arb this to make money? Well, let’s do some back of the napkin maphs. This is where I could mess things up but let’s try not to.

Lets say you short the hyperliquid perp at current price (93.362) and long the June contract at (86.56) each for $10,000. We know that they will trade close together on the 14th, so we have a basis to collect $6.8, which amounts to a hypothetical $729 return from the pure convergence.

However we have 406% funding to pay, lets again assume it stays the same and we will be paying 1.11% a day for that exposure. That will be a total of 156.5 hours until the full roll which is about 6.52 days. Given the $10,000 HYPERLIQUID leg exposure we will pay about -$725.24 in funding and the arb is wiped out.

That trade then gets REALLY negative if we factor in fees and market impact, which I won’t even get into here.

Now there’s a few things I am too lazy (maybe) to estimate for.

1. The equal position size at different prices means I don’t have a completely market neutral position. This was purely used as an example, one should factor in the different prices into the sizing formula.

2. I assumed the full 400% annualized funding we’re seeing now. Well as the basis collapses, since hyperliquid oracle will begin adding weight to the new month contract, also will these fundings begin collapsing (or increasing towards less negative, in this case), so using the same funding reading as of now, might be a bit too conservative.

There’s other ways to look at this that might offer good potential trades, but that’s your job now, not mine innit?

I hope it clears things up as it did for me.

Disclaimer: The content and information provided by the Trading Research Hub, including all other materials, are for educational and informational purposes only and should not be considered financial advice or a recommendation to buy or sell any type of security or investment. Always conduct your own research and consult with a licensed financial professional before making any investment decisions. Trading and investing can involve significant risk of loss, and you should understand these risks before making any financial decisions.

If you’re not yet a part of our community, subscribe to stay updated with these more of these posts, and to access all our content.

Hey friends,

In part 1 of our crypto carry experiment, we’ve analyzed how a simple carry strategy performed in crypto, without any prior data analysis.

We went straight for the backtest, and it showed. As with most backtests, that lack the careful thinking about the underlying idea, it didn’t yield much on its own.

So we’ll continue down the logic line to find a solution for this trade.

In this series of articles (oh look how fancy we are talking about series) we will go into multiple steps to gain some intuition about the trade we want to put on first, and then construct the strategy to harvest it. The first thing we’ll do will be to analyze how sticky the funding rate is. If we’re going to be harvesting funding, we want to be able to figure out if what we’re harvesting sticks long enough to compensate for fees and market impact.

Lets start with a simple analysis.

We will analyze if and how current day’s funding is predictive, of funding N days into the future. I’ve removed the tails from this starting analysis as I don’t want those skewing our results for now. We will touch on the tails (no puns intended) later in the article.

We can see that fundings seem to be indeed predictive and kinda sticky.

After the 5d mark the correlation begins to decay, and that makes general sense.

This tells us that if we’re trading within that 1d-5d window we have a reasonable expectation to get paid funding during those days. All that we should care for is how much we’re getting paid vs the costs of trading in an out. Topic for later!

One thing I often see discussed is the tails of the funding. We removed the tails from the starting analysis as they were skewing the results. But my intuition tells me that when funding gets too crazy, it might be more prone to mean revert than normal funding. And it makes sense right? If something is paying so much, everyone’s looking at it because they want it. It’s a beacon of light in the mud.

Now I’ve ran the exact same analysis that I did on top, but now ONLY on the removed tails. We can see that indeed the stickiness is weaker and also decays faster than the “normal” funding readings.

That’s what we expected.

I’ve also traded this myself without any data. I tried to harvest the largest fundings, but they often decayed too fast and forced me to exit of positions that were not paying me anymore, and to be frank, were quite illiquid.

Lets try to prove this though.

I will use dollar volume as a proxy for liquidity. I say proxy because volume isn’t really liquidity innit? But I don’t have book data available at the moment, so dollar volume will do, let’s not be nerds about it. We’re paid to trade not to improve our accuracy by 2% (totally made that number up).

What’s interesting is that although there’s no clear relationship between dollar volume and funding, there’s an interesting behavior at the right end of the tail of dollar volume.

We can see that as volume scales higher, the variance of funding decreases. Goes right into what I was just mentioning from my own observations when I attempted to harvest the largest funding rates. They were usually hard to get because these were smaller coins, and many times, I found myself squeezing some allocation in the books, for little to no profit after it was all and done. Just headaches. Not saying that there’s no opportunity there, just my observation trading it, skill issue perhaps.

This makes general sense because assets with more volume correlate well with larger market caps, and those tend to not get large deviations from their underlying, after all, they’re harder to move. We can even look at the deciles of funding by volume, and observe that the left tail, and the middle section, are more funky than the higher liquidity stuff.

If we look at the outlier frequency by volume decile it shows that the left tail of volume (smaller coins) are a factory of outliers.

The takeaway of this is that although volume isn’t predictive of the direction of funding, low volume coins have far more extreme funding swings. The variance is quite high at the left tail.

So after all this data we have an idea of what we might want to go for. We don’t want a pure raw funding signal, but rather restrict it to the highest volume traded universe, so that we can harvest more “stable” funding and not be misled by extreme funding readings that don’t help us much.

Now let’s move on with the nerdiness and see how we can make money off of this signal.

But that’s for the next post, this one getting already quite extended.

I hope you’ve enjoyed today’s simple exploration of funding readings.

Disclaimer: The content and information provided by the Trading Research Hub, including all other materials, are for educational and informational purposes only and should not be considered financial advice or a recommendation to buy or sell any type of security or investment. Always conduct your own research and consult with a licensed financial professional before making any investment decisions. Trading and investing can involve significant risk of loss, and you should understand these risks before making any financial decisions.

If you’re not a subscriber, here’s what you missed this past month so far:

If you’re not yet a part of our community, subscribe to stay updated with these more of these posts, and to access all our content.

Hey friends,

A few months ago I learned about a new perp dex running a points program with a pretty reasonable innovation compared to other perp dexes. The problem was, they didn’t have an API. So I had to execute manually each day.

That’s what I did. Below is the equity curve since December — the major jumps are deposits.

Total deposits have been about $12,500, and we’re sitting at $12,455 equity right now (not doing so great!).

Today I want to actually test this strategy properly — go through the data and figure out if we’re doing something right or not.

For context: when I started trading this, since there was no API, I had to pull data from another exchange and generate signals there to trade here. The goal wasn’t to make a lot of money — it was to not lose much while farming points. We’ve done over $3M in volume on a small account, so we’re achieving that. But I can’t scale it further unless I can make the strategy actually work.

So I collected a bunch of data and today we’re running through it.

The idea is pretty basic. Perpetual contracts pay funding to incentivize one side to converge with the underlying. When the perp drifts too far from spot, the mechanism kicks in: it pays the side that pushes price back toward convergence, and charges the side pushing away from it — based on their notional size.

A simple carry strategy would be to go long the lowest-funding coins, short the highest-funding coins, collect the spread. This is simple enough that we can use to collect points on these perp dexes, since they reward OI (open interest) more than most other metrics, it’s pretty easy to maintain, and it makes sense why we should get paid for it.

But turns out, it isn’t as simple as I thought. Story of our life innit?

Let’s look at some data first. How do funding payments actually distribute across assets?

Read more

If you’re not yet a part of our community, subscribe to stay updated with these more of these posts, and to access all our content.

Hey friends,

Last week, on the 16th of March, I woke up, did my usual morning portfolio check, and saw a -6% loss on the portfolio, staring back at me.

My first instinct? Exchange bug. Sometimes wallets don’t read capital correctly, happens. I refreshed. Still -6%. Day ended slightly better, below -5%, but the damage was done.

So what happened? Memecoins happened.

“Risk Neutral” — but not really

My portfolio runs a net-neutral approach. Long and short in volatility-weighted amounts, so in theory, I’m not making big directional bets.

I say “risk neutral” very loosely though. Because we’re only neutralizing volatility. Not all risks are mitigated.

When I broke down the major losses that day, there’s exactly one thing they all have in common…

Memecoins.

74% of all losses that day came from memecoin exposure. Just that one sector. And when I checked the gross allocation, memecoins accounted for over 45% of total exposure that morning.

Imagine having close to half your portfolio in a single sector. On a day where that sector implodes together.

So what do we do about it?

The idea isn’t complicated. Set sector caps so you never have too much gross exposure to any single sector.

This is the same logic we apply to individual positions. You wouldn’t put 20% of capital into a single name (in most cases). So why let sector exposure run wild when clearly some sectors carry more tail risk than others?

The sector tail analysis confirms this pretty clearly. Memecoins and AI have the fattest tails of the lot — the distributions have long right and left extremes, which means the possibility of truly violent days is higher in these sectors than anywhere else.

Now, I want to be clear about one thing. I am not saying we should pick and choose which sectors to trade. That would be discretionary garbage layered on top of a systematic model. What I’m saying is capping how much exposure we ever take to a single sector seems like a reasonable, rules-based constraint — the same way position size limits are reasonable.

I also wouldn’t go optimizing the cap levels based on historical sector volatility. That’s an overfit waiting to happen. Sector dynamics shift. Keep it simple.

So let’s just test it and see what the data says.

The Tests

I ran three configurations on Hyperliquid perps:

Iteration 1 — baseline (no caps), my current live system

Iteration 2 — 20% cap on all sectors (both sides)

Iteration 3 — 20% cap on Meme sector only (short side)

Baseline — what I’m running now

First, the benchmark. No sector caps. This is the system live in production.

CAGR of 77.4%, Sharpe of 1.61. Max drawdown -23.6%.

20% cap on all sectors — ouch

Now let’s see what happens when we cap every sector at 20% of capital.

The performance gets a massive downgrade.

CAGR drops from 77.4% to 19.5% — a -75% relative hit. Sharpe falls from 1.61 to 0.81. And while volatility compresses from 40.8% to 26.3%, the return compression is disproportionately worse. Worst of all, max DD duration worsened from 145 days to 223 days. So you’re taking less risk but spending more time underwater. Not the deal you’re looking for.

Is that surprising? Tbh, a bit. But when you think about it, it makes sense. In crypto, returns are heavily concentrated in sector rotations. When a sector is running, the edge is in riding it. By capping how much you can be in each sector, you’re basically forcing the model to spread capital into weaker bets.

It’s like saying, in equities, that a large chunk of returns came from the tech sector. Is that wrong? I don’t think so. That’s just where the momentum was.

Let’s at least do a sanity check on the backtest before moving on — always do sanity checks.

Good. The mechanism works correctly. Every time a sector goes above 20%, it gets rebalanced back to target. The plumbing is clean.

20% cap on Meme sector only — interesting

Now the more surgical version: cap only the Meme sector, on the short side.

Performance barely moves.

CAGR: 73.8% vs baseline 77.4% — only -4.6% relative. Sharpe actually ticks up slightly, from 1.61 to 1.63. Max DD improves from -23.6% to -20.8%, and max DD duration compresses significantly — from 145 days down to 97 days.

This is the best risk-adjusted result of the three.

Surgical precision beats a blanket rule. The Meme-only cap shaves a small amount of return while actually cleaning up the drawdown profile. The day I described at the start? That cap would’ve saved a chunk of it.

Putting them all together

The baseline is the historical winner. Meme-only cap is close. All-sector cap is not a competition.

Honestly, the gap between baseline and Meme-only is small enough that I don’t particularly care which one to run. If the delta isn’t large, I’d rather stick with the simplest version. Occam’s razor and all that.

Binance confirms it

This is still a limited dataset. Hyperliquid only goes back to early 2023. So let’s run the same sequence on Binance spot, where history goes back to late 2017.

Iteration 4 — Binance baseline (no caps)

Iteration 5 — 20% cap on all sectors (both sides)

Iteration 6 — 20% cap on Meme sector, short side only

Same pattern. Baseline wins. Other iterations lag.

The all-sector cap drops CAGR from 52.0% to 8.9% — that’s -83% relative, even worse than on Hyperliquid. Sharpe collapses from 1.22 to 0.45. Max DD duration goes from 487 days to 1,502 days — over 4 years spent in drawdown. The strategy was barely trading.

The Binance meme short-only cap holds up much better: CAGR 48.8% vs 52.0%, Sharpe 1.18 vs 1.22. The damage is minimal.

Why does capping all sectors hurt so much? Theory time — though I’ll be honest, I could be totally wrong here.

In crypto, returns are heavily driven by sector rotations. When momentum is running in a sector, that’s where the edge concentrates. A blanket cap forces the model to cut the winners and spread into the laggards. In a trend-following framework, that’s basically the opposite of what you want to do.

The Meme short-only cap is a different story. That’s a pure risk constraint, not a return bet. You’re saying: I don’t want to run heavily short a sector that’s structurally driven by speculation and narrative. That’s reasonable. And the data shows it costs almost nothing in returns.

What I’m taking from this

Blanket sector caps are not the way to go. The momentum strategy’s alpha comes from concentrating in sectors where trends are strongest — cap everything uniformly and you’re killing the engine.

Surgical caps, targeting only the most volatile or structurally weird sectors (memes), are nearly free. Small CAGR cost, cleaner drawdown, and you avoid the kind of morning I described at the start.

For now? I’m keeping things simple. The baseline version is what I run. But if I were to add any constraint, the Meme-only cap is the obvious candidate — it’s the closest thing to a free lunch this analysis produced.

Disclaimer: The content and information provided by the Trading Research Hub, including all other materials, are for educational and informational purposes only and should not be considered financial advice or a recommendation to buy or sell any type of security or investment. Always conduct your own research and consult with a licensed financial professional before making any investment decisions. Trading and investing can involve significant risk of loss, and you should understand these risks before making any financial decisions.

If you’re not a subscriber, here’s what you missed this past month so far:

If you’re not yet a part of our community, subscribe to stay updated with these more of these posts, and to access all our content.

Hey friends,

Today we’re testing a trading strategy I found in a paper where the author claims it produced a Sharpe ratio of 2.41. It’s a trend following strategy, so naturally I’m curious — how did they achieve a 2.41 Sharpe on trend following?

Over the past couple of weeks I’ve been heads down on a bunch of different things — new mean reversion models, infrastructure work, the usual. Those models are now running on their own, collecting data for me to analyze, so it’s a good time to go back and test new variations of what we already work with.

This one’s interesting because the paper adds complexity my systems can’t handle yet: intraday trading. My infra is built for daily rebalancing, not hourly. But that’s something I need to work on in the near future anyway, so why not start today.

I’ll publish a deeper article on why I want to trade faster, but for now here’s the tweet where I got a first taste of it.

So let’s start with the simplest possible daily implementation and iterate from there.

Spoiler: the results we find by the end are quite decent.

Let’s get into it.

First Test — Binary Signals

Read more

Hey friends,

Today is Saturday the 28th of February, 2026. I am taking the afternoon off from building my trading infrastructure to write about my trading performance in 2025, 2026 so far, and the plan I am implementing to hit sensible risk targets for my portfolio going forward.

I am building at a speed I have never had before, but sometimes it is good to take some time to reflect and plan. I’ve gone from managing a small portfolio, to having days where I was up or down 6 figures in a single day.

For the past few years I’ve always written end of year performance reviews, on top of the ones I write through the year. I want to continue that trend as I do want to build and keep a long track record.

Highlights of today’s post:

• 2025 in review — from +70% YTD to ending the year negative, and everything in between

• 10/10/2025

• Going to the ER

• 700+ hours of infrastructure building since October

• Systematic portfolio up 15% while BTC down ~40% and alts down ~35%

• Risk objectives and the plan to get there by end of 2026

• The HYPE exit strategy

Let’s get started.

Subscribe now

2025 — A Year to Remember

This last year was an emotional roller coaster. I will tell the story of how in late October I ended up in the hospital emergency room due to what I can now look back and analyze as the accumulated stress of this insane year.

Entering into 2025, I was coming off one of the best periods of my trading career. Just 2 months prior in November, I had been airdropped an amount that doubled my portfolio, and I decided to hold the entire bag for a long-term investment that ended up going from $2 (where it opened) to close to $30 in less than a month.

This was a very significant moment because this single trade made it possible for me, for the first time, to live off of trading returns alone and with some margin to spare.

When that was all said and done I had 2 options: send some of that to my bank account and put it into “safer” investments, or full ape the entire stack into my trading portfolio and roll the profits.

Oh well… If you know me, you know my choice.

I am a very aggressive kind of trader and I have not spent all of this time learning this craft to make the US median income. I want to shoot for the fences, while being fully aware of the risks that come with that. There’s no room for naive optimism in this business, and if you allow any, you will soon find out the reason I am saying this. I knew the risks I was taking, but I carried on.

One of my many talents is the ability of making significant financial decisions at the absolute worst possible times. So yeh, when I full ported this infusion of cash into my crypto trading portfolio, the crypto market peaked. I should have been expecting it.

At this time I was still mostly a long-only trader. I had a momentum model, that we discussed here many times, and being net long during this period was naturally a money loser. My portfolio took a hit, and by April, I was down more than 30% from the initial starting point.

This was even worse now because I had an account magnitudes larger than what I ever had. The losses I was seeing in my portfolio were unprecedented for me. Considering I could be up or down 6 figures in a single day, I am happy how I handled it… I was pretty numb to it all, and remained rational through the process while executing the trades without allowing externalities to control my decisions. After all, I’ve been here before, just not at this monetary scale.

April comes along and we start getting the first time where the TACO trade was still somewhat of an unknown. The US administration announces tariffs to all countries, and naturally risk assets tank in the expectation of this devastating news. I had the sense that this was probably all bogus and I wrote about it at the time, so I began buying even more aggressively than before.

I bought stocks…

I bought crypto…

Every asset in sight that I knew something about, I bought.

I want to say that I am for sure NOT a macroeconomics trader. I am just a systems trader and that’s what I focus on. But this time I was right in my thesis and the market just went up after that. Being right in my thesis doesn’t mean it was a particularly intelligent one, just the one that worked out.

My portfolio not only recovered it all but by the summer, I was up over 70% YTD at the peak and over 120% from the April lows.

It was a phenomenal time. I’d literally wake up every day to a higher net worth. I went on vacation, just chilled out, and I was having a blast.

The thing is, there was something lurking underneath all of this that I did not pay attention to. There were weaknesses that I knew my portfolio suffered from but that I chose to ignore due to my recent success. I have this tendency when things are going well to take my foot off the gas and that’s what I did. And that cost me dearly over the next quarter.

10/10/2025 — A Day I Will Never Forget

Before we move into the last quarter, I can’t do a 2025 review without mentioning the 10/10 nuke to crypto. That is a day that has been ingrained in my mind that I can’t forget. It was a career defining moment for me, and a career ending moment for many. I can only remember one time when I felt as hopeless as I did for a brief few minutes on that night.

I have this habit of writing notes about something important that happened that day. This is the note I wrote the day after (pls ignore the typos):

That note doesn’t reflect how bad it was. In the span of a few hours, perpetual contracts on the most liquid traded coins collapsed almost 50%. You know what it’s like to take a 50% haircut in just a few hours? 50% is a lot for any portfolio, now imagine when that portfolio is 95% of my liquid net worth.

There was absolutely nothing I could do. I remember watching the tape go lower and lower and thinking, ok, this might be it. This might be the moment I have to restart fresh. Looking back, those feelings were purely based on panic, but for a moment, I was okay with it all. I did not care if it all went to 0. That is something that I still reflect on to this day — I literally felt nothing. I just observed in awe with absolute admiration for the events that unfolded.

10/27/2025 — Going to the ER

The roller coaster of the year, the highs and lows, the bloodbath on the 10th — I think it all contributed to an accumulation of stress that culminated in me ending up in the ER on the 27th of October.

For about a week or so, I was waking up every night at 4am, feeling chest pains, difficulties breathing, sometimes almost passing out and vomiting. I got a full check done and they determined it was stress induced.

This never happened to me before.

I am pretty chill and I don’t stress that much about stuff. But I guess it got to me, even if I didn’t acknowledge it. I do remember that when I woke up at night, the first thing I’d do was check the market. Imagine waking up at 4am to check your TradingView app. Yeh, that was me.

That was probably the lowest point of the year for me and I had to do something. I had to change. I could not live exposed to one thing for much longer.

So I decided something had to change. I wanted to add protection to my portfolio — I was no longer trading a marginal amount of money. I also wanted to be able to profit across regimes, not just when crypto went up. So I went all in into building the next version of my trading infrastructure, which I’ve spent close to 700 hours alone on since.

We will get into that, but let’s finish 2025 first.

The remaining quarter was also not very good. The crypto market was extremely hard to trade, ESPECIALLY if you were net long like I was up until that point. Ended up the year negative on what could’ve been a phenomenal year had I played my cards differently.

The thing is, even though I ended up the year in the red, it was one of my best years looking back. I cannot express how much I’ve changed as a trader from just October last year. It has been almost 6 months, but feels so much more than that to me.

I look back at 2025 as a career defining moment. I am happy it all happened.

I really do believe that in this business sometimes it takes the absolute worst pain to make you change and be better. I don’t know, maybe some people do it just by improving a bit each day, but I work differently. I have to be taught through pain, and in 2025, the market disciplined me.

If you look at all the painful lessons as lessons, not as something negative to avoid, you indeed grow. The problem is that people tend to throw unpleasant experiences under a rug, because it’s painful to remember. I get it, but force yourself to be glad for the lesson, and see the lesson for what it is — an opportunity to become a better version of yourself.

2026 — Where Everything Changed

Following October, I began pouring ALL OF MY TIME into crafting out from scratch my vision of what a trading infrastructure should look like. I had this idea for a long time, but I did not want to spend the time on it. After all, I was doing fine.

The idea was simple. I could no longer be exposed to just one side of the market and one strategy. I had to create something that was robust, that could trade across strategies, exchanges, brokers, wallets, and markets — and make it in a way that could be scalable and hassle free to work with. I see good systematic trading almost like a factory where research is the core product you monetize. You need infrastructure to do this at capacity and that’s why I decided to take a “step back” and build everything back from first principles.

I’ve spent over 700 hours on this since October. How do I know? Well I track where my time is spent…

The output is that now I have an infrastructure that can handle everything I need to make a systematic portfolio successful.

A research and backtesting module:

A position and portfolio management module:

An execution engine module that handles all the positions generated for each portfolio from each individual strategy. This became increasingly important as I traded larger size and naturally needed to execute efficiently. This also tracks all trade execution details so we can measure and track performance and costs.

And obviously the global performance report against benchmarks.

This may seem relatively simple to put together but it takes a long time to make something work at this scale and well. The backend that manages this entire infrastructure — making sure that all strategies across portfolios, wallets, exchanges is working like clockwork, all the data is tracked appropriately, safety protocols are in order — that’s where all the development is.

Also another big feature is that I made the backtesting module completely integrated with the live trading execution module. This means that I can just plug in new strategies that I researched directly into live execution with the push of a button.

All of these small features require a lot of thinking and development time to ensure they’re working at capacity. Only when you get into the weeds of it, and decide to build everything from scratch and not take shortcuts, will you see how complicated this gets at the tails.

All of this work culminated in a performance that I am really proud about. Bitcoin is down close to -40% and altcoins -35%, while my systematic portfolio is up 15% since inception.

But now that work is just getting started. I’ve spent the last 6 months mostly putting infrastructure together, in a way that I am confident to put my money behind it, and to manage my portfolios at capacity. As you can see, the AUM allocated to this infra has been scaling steadily, and is now a large chunk of my portfolio.

Now I still have goals I want to hit in my portfolio and I am far from hitting them. These are not return goals or something like that, but rather from a risk perspective. Let’s talk about them.

Objectives to Reach in 2026 and Beyond

I wrote this thread back in December 2025 with the goal of putting on paper my portfolio performance from a risk perspective, and today I will expand a bit on it as it touches on all the targets I want to hit this year.

Let’s start by briefly touching on the entire performance up until the end of 2025. I don’t want to talk too much about it as that is part of all the posts we did on the portfolio performance updates page, but I do think we did not touch on it from a risk aspect.

Until the last day of December 2025, the global portfolio grew by 1874.1%.

Obviously the volatility is insane, and something that no normal person should go for. I have taken very large concentrated bets at times and will probably take more in the future. I don’t particularly see something wrong with that if you know the risks you’re taking.

However that’s not a sensible target I want to set for the majority of my portfolio right? The bulk of my portfolio should ideally be mostly controlled volatility, risk aware, robust long-term strategies. Or at least that’s what I want. I will always have a portion of my portfolio allocated to more “degenerate” bets, but it shouldn’t be a significant portion of the portfolio. It’s just not worth the risk and the stress MOST of the time.

Up until late 2024, for a long time, I had essentially a long alts, short BTC exposure in my long-only momentum model. I had a few manual interventions in late 2023 which caused me to actually be uncorrelated to alts, but that’s all that it was. At the peak, you can see that most of my portfolio was just dominated by altcoin performance. I noticed that, and it was super uncomfortable to have your entire portfolio dictated by what the altcoin market was doing.

Approaching the end of 2024, I began scaling down as the market was also dying out and my beta to the market approached 0. What’s interesting to me is that my beta to a standard trend model was practically 0 and at the peak of April it was negative. I actually thought this was wrong but no. What I was doing was fundamentally a long alts trade, and that is what is reflected here. I was not aware of this risk at the time, but now looking back, how dumb was it!

In November I received the HYPE airdrop and for a long time my performance was mostly dominated by that massive HYPE exposure, as expected, given that most of this time my portfolio was at least 50% HYPE, sometimes more, and the remaining in long-only trend.

To this day, the beta of the global portfolio is still close to 0.5 to HYPE and 0.8 to alts. But if we zoom in after October, you can notice how my beta to trend has been creeping up a lot. That is because of the new models I have been working on, and that is something reflected here.

Ideally, we don’t want to overbet on a single asset when we can put that cash to work on the systematic side.

My thesis for HYPE played out. I did well. I took risk when risk was there to be taken. But having that much exposure going forward is something I just don’t want to do.

To this day I still carry ~46% of HYPE exposure but I will explain the reasoning of why I still carry it later on.

Also I don’t want that beta to trend to be so high. This is where adding truly diversifying strategies over the next few months will come in. Ideally by the end of 2026, I can refer back to this post and show that my risk exposures got to the target we set out, and that’s what I’ll be working on.

Plans for my HYPE Exposure in 2026

I am going to be very straightforward here. I’ve never held any significant position for more than a few days or weeks other than my long-term US equities portfolio, some BTC, and now HYPE. I really like the project, and I do believe it will continue to succeed. But I must think from a systematic trader point of view.

There have also been tax reasons for me to hold this position — in my country it’s tax free on crypto gains that are held for over 1 year (with some caveats). And I have been holding this for almost 1 year and half now.

My thesis is still solid and I think it has good legs to materialize. Not only that, but in an era where increasingly AI agents will take financial decisions, they will require fast venues to do financial operations and Hyperliquid is positioned to capture a lot of that flow. We’ve started to see some of that, for example with the Alpha Arena context where they deployed AI agents to trade and used Hyperliquid as the venue.

There’s concerns I have, especially regulatory in future less lenient crypto administrations that are hard to estimate for. I am not a fundamental investor, that’s not my forte. If I see something extremely mispriced, be sure that I will be there. But if something is fairly priced or the upside is just not that significant for me to take a big swing on, I will refrain from that and let my systematic portfolio capture those returns.

By the end of 2026 I expect to have lowered my HYPE exposure from 45% to less than 15%. That is a chapter I look forward to putting an end on.

There is another interesting dynamic though that I have found just by looking at my day to day exposure. Right now having HYPE in my portfolio is giving me a free-ish hedge. I say that carefully though. What do I mean? Well, my entire book is mostly net short (vol adjusted) at the moment, and the only thing giving me net long exposure is my Hyperliquid long-term position.

This means that when the market goes down and HYPE with it, the portion of the portfolio on the systematic side is dampening that drawdown. Especially because HYPE has held so much better than the vast majority of altcoins.

When the market goes up and my portfolio suffers from being net short, the HYPE leg covers that and usually even more. And we can observe that from the global portfolio view — the portfolio still does move with HYPE, but the volatility is dampened by the systematic side of the portfolio, which is a good side effect I didn’t plan for.

There’s a few risks here and this shouldn’t be seen as a hedge obviously — this is just what I observed so far. There’s a bunch of idiosyncratic risk (regulatory, exploit, etc), and if those risks were to materialize, the short book ain’t saving my HYPE long losses then. So the protection is really just on broad market beta moves, not HYPE specific tail risk.

So what I am thinking about exiting my long HYPE position is that when/if the market recovers and my portfolio becomes tilted net long alongside HYPE, I start scaling off my HYPE position, because then the portfolio would be completely net long. That’s the strategic approach I’ll take. Right now let it act almost like a hedge while at the same time I do believe it has significant upside potential. As that upside starts materializing, I begin scaling out of it and shifting that risk back to my systematic portfolio.

Conclusion

There’s not much to say to end this post that hasn’t been said already.

Here’s the risk targets we’ve set through today’s post:

• Lower HYPE exposure from ~45% to under 15% by year’s end

• Reduce beta to trend by adding truly diversifying strategies

• Shift the bulk of the portfolio to controlled volatility, risk-aware systematic strategies

If you take one thing from my experiences, it is that although I have been in this game for 8 years now, I still have a lot to learn. There’s a lesson in there.

It seems that this is a never ending game of perpetual learning. You have to look at yourself as an eternal student. There’s no masters here, just people a few years ahead on that learning curve.

I feel like every year I have so much more to learn. I can’t really describe it. You’d think that by a few years you’d have a broad grasp of what good trading is, but no — the more you dig, the more you find your own inadequacies.

The market also forces you to be extremely honest with yourself. There’s no office politics, charm, or anything else that can save your performance, like in other careers. Here you have to be at the peak of your game, or you will pay. Maybe that’s why there’s so many people that like to talk trading, but very few willing to open their track record for public scrutiny.

I am just learning the game. Every year I become a bit better, hopefully.

There’s so much more to do and I am going to get back at it.

I hope you liked today’s post. Have a great weekend.

Disclaimer: The content and information provided by the Trading Research Hub, including all other materials, are for educational and informational purposes only and should not be considered financial advice or a recommendation to buy or sell any type of security or investment. Always conduct your own research and consult with a licensed financial professional before making any investment decisions. Trading and investing can involve significant risk of loss, and you should understand these risks before making any financial decisions.

I have this strategy that I’ve been trading live since November — and it’s been doing really well. Not only has it contributed the majority of returns to my overall portfolio, it’s been surviving an awful period in crypto markets.

But I’m having an issue with it. Maybe not so much of an issue but rather an inefficiency that I want to fix. This is one of those problems that you either KNOW about intuitively when you’re building a strategy, or you discover when you trade it day after day.

So this model is very noisy at the boundary of trade selection.

What do I mean by this?

Imagine that you want to hold a maximum of 20 positions on a specific portfolio.

On average, the 20th position is a weak signal, no different than the 21st. But you take the 20th because it ranks higher than the 21st. Now, tomorrow the weak signal of that 21st position becomes a tiny bit stronger and now it’s the 20th position. So you shift your book from a full position of yesterday’s 20th, to today’s new 20th.

You can see where I’m going with this. We can easily estimate that the edge increase in this case is meaningless — it’s not like that small increase in ranking is going to meaningfully change the average outcome of that positioning. Which means we’re paying full cost of trade execution for a trade we don’t get paid to do.

The strategy does well on its own on a historical test, and the live performance has shown that too. But I believe this behavior can be optimized.

Digging Into the Data

Let’s start by plotting out the data. This is a use case you can use AI for — take the data your backtests generate, pass it through Claude or any other LLM, and direct it to the thing you want to explore. In this case, churn and costs.

The annualized drag of costs for this strategy is around 7.7% and the median hold of a position is 2 days.

We can also see that the VAST majority of these costs — almost 90% — comes from round trips during that 3-day window. That seems awfully inefficient for the type of strategy that this is.

If we dig even closer into the type of exits generating the most churn, we see that 90% of the exits — their primary reason is that they’re not in the top signals for that day.

Now, that might be a reasonable exit. If an asset isn’t the highest ranked in a strategy, we don’t want to be in it and should make room for better trades, right? But like I said at the start, we need to ensure that the change in signal is meaningful enough so that the edge we’re getting from executing that trade is enough to overcome costs.

Another important data point: 59% of trades closed within that 3-day period. On a low frequency strategy like this, that might not be optimal. Almost certainly not the 35% of trades that close on that 1-day period.

Another interesting data point — over 900 trades re-enter within a 3-day period. Meaning the asset drops from that top 20 list, and then the other day a tiny adjustment makes it get back to 20 again and we issue a full rebalance once more.

Look at this concrete example. ETH in June 2023 — see how many times we enter and exit back and forth. 6 flips in 10 days for a cost of ~$15.8 and zero net movement. But this is a different problem we might have to deal with later, because it’s whipsawing back and forth from long to short. That might mean we need to also add some EWMA or something to smooth out the signal, because in range-bound environments we don’t want to be whipsawed by these small oscillations.

The Fix: A Buffer Band

Now, let’s get to what matters and try to fix this with the simplest possible implementation for now. Quick win territory.

We’re going to implement a minimum buffer band at the boundary so that for something to enter the top positions, it needs to cross a threshold of signal strength relative to the current holder. Let’s start at 50% for this specific signal. So a new signal needs to be at least 50% stronger to displace a ranking.

Let me describe practically how this works.

Imagine you have 2 signals, A and B.

• A is today’s 20th signal.

• B is today’s 21st signal.

Signal A, on a scale of 0 to 1, has a signal strength of 10%. Signal B has a signal strength of 11%. On the old version, signal B would automatically pass to be the 20th signal of the day, issuing a full round trip of the other signal without even knowing if that 10% increase in signal strength is in any way, shape, or form economically worth the round trip.

So what we do is — ok, signal B is indeed stronger, BUT it needs to be at least 50% stronger to issue a full rebalance. Which would require in this case to be 15% strength or higher. In this case, it would not issue the new trade.

The % threshold is completely arbitrary right now, and a more “nerdy” (professional) solution would be to really understand how much you’re getting paid for this exposure after fees and issue the rebalance when it’s higher and with some margin. But today we want a quick and dirty solution that stops us from bleeding money tomorrow, not in 1 month from today.

Let’s see how this works out.

The Results

The first thing we see, as expected, is that we save a lot in costs.

Even through its short history, we can see how the gap between the old legacy model vs the one with the churn control becomes larger and larger. The more time passes, the larger this gap becomes because we’re trading way less.

One thing I want to add here — this doesn’t even factor in the cost saving we get from slippage when crossing the book. Because in lower liquidity assets, and trading on newer exchanges, which I do, this gap becomes even larger.

Now, let’s dashboard this thing out and see the comparison between the old and new model.

The one that pops out right away is that we slash about 467 whipsaw events (-34%) and we make even more returns than before.

We were also able to reduce our total trading costs by about 15%, and reduce the annual drag to 6.6%. Quite good.

Our total trades decreased by over 600 trades, which is insane. The less we trade and the more we get out of it, the better. If I were to reduce returns and reduce 600 trades, I’d still take that trade any day of the week. Trading costs money, and the impact of those costs is usually higher than what a backtest assumes.

Our median hold also went from 2 days to 3 days. Both the median and the average increased by a lot.

The returns at the end stay relatively stable, which is what we want to see in a truly robust system, with a slight increase in performance.

And that’s how we saved over 600 trades without a decrease in expected performance.

It’s usually always better to decrease the amount of trading done and to be more efficient.

The only guaranteed thing in trading is the cost you pay to trade.

I hope you’ve enjoyed this one!

Share

Disclaimer: The content and information provided by the Trading Research Hub, including all other materials, are for educational and informational purposes only and should not be considered financial advice or a recommendation to buy or sell any type of security or investment. Always conduct your own research and consult with a licensed financial professional before making any investment decisions. Trading and investing can involve significant risk of loss, and you should understand these risks before making any financial decisions.

If you’re not a subscriber, here’s what you missed this past month so far:

If you’re not yet a part of our community, subscribe to stay updated with these more of these posts, and to access all our content.

Hey friends,

The other day I figured out that my crude volatility targeting formula was, well, too crude. That meant that a lot of times I wasn’t hitting the volatility target I set for the portfolio.

Why is this even relevant? Well, I was capping the portfolio too much compared to the risk I was willing to take. I thought that my portfolio would “wiggle” by a certain amount, but it only did by half of what I targeted. By the end of today’s article, we’ll fix it. The historical equity curve jump looks like this:

The volatility will also go from half (20%) to the target of 40% as we intended:

We explored the debugging that led me to this conclusion, in more detail in the previous article, when we figured it was due to the underlying assumption that all positions were correlated with each other.

Today we’re going to get deep into the nitty gritty of correlations the math formulas and the solution for now, to this issue.

Don’t worry, I’m not going to make it math heavy like some of the quant substacks like to overcomplicate. I am not a quant, I’m just a trader using simple, robust solutions to my systematic trading. I’m going to keep it as simple as possible, because that’s what fits me. I’m not great at math. I can understand basic concepts, so I have to keep it basic.

Personally, the most important thing is understanding. I need to understand what I’m doing on the backend, otherwise I can’t do it. I think that’s a concept I share with you a lot of time on this blog.

Let’s get into it.

The Setup: Why Equal-Weight Sizing Needs Volatility Adjustment

Let’s start at the very beginning.

Imagine that you have multiple positions. You want to size them equally because, well, you don’t want a single position to contribute disproportionally to the portfolio right? Holding a SHIB position is way different than holding a BTC position in terms of volatility. One wiggles around way more than the other.

We’d already fixed for this with our crude formula that was getting us half the target. We’d take our target volatility, divide it by the positions we wanted to hold for that day (max_signals_for_day), and divide it by the annualized volatility of the asset.

The size magnitude becomes inverse to our estimate of future volatility. This means that an asset with higher expected volatility gets a smaller position, and an asset with lower expected volatility gets a larger position.

volatility_targeted_size = (target_volatility / max_signals_for_day) / annualized_volatility_20 * total_equity

Now, like I said above, we weren’t factoring in correlation of the individual positions. So if we had a book that’s long/short, market neutral, well half the book PnL is uncorrelated to the other half, right? So there’s a bit of an issue there.

First let’s talk a little bit about correlation and what it really means.

Part 1: What Correlation Really Means

Correlation is just a way to measure something like “when asset A moves, does asset B move with it?”

Think of it as a score between -1 and 1:

-1          0          +1
 |          |           |
 ↓          ↓           ↓
A up,      A tells     A up,
B down     you         B up
(always)   nothing     (always)
           about B

The formula for the correlation of two return series X and Y is:

ρ(X,Y) = Cov(X,Y) / (σ_X × σ_Y)

Where the covariance measures how X and Y move together:

Cov(X,Y) = Average of [(X_i - mean_X) × (Y_i - mean_Y)]

Let’s make it practical. Say we have 7 days of returns for BTC and ETH:

Step 1: Calculate the means

mean_BTC = (2.0 - 1.0 + 0.5 - 0.5 + 1.5 + 0.0 - 1.0) / 7 = 0.21%
mean_ETH = (1.5 - 0.5 + 1.0 - 1.0 + 0.5 + 0.5 - 0.5) / 7 = 0.21%

The mean is simply the average return over the window, in this case 7 days.

Step 2: Calculate deviations from mean

Read more

I’m reading The Creative Act by Rick Rubin.

Honestly, I didn’t think much of it at first. It’s a book for artists. I am no artist. I am a trader.

But the more I read it, the more I agree with his concepts around a lot of things being an artistic expression. Even as I’m writing this, I’m thinking about what you might enjoy reading, whether this is going to connect. But is the objective merely to please an audience, or to write new thoughts, new ideas that might be interesting and useful to you?

We allow a lot of our original ideas to be disturbed by logic and filters that we set. But what if we allowed our work to flow with whatever we’re thinking? What if we could be more in tune with our creative process to create new things, in fields where questioning the rules isn’t always accepted well?

Trading as Creative Expression

How can I be an artist in trading? Is there an art to trading?

Well, as laid out in the book, anything we do that creates something that didn’t exist before in the world, that we transmit from our minds into the “real” world, is a form of creative expression.

As a trader, I sometimes land on thoughts and ideas that deeply connect with me. Today I was reading a tweet from ShortBear. It deeply resonated with me because I find it unbelievably true.

I immediately saved it to write about it after I carefully thought about it.

Markets are indeed forward looking machines. We all get so stuck in past data, not realizing that everything with an added “confirmation”, like an abundance of past data, means that it’s probably no longer as good as something where the future is so uncertain that you’d probably get mocked for sharing your idea publicly.

I have traded models with vast amounts of data available, papers written on them, confirmation from good friends of mine that trade those models. None of those models ever paid me like bridging funds to new perp dexes and experimenting with their technology, fully aware that I might lose all my funds.

None of those ideas ever paid me like betting on copies of successful protocols at the peak of the 2021 mania.

None of those ideas paid me like launching a strategy in a new market, with only a few weeks of data and a few heuristics on why that strategy might work, and why that effect should pay me.

Being in those unknown future outcomes seems to be where the highest RR is for me, and I take them each time I find something with that potential. Being in tune with the information that is provided to me, listening and acting on it.

That is art. That is artistic expression as a trader. The ability to consume information and translate it into a system that works.

Be Guided by Beauty

One of the quotes I most like from Jim Simons:

“Be guided by beauty. I really mean that. Pretty much everything I’ve done has had an aesthetic component, at least to me. Now you might think ‘well, building a company that’s trading bonds, what’s so aesthetic about that?’ But, what’s aesthetic about it is doing it right. Getting the right kind of people, and approaching the problem, and doing it right [...] it’s a beautiful thing to do something right.”

— Jim Simons

Building a trading company right, doing it right, is the beauty in it.

Trading is the ultimate expression of art, because we’re on the cusp of the unknown. We’re all betting on a future that has not happened yet, using information from the current. If art is bringing the non-existent into reality, then trading is an art where we’re all trying to make a living despite that unknown future.

To live as a trader is the choice of a hard life. It’s the choice of volatility and never knowing what will happen tomorrow. Some seasons you feel like you’re the man, other seasons you wonder what you’re doing with your life.

Yet you must take these moments and despite them, continue to move forward, with the expectation that your creation will continue to adapt to an ever changing world.

Building and Reflection

Now to get more practical. How can this help me get better at trading?

There’s times I have to build. There’s no way around it. I have no developers around me to do all the coding work while I think about new ideas. So I have to build myself, and at times, that consumes me for months.

That happened for the past 4-5 months.

I’ve done nothing but relentlessly build my new trading infrastructure to handle scalability across multiple models, exchanges, data sources, etc. I literally poured all of my energy into that these past few months.

All because of an idea I had during the summer. An idea that was not just generated from me—it was generated from reading about what other traders I respect were doing and seeing them do well in their fields by adding a bunch of different models together. So I wanted to do the same.

For a long time I was under the impression that I should focus on one thing and one thing only, because I couldn’t do anything else without spending a ton of time on IT, development and maintenance. Although that was true a few years ago, with the help of AI, I am able to express my ideas of what I believe to be solid infrastructure much faster. In a few months, I was able to build something that I never came close to having before.

This takes me into a different topic of how AI can aid you boost your creative efforts. Not REPLACE it, but rather boost.

A coder is an artist. A coder that produces beautiful, efficient, scalable code is an artist in my eyes. Maybe that coder can and will develop at much faster rates than before. I am sure that I am. I am no coder, I am just a trader that codes to trade. AI is amazing to bring my ideas of what good trading is into beautiful, efficient and scalable code, without spending too much time on that task that is not my main occupation.

The Value of Stillness

In any case, there’s times to build, but there should also be time for reflection. Perhaps time where you do nothing but sit on a beach reading, thinking, doing nothing. Just reflecting on what’s around you and where you want to take your venture next.

As I’m writing this, I’m at a cafe. I walked here, about 30 minutes walking, with nothing but my computer and Rick’s book. My ideas flow much better because I can just stop. I’m not mindlessly consuming or focused on building. I’m just thinking about new ideas that came from this book.

That time is useful for me.

The best ideas that I’ve produced in the past, whether in trading or business—have arrived at the times I was having a walk in nature, on a path I have nearby my house. That hour, stuck in nature, with nothing but my thoughts, forces me to think and to create. You can’t ignore your inner thoughts and you have to work through them. To think through them.

And there I created a ton of ideas.

But it doesn’t have to be full quiet. I have listened to great podcasts from traders I respect, and many times I had to pause the podcast to deeply consider an idea discussed. Almost every time I’d get a new idea that I didn’t consider before and took note.

Also these ideas made so much sense to me. Almost as if it was familiar to me, yet I never heard about it before.

Rick, in his book, talks about that. Sometimes you get these familiar ideas that deeply connect with you, that make sense, but that you never heard about before. Those are the creative moments we should seek as traders and researchers. Those ideas that make sense. That can drive you forward in your own trading.

How do we tune ourselves to those good ideas, is the question.

To me, it certainly helps leaving my habitual environment like I said—go on a walk in nature, and consider these ideas with purpose. Allow the mind to wander, to flow, and not put too many constraints, yet try to pursue those ideas.

Also adding external input from traders you respect is a good habit, because it puts ideas in your mind that might manifest in the future. A lot of times I heard a trader say something so out of reach of my current understanding that it goes unnoticed. But after a few months or a year, you run into that problem, because you’re now dealing with more complex problems. And there, the thought, the idea, comes to your mind again and there’s the answer.

Immersing oneself with your field, and consuming everything related to it, is probably a good way to do it.

The Frictionless Testing Environment

On testing ideas.

I had to build an environment where I can test ideas freely and with very little effort.

I mean, I think about an effect or a signal that might have edge, I have to be able to test it within minutes. I need to test with a few lines of code. Each test can’t require a week of changes in the code to make it work. Initially, that might be the case, but for that we need to ensure that each new feature we add to our infrastructure is built with the thought in mind of future scalability.

Because if we view trading as an art, the generation of ideas is part of the creative process. If every time you have an idea, you’re burdened by the dull process of having to re-structure your entire codebase to accommodate for that new idea, you’re going to break off your creative flow and subconsciously associate new ideas with a boring, dull process, of having to do very painful work to get at the result.

That happened to me a lot.

As a result of that—and my own laziness, let’s not try to fool anyone here—I delayed my own expansion into other systems because of that. Every time I wanted to add a new model, I had this feeling of deep pain associated with it, with the knowledge of the work I had ahead. So I accepted ideas such as “I am a trend follower” or these self-affirming beliefs that I must do one thing or the other, rather than viewing my situation as a retail trader advantageous in the form of flexibility.

The process of testing needs to be efficient, almost like a well-oiled machine that flows beautifully and smoothly, not a rusty old machine that breaks each time you add some small degree of improvement.

Here we go back to beauty, like Jim Simons so well put it.

These are just my thoughts on trading and research as a form of artistic expression.

I could be wrong.

Subscribe now

Disclaimer: The content and information provided by the Trading Research Hub, including all other materials, are for educational and informational purposes only and should not be considered financial advice or a recommendation to buy or sell any type of security or investment. Always conduct your own research and consult with a licensed financial professional before making any investment decisions. Trading and investing can involve significant risk of loss, and you should understand these risks before making any financial decisions.

One of the things I’ve been struggling with lately is that some of my models WERE NOT matching the volatility target I set for them. Targeting 40% annualized volatility, only realizing about 20%. Half. And I had no idea why.

This obviously affects the returns since we’re not hitting the risk targets we set. Even though we win on sharpe against the benchmarks, the absolute returns stay far behind what we want them to be.

Spoiler: by the end of this post, we did achieve the target. But man… It was a lengthy task, took me the weekend. Here’s how the equity curve looks by the end, after we’ve fixed our formulas.

I had a few hypotheses to start with. Most of them were wrong. But I’m leaving them in this post because I think there’s a lot of value in dissecting your models and what they’re actually doing on the backend. Sometimes we learn more from digging ourselves off the wrong paths, than going straight to the right ones.

Let’s get into it.

Breaking It Down

First thing I like to do is plot things out, right? So, we’ve been running this strategy at around 20% realized volatility. Mean of 23.65%, median of 22.28%.

Is that bad? Well, not necessarily… if your target is 20%. But I was targeting 40% for this model and I’m only realizing about half. Which I don’t understand why.

Read more

Today I want to walk you through a problem I ran into while building out my trading infra. One of those annoying bugs that shows up when you’re trying to scale the code, to handle a lot of different things, from a centralized place.

I’ve been designing my system to scale across asset classes, exchanges, different strategy types, etc etc. The goal is always: build once, use everywhere.

But then I ran into something.

By the way, if you’ve been reading my blog as of late, this all “new infrastructure” talk is starting to produce interesting results, although still a small sample. Crypto has been doing really poorly, and our new portfolio of strategies, has been handling it quite well. I’ll write a performance update, soon…

When testing a binary signal (which, in most cases, is not the most efficient way to look at a signal), I noticed we were creating a very low gross exposure profile for the strategy.

Mean of 10.9%. Median of 5%. That’s... not great.

What’s Actually Happening Here?

Let me explain how a binary signal works in my system, with a very simple example:

• If close > xSignal, you get a long.

• If close < xSignal, you get a short.

Simple enough.

The issue is that on a day to day basis, this doesn’t scale well with my overall infra.

Here’s why. Let’s say I don’t want to just measure my signal at entry or exit, but I want to measure each day against a pool of signals. Why? Because imagine you’re targeting a certain volatility AND you have a bunch of signals “feeding” into your overall weight on the market. You can periodically adjust your exposure based on that volatility target AND signal weight.

But that means periodic checks when you’re designing the system.

When I designed mine, I added a bit of something extra. Say you only allow X amount of positions in any given day, and the current signal is the weakest out of all available signals. What I do is remove that position from my open positions and allocate to the next one on the list.

# If not in signals, or if using continuous signals
# and not in signal list, set target to 0
if not asset_in_signals:
    target_position_size = 0
    would_flip = True
    reason = "asset not in top signals"

Makes sense right? Rank signals, keep the best ones, rotate out the weak ones.

Here’s where it breaks.

The Binary Signal Problem

A binary signal is a single event. The way I have my infra coded, binary signals are only accounted for at the entry/exit of their signal, not continuously. Why that design? Because if we fill the following dates forward with the binary signal, we might enter a new position when another position drops off our current open positions - but we didn’t get a binary signal for it on the current day.

Let me use a concrete example.

BTC gave a breakout signal on the 22nd of April 2025. That gives us a “1” signal, full allocation for that asset.

However, if this signal is now long for the following days until an exit, the code could open that position any time between that entry point and the exit point. Because now it’s always 1 going forward.

So the design to handle continuous adjustment of signals does not work well with a binary type event.

We need to re-think it.

Finding The Bug

The bug that was creating the 10% average gross exposure, is that a position exits just because it didn’t had a signal today. Remember how we check for positions on a day to day basis to scale well with continuous models?

Look at how XRP exits the position on the 6th of January 2026.

While at the same time, the signal strength on the individual asset, is actually 1, and never exited the position.

Obviously this is a design flaw that we need to fix. The first thing I did was pinpoint exactly why XRP exits on the 6th. The reason on the logs is that the asset is “not in top signals anymore.”

But why is it not in the top signals?

Because as mentioned above, being a binary signal, it only generates a long/short signal at the binary event. But the infra design choice requires it to be a “long_signal”, or “short_signal”, in order to be included in the valid signals list for that day:

if signal_calculation_method == "continuous":
    valid_signal = (pd.notna(row['signal_strength']) and
                   row['signal_strength'] != 0 and
                   pd.notna(row['date']) and
                   row.get('annualized_volatility_20', 0) != 0)
else:
    # For binary signals, check for any long or short signal being True
    long_signal_cols = [col for col in row.index if col.startswith('long_signal_')]
    short_signal_cols = [col for col in row.index if col.startswith('short_signal_')]
    has_long_signal = any(row.get(col, False) for col in long_signal_cols)
    has_short_signal = any(row.get(col, False) for col in short_signal_cols)
    valid_signal = ((has_long_signal or has_short_signal) and
                   all(pd.notna(row.get(col)) for col in ['date']))

So the issue is that a long_signal and short_signal only occur at the binary event. That’s the crux of the matter.

Yes, we can’t just fill forward until there’s an exit, because my infra is more complex to handle a wide variety of signal types and that wouldn’t merge well.

So I need to figure out a way to ensure it’s scalable and handles these variations.

Breaking It Into First Principles

I always try to break things into first principles.

The first thing is that we want to ensure binary signals stay on for as long as it takes until they exit. So my first idea: if the model is based on a target volatility BUT with no rebalances, should we just skip checking if the position is still in the top signals?

As I considered it, it actually makes sense. If we don’t intend on rebalancing or making any adjustment before the exit signal, we don’t have to check if there’s a better signal on a day to day basis.

However, it might matter if we have a bunch of binary signals informing the exposure and suddenly one is better than the other. Then I might want to issue a rebalance.

But no, actually. If we think about it practically, let’s say we have a new signal that’s now higher ranked than one we already had. If we exit, we basically have to re-issue the entire exposure for this new signal at a price that is not the binary event price (entry/exit). That might not be the behavior we want.

So let’s start by testing this first idea.

What I did is add the option to require a material change on the signal to issue new exposure. For example:

• If the signal inverts, we target a position of 0 (signal flipped)

• If the signal directionality is the same but increases, we issue a new position based on current volatility and signal strength

elif volatility_targeting and rebalance_threshold == 1:
    # When rebalance threshold is 1 (no rebalancing), maintain current position
    # But exit if signal inverts (changes sign)
    if (previous_signal_strength > 0 and current_signal_strength <= 0) or \
       (previous_signal_strength < 0 and current_signal_strength >= 0):
        target_position_size = 0
        would_flip = True
        reason = "signal inversion with no rebalancing"
    # If signal increases in direction, calculate new position
    elif (signal_type == 'long' and current_signal_strength > previous_signal_strength) or \
         (signal_type == 'short' and current_signal_strength < previous_signal_strength):
        target_position_size = (target_volatility / max_signals_for_day) / current_volatility * total_equity * current_signal_strength
        target_position_size = min(target_position_size, total_equity * max_position_size_pct)
    else:
        target_position_size = current_position_value

This ensures the position only gets removed when there’s an actual exit signal. And then at the end of the day, the code goes back to all available signals and adds another signal that might be available.

The Result

What happens practically?

Our gross exposure goes up to 50% from the initial 10% we had.

Now, is a gross exposure of 50% the best allocation of capital?

Well, we’ll look into that in a later article.

I hope you’ve enjoyed today’s more “exploratory” post!

Subscribe now

Disclaimer: The content and information provided by the Trading Research Hub, including all other materials, are for educational and informational purposes only and should not be considered financial advice or a recommendation to buy or sell any type of security or investment. Always conduct your own research and consult with a licensed financial professional before making any investment decisions. Trading and investing can involve significant risk of loss, and you should understand these risks before making any financial decisions.

All was fun and games when I was trading a small portfolio, purely focused on high capacity, low frequency models.

So this issue is relatively new, at least in practical terms. Obviously I’ve spent a lot of time on this issue over the years, we also talk about it almost every chance we get… We’ve ALWAYS factored in costs and slippage, even in the most naive of simulations. Here’s a tweet I just randomly found from 2023 talking about it. I can go back to 2020, and even before, as far as I remember, I’ve always considered costs.

But tbh, I never REALLY had faced them. I didn’t have enough capital for it to be a real issue. So knowing about it theoretically, and having to solve for it, is a completely different ball game.

Today we have a clear example of why we say that backtests are mere sanity tests… The more I scale into new strategies, and new markets, the more I find that to be true.

Also this is something that is not just theoretical nonsense that adds no PnL. Saving on costs, is one of the most important things you can do in your own trading. To give you an example, from the past week alone, I am probably saving between $200-$400 A DAY in execution costs on this particular strategy.

You might look at that dashboard and say:

“Did you really paid 40bps to trade? Are you dumb?”

Well… a bit, but allow me to give you a bit of context where it all started and why I let it run for longer than I should.

There’s this new market that I won’t name, but that has pretty restricted capacity. I didn’t expect to be paid a lot to trade there tbh with you, I just wanted an allocation, to be able to farm future reward programs. That’s it. That was the whole purpose.

In crypto there’s these things we call “metas”, or traditionally known as trends. Basically after Hyperliquid TGE’d, a bunch of people got rich from its airdrop. There was nothing else to talk about other than Hyperliquid for months.

That captures a lot of eyeballs… and with that much attention a lot of teams saw opportunity. So they either began developing or the ones that were already developing competing protocols, adopted similar rewards structures to what Hyperliquid did.

So I began betting on all of the most prominent new markets/projects by being very early with my exposures. Let me give you a few examples:

• I was early to Hyperliquid, got a decent airdrop from it.

• I was early to Lighter, barely did anything there, got a $10k airdrop from it.

• I am early to Extended, having more than 11k points at this moment, estimations of valuation fall between $30k-$50k. I even wrote this article in March.

• I am farming Variational.

• and others…

This is one of the things I’ve been doing to add a “cherry” on top of my systems. Just being early to the next new thing with potential. Obviously there’s added counter-party risks here, but that’s a topic for another day.

So the same applied to this new market. I wanted to be early, had a few heuristics on a model that seemed fairly decent to deploy, and went ahead with it. I didn’t have a backtest for it, no idea if it would work out, but I see that as a plus. Whenever something is weird and hard to estimate for, so it is for my competitors. And I don’t shy away from taking risk.

But that often comes with problems... When you’re in unknown territory, you will run into issues. Mine was execution.

A few weeks pass, and the model is making consistent, decent returns. Best of all, uncorrelated to the market.

(If you’re wondering why the excel and the exchange’s equity curve is different, is because of sizing. As the model started performing well, I began sizing up, and that causes the % returns to be different than the $ returns. Just wanted to point that out.)

The funding I was collecting, was also fairly consistent across time. (just by this comment you probably can have an idea of what I was doing eheh)

If we take a look at the cumulative funding, net of the cumulative costs, we can see that it was doing pretty well, even though the costs weren’t even close to being as efficient as I wanted them to be. I’ll talk about it later in the post.

As it became a decent contributor of PnL to my portfolio, I went ahead and scaled size. But I did not do it properly. I did it like an ape. I increased the allocation to that strategy to over 6 figures over night.

As I should’ve expected, over the next few weeks this happened…

Well.. It isn’t a big loss. Right now sits at -3% of the total allocation to that particular strategy, which is just a portion of the portfolio. But it shouldn’t have taken me this long to realize what was going on….

I was getting extremely bad execution. As I dug through all the records, which to be fair, have been pretty poor in this new infrastructure that I’ve been building over the last 3 months, I noticed that it was due to a few things:

• Adding more size than that market’s capacity allowed (obviously).

• As more markets were added, I added them to the pool of tradable assets, once again, completely disregarding capacity.

• Crossing the spread with no controls.

• Being too careless about margin requirements.

So enough is enough, I said, something had to be done…

As a side note, I’ve been trying to set time to strictly focus on signal/feature research efficiency. One of my priorities for this year, is to make the existent models more robust and also deploy new ones. But it seems that always something comes up. Hopefully after these last few days building the tools to be a bit smarter about trade execution, we can move on to better things.

Let’s analyze the costs I incurred in trading this new market. I’ve exported all of my data and will go through it bit by bit, and then we will go over what has been implemented, with this new learned experience. I can make mistakes, I make them often, as you can see, but I don’t accept making those same mistakes twice. That is not acceptable.

One interesting metric is the costs as a function of the allocation. We can see that it began began increasing a lot when we moved our size up, on the 24th of November.

However, now a question comes up:

But wait a second, how are the costs as a function of capital, scaling? Are you trading more often? Is the daily turnover higher now than it was before?

Well, yes and no. I should be having a similar amount of turnover as it was before, because nothing fundamentally changed about the strategy. But don’t take my word for it. My intuition could be wrong. Let’s take a look at the modeled turnover over time.

We can see that despite being an absurd amount, which we’ll get into later, it’s pretty stable, according to my estimates.

I think that the increase in costs comes from the fact that I was having way more “problems” than before. Now I was getting picked off, like this morning. I got liquidated at 1 am. I can’t show the chart without disclosing my wallet or the market I am trading, since it was pretty illiquid at the time. But it was becoming a regular occurrence being liquidated on a specific pair, for things external to my system, like someone manipulating price overnight. Real world trading type things, innit?

Read more

Following the improvements to my trading infrastructure, that I’ve been working on the past few months, we’ve had a pretty solid month of performance

While BTC and altcoins had a pretty rough month, and continue down that trend, the portfolio did quite well and we’re up over 7% from last month.

This is exactly what I am aiming for, and why I’ve devoted the sheer amount of time over the last few months re-designing my entire infrastructure. The goal is to increase sharpe, consistency, and risk management. I will achieve that goal by trading a wider breadth of edges, rather than being focused on just one thing.

On top of the performance above, it’s not yet reflected the performance of a new strategy I’ve deployed, that is a bit non-traditional in the sense that it is a new market. I also wanted to harvest returns right away, so I had to build a temporary external execution engine to be able to trade it. As I am writing this, I’ve already integrated it with my main infrastructure, but at the time I had deployed it, I hadn’t.

I’ve allocated a significant portion of my resources towards that strategy, as it has performed quite above expectations and it’s adding a significant chunk of P&L to my portfolio.

The interesting thing is, I didn’t even have a backtest for it, or any test done.

I took a few heuristics of why it should work, vibe coded an entire execution engine, gave it a few rules, and each morning I began sending the orders through the exchange.

What’s funny about it, is that I thought it was a completely different trade than what it actually became.

Like I said, it’s now a significant contributor to my PnL. A lot of people get stuck in the details and making a backtest perfect, while the market is right there.

What do you think you’ll improve in a backtest that you wont find out live trading it? Even if with small capital. I’ve only discovered this new strategy because I was willing to get to the market quickly, and not spend too much time on the details. Sure, I’ve worked on those details after, and now, I’ve constructed a pretty solid model of the effect I’m harvesting, but I didn’t do it before. I could’ve spent the last 2-3 weeks designing a simulation, not knowing if it would work or not. I’d rather not do that, and get immediate feedback from the market.

But taking risk sucks innit? That’s why most people aren’t willing to go there.

There’s a few characteristics specific to this last strategy that I won’t discuss publicly. I do think there’s some alpha here so I don’t want to get into it too specifically, as you may understand. I do trading as my primary business, and diluting alpha is not a good long-term decision for my business. I am willing to share risk premia strategies and their details, like I’ve done over the last few years on this blog, but not new strategies that are capital constrained.

I’ve also been working a lot on the scalability of the systems. What do I mean by this, more practically? Well, if I design a new model, how long does it take to incorporate it into my main portfolio? Personally, there’s nothing that suggests that I shouldn’t be doing it with a click of a button. It just changes when completely new markets are added, and different execution types are required. But that gets weeded out over time, as we build those tools for individual strategies, and they become part of a well rounded infrastructure.

Indeed achieving that functionality of sending new strategies/features from design phase, to live production, with the “click of a button”.

This has another added importance. You always want to ensure that your live trading (green line) matches your simulation (purple line).

To do that you must ensure that the engine that designed the strategy, is the same engine that produces the target exposures.

For example, you can see that I am running at a slightly positive delta to the simulation, and that delta has been broadening a bit over the past few days. You may ask me, why is it? Well, I am going to be very straightforward with you, I am not sure yet. I have few assumptions:

• Cost/slippage assumptions on simulation are more aggressive than reality.

• The difference in execution time of the live model vs the assumption of the simulation. More specifically, I execute my positions in the morning between 6am-8am, as opposed to the simulation that is set for 00:00 every day.

• Compound effects on the calculation of the simulated curve on the backend.

But I don’t yet know for sure as I am writing this, and I’ll have to investigate. My point is that I know I have something to investigate, because I track it. Always make sure that your modeling engine matches your live execution engine!

You can see below a table of exposures for the portfolio I so creatively named “Portfolio 1”. This is the portfolio that generated the equity curve at the start of our post today.

One of the most important things I’ve learned when designing trading infrastructure is that I should always be thinking about scalability and future integration. Always think about how scalable this new feature really is given a wide breadth of future edges, and if it isn’t, design it to be. It will save you a ton of time.

This is it for this week. Just wanted to give you this more practical update on the portfolio. It’s doing really well, on par with the expectations I set for it. We’re still scaling the capital that is allocated to it, and as we get more feedback, the more I’ll allocate. I am also designing more strategies that will be launched soon. I have more ideas on the pipeline than I have time to get to.

I am trying to answer the question:

“If I had to purely live off trading gains each month, could I do it?”

Now, that question explicitly excludes savings, external income, etc. Purely “eat what you kill” kind of approach. I am deeply interested by that question at this moment in my career. That means that my sharpe has to be much higher than it is now. It might be a foolish question, because we know different regimes affect strategies in different ways, and no single strategy makes money forever, all the time.

But I want to question that idea…

I personally know people that do it, why shouldn’t I? There’s short/medium-term alphas/edges that we can run that would indeed increase the consistency of PnL on a month to month basis. I know that I can make money over a few months or a year, that is already established, I’ve been doing that for a long time. But my next evolution as a trader/capital allocator will answer this new question practically.

I hope you join along.

Thanks for reading and wish you a great Sunday.
Pedma

Subscribe now

Disclaimer: The content and information provided by the Trading Research Hub, including all other materials, are for educational and informational purposes only and should not be considered financial advice or a recommendation to buy or sell any type of security or investment. Always conduct your own research and consult with a licensed financial professional before making any investment decisions. Trading and investing can involve significant risk of loss, and you should understand these risks before making any financial decisions.

Following the 10/10 crypto crash, I’ve spent over 220 hours (yeh I track time spent on tasks), re-structuring my entire trading infrastructure with the following purpose:

• Make it scalable enough that I don’t have to be re-designing it every time I want to add a new model/feature, that isn’t traditional.

• Being able to trade across different exchanges so that I can:

  • Harvest rewards programs (points)

  • Spread volume to decrease market impact

  • Spread risk of exchange blow ups and hacks

• Track performance, balances, positions, orders, from a single point of control.

• Developing/researching new models under one unified infrastructure.

• Deploy new models with the same engine that designed them, to ensure that simulation/research environments match reality.

The timing on that couldn’t have been better.

Over the past month alone, Bitcoin is down more than -20% and crypto in general also down -20%. Our new portfolio has been quite helpful in mitigating losses, and not only that, we’ve actually generated a 7.38% return on the portfolio from previous month up to the date I am writing this article.

Crypto has been a tough market to trade directionally this year. You can see that everything is down at least -40% from 2024 highs, and even BTC is down -30% from the this year’s highs.

But as systematic traders we shouldn’t care that much for directionality, as long there’s volatility and liquidity, that’s what we should care for.

As I expand my view on what can be done, there’s so many trades to do, that I didn’t realize that I could do them, or that I was too lazy to go after. I’ve just found a new one a week ago, sent $10k to the exchange, and started harvesting it.

The interesting thing is that what I initially thought was a pure arbitrage trade, it turned out to be something else entirely, and now I have a different edge that I am harvesting.

Just because I was willing to play around, experiment, and observe behavior around what I was doing. Still low sample, but it just goes to show how speed can beat being a perfectionist that never deploys.

This year, as I am scaling my infrastructure, developing new models, new risk tools, new execution engines, going into niche trades I never went before, and putting all of these intro the market, I realized something…

Let me just say that this is my own perspective and experience, and many may not fully relate. But going into the market and testing my ideas was a hard thing to do for a long time. We all know that it can cost money, but I do think it’s a bit deeper than that. Because if I wanted to mitigate cost, I could always have tested it with small size and pay a negligible fee to do so.

Personally, I think I feared for a long time launching new models, because subconsciously, I knew, or thought that I knew, that I couldn’t compete outside what I already did well, or I thought I did well. Every time I picked up some courage to start this massive undertaking, I’d eventually stop myself, and do something else, always under the guise that I was already doing well and should focus on that.

Maybe that’s true, and maybe I should scale what I already know. But being a bit introspective, I do think it’s deeper. I do think that for a long time I was afraid of revealing what my subconscious fear already was…

… of not being able to compete in other forms, and often, more complex trades.

But rather than trying it, I hid behind short-term success. However the problem with hiding behind short-term success, in an industry that requires your peak capacity, is that soon enough, that chink in one’s armor will be revealed. Mine certainly was.

Anyways, now that I am focused on scalability of my trading business, I see so many more ideas, that are worth pursuing. Almost like a new world has opened up, of things that I could do. I’ve always stored trade ideas on a database that I keep to myself. But now, I just see trades everywhere.

Every time I see someone mention a trade idea on social media, or something related to trading, I always ask the question:

“Is there something I can do around what he just said? Even if it’s just facilitating his flow or taking the other side, is there something I can explain fundamentally of why those would be right/wrong?”

I think that when you frame it that way, you start seeing information from the eyes of a trader. From someone that can take information and transform it into a tradable idea.

Just now I saw this post and immediately asked myself the following questions:

• Does retail really love high volatility assets?

• Does it matter if it’s an uptrend or downtrend?

• Why did it happen in 2020-2021? Can you prove it?

• If retail loves it, how can I trade around that effect?

  • Should I sell it to them for a profit?

  • Should I front run them for a profit?

There’s actually a lot of research on this idea, but I’ll leave that to you to find out 😉.

Right now, my problem is not lack of ideas but rather that I just don’t have the time to get to all of them. I have an entire research pipeline with over 200 trade ideas, that I can’t get to, because I am so busy with other models that I am working on.

My main objective for this year and the next, is to scale the diversity of models I trade, so that I can ensure stability across market regimes. To not only be dependent on favorable markets to make money, but rather any market.

Is that a feasible goal? Well maybe that’s the wrong question. Perhaps the question should be:

How much extra return, compared to the market, can I make on average, despite really “bad” regimes?

The answer to that, is that I don’t know yet, I need to trade through those environments again and see if I can beat that goal.

I say the word “again”, because I’ve already survived all the cycles since 2021… but I want something more than just survival, I want to thrive in any given market. That’s why I’ve been putting so many hours behind this one goal.

I think I am getting there, but there’s a lot more to be done.

What Type of Systems Does my Portfolio Hold Now?

Read more

I haven’t posted in a while, but I haven’t forgotten about you. Every day I remind myself I have to write something useful here, and that has been bothering me.

This blog started off as a personal log of my own trading experiences. A place where I could log and document the path of becoming a self-sustainable trader. I want it to be more of that. I don’t want it to be just about long research articles and theoretical stuff. There’s a lot of people out there with far better academic capabilities than me.

Actual risk taking is what I do, and this last month, has been one of the best months in terms of getting my trading infrastructure to the next level.

I’ve been researching and deploying more models, with the purpose of achieving better performance under different regimes. The crypto world is not just trend following and there’s other things I can be doing on top of it—that I have realized.

But that’s a topic for another day.

Today I want to tell you about how I did on the 10th of October—the crypto flash crash.

If you don’t trade crypto, you may not be aware of it, so here’s a quick snapshot of the altcoins performance on that day. In the span of a few hours, the entire crypto market crashed ~50%, taking many out of the game, and leaving most traders with wounds that will take a long time to recover from.

How did my portfolio handle it? What mistakes did I do? What am I doing to fix them, so that they never happen again?

Well, that’s what I’ll get into in today’s post.

I hope that by sharing my experiences, you can avoid similar mistakes, and also forces me to “put in paper” both the mistakes I did, and what changes I am implementing going forward.

10th of October 2025

This day, is a day that I will never forget. It’s one of those days that is either defines your career, or ends it. It will be on the minds of every crypto trader/investor for a long time.

This year’s summer had been particularly difficult for traders that ran the same style of models that I did—trend with a net-long tilt. It was a very choppy summer. However, despite that, by September I was up over 60% ytd at the peak.

But as I said above, it was a very choppy summer, and I ended up going into a ~24% drawdown right before the 10th. I mean, that was normal to me, I expected it. So the day started off as a completely normal day—I was red on that day.

It was night already, I had probably half an hour before going to bed, and I was watching a show on my couch. I glance over at my phone and all is fine. I go back to watching the show, and was almost falling asleep.

Some time passes by, I open my tradingview app, and I see the altcoin index down over -30% and speeding. (now that I think about it, I do check twitter and tradingview perhaps a bit too much…)

That is insane… I think, as I refresh my app thinking it’s an error. Turns out, it wasn’t…

I jump off the couch and rush to the computer to find out what’s going on.

I had a few longs on the long-only trend portfolio, and some discretionary positions that also got closed out. Many of the longs were never meant to be closed out, but they got ADL’d (auto-deleveraging).

ADL is auto-deleveraging, a last resort back-stop mechanism of a crypto exchange to maintain solvency when standard liquidation mechanisms are exhausted. In very simple terms, if the platform can’t find enough liquidity to close bad (losing) positions, the ADL mechanism will forcefully reduce or close positions, even profitable ones, so that the platform doesn’t end up with bad debt.

But nothing compares to my hyperliquid exposure. As you know if you’ve read my performance updates, hyperliquid is a pretty significant exposure I hold, to the point that I think around 50% of my liquid net worth is composed of this one hold. Why do I have such an exposure as a systematic trader you may be wondering? Well… that’s another topic for another day.

But watching it take a 50% cut in just a few hours, is something I’ll never forget.

Imagine your entire net worth get cut in half in a few hours… Yeh, that happened.

The interesting thing to me is that, and I am being completely honest here, I felt nothing.

I was not scared, panicking, or had feelings that would’ve been completely warranted given the amount of exposure I had, and the volatility my entire portfolio was going through.

I was just numb to it. I accepted whatever fate it was stored for me even if this went all to 0. It’s a weird thing to say, but that’s the feeling I had on that day. Obviously I have safeguards, and even if crypto somehow collapsed and stuck there, I’d be fine. Even it did go to 0, I’d still be much better off than just a few years ago. I accepted that risk. I always knew there’s no excess return for free. I traded in the wild west of financial markets for a reason. I was net long for a reason. I was just getting handed that left tail risk I talked about for so long.

Anyways, I checked exchanges, most of them were broken, even those I had discretionary leveraged positions on. I could not close them. I could not even access the apps.

The major exchanges all had issues, there were no bids, and my positions got closed at the absolute worst levels due to ADL. Remember when I said I ran a long-only portfolio? Well, if you did that, you got pretty bad execution when the ADL hit them. If you were long/short, the shorts would’ve closed at the bottom of some of these wicks, but again, I wasn’t.

As I start refreshing the news, the liquidations numbers start pouring in. They’re just out of the charts. By the end of it all, over $20B of liquidations were reported. (The numbers are probably much higher as the major exchanges under-report these figures.)

To put it into perspective, the average move from open to low on that day was -63% across the board.

And we’re not just talking about some small caps on low liquidity. We’re talking about the multi-billion $ cryptocurrencies.

ETH, the second largest crypto asset by market capitalization, went down more than -20% in a few hours.

If we move a few steps lower on the market capitalization scale, and we go to linkchain for example, it went down -64% in a single session from its highs.

I survived it, but I was wounded. I still am. At the lowest point of the drawdown, I was down -48% from my equity peak. Granted, I was already on a 24% drawdown, but it was still massive.

Again, much of that drawdown is explained by my discretionary buy and hold of hyperliquid, but some of it was also derived from the net-long trend model. The market was already weak, so the long exposure wasn’t a lot, but it was enough to do damage. The largest unrealized gain on that day was MNT.

All of these positions, as I mentioned earlier in the post, were exited at very unfavorable prices due to the ADL mechanism of the exchange, which led to larger losses than what would’ve otherwise been expected.

As you can see below, a highly profitable position, got closed out almost at the bottom of the chart when the platform triggered the ADL.

This was one of my major lessons that we will get to later down the post.

On this MNT alone, at one point, I had over $20,000 unrealized, and by the end of it, I lost almost $7k on it due to the ADL.

I was at my desk until late that night processing all of this and refreshing X to see what was going on.

Eventually the market rebounded quite aggressively but the damage had been done—a nasty -26% hit to the portfolio on that day.

Still writing about it feels insane, but considering that I was up over 40x my portfolio from 2 years ago, you got to expect some tail events that will happen. Like we talk a lot about in here, it’s unreasonable to be in a market where the expected returns is so much higher than traditional, established markets, and at the same time, expect the security and safety of the latter.

However I am not trying to throw it under a rug also. I made mistakes, I had inefficiencies. I am fixing them so that next time we’re not going to be so exposed to it as we were then. We might be exposed to something else, but not this.

I always try to be brutally honest with myself. I’ve not done as well as I should’ve. That won’t happen again.

Let’s now get into lessons learned and what I am doing going forward.

Lessons Learned

1) Delta-neutral models look way more attractive than before

After the 10/10 debacle, I gained an increased appreciation for boring, market neutral models. They do look way more attractive to me, than they once were. I was once focused a lot more on absolute returns, and maximizing portfolio growth—that is changing. I am not saying that my book will shift 100% market neutral, but it will trend closer to that. I am at a capital stage that I can afford slower growth, for increased stability, while at the same time doing risky trades on the side. Not my entire book needs to be aggressive risk on. We can chill a bit.

I don’t want to over-romanticize it also, because like anything else, it carries its own risks and tradeoffs. The largest tradeoff being that when the market is hot, these sort of models will underperform in absolute returns compared to other models. Why? Because they’re supposed to have 0 net market exposure, and market carries a lot of performance doing those times. That means practically that when the market is going up, the short legs will carry a negative burden on the performance.

However during lengthy periods of market underperformance, like we’re going through right now, the performances are wildly superior to traditional models with net market exposure, for reasons I won’t get into in today’s post.

2) Being a good trader is a permanent state of research, improvement and iteration

This I knew already, but when you’re making money, you don’t care for it that much.

I’ve developed the long-only trend model in a rush during the summer of 2023 to get some exposure to the crypto market. I wrote a bit about it in this blog post in 12/2023:

At the time, I had a very naive notion of how these models work. Despite that, I am glad I launched it. It carried me forward for the next 2 years, instead of watching it from the sidelines. Whilst at the same time making decent returns during that period.

Had I stacked all my cash in the S&P500, as a standard example, I’d have half the growth that I got from this simple, rough model. Obviously the volatility is much higher, and the risks lurking underneath it, also are… but I didn’t care for that at the time, I cared about growth of the small portfolio I had.

However, let’s be real. It’s been a hard 2 years for this style of trading. We talked about it already a lot, but crypto trend has been really difficult to trade. If you put any trend trader doing the traditional stuff, you’ll see that their performances don’t differ from mine, at least directionally.

However, during this summer, I had already started branching out a lot. I’ve been exploring new models, that are not correlated to what I was doing before, but with the potential of offering diversification benefits to the portfolio. After all, I pay my bills with trading, and I am not constrained by the mandates that large funds are, so there’s no reason that I can’t go where they can’t.

One of these models that was under development was already deployed last week, and is doing its job during a tough market.

Now I am not saying I will be making money every day, week, or even month, but at least I will be adding uncorrelated sources of return, with the goal of smoothing my portfolio growth over time.

One of the main reasons I made money last year was due to my discretionary bets on hype, pump, the “taco” trade whenever Trump nuked the market, idiosyncratic plays that happened here and there, farming protocol airdrops, etc. Many of these are not yet realized, but I have significant airdrops coming up in 2026 due to all of this work.

All of these were and are major contributors to PnL, and that showed me that there’s a path forward that is not only reliant on general market beta, even though a portion of the portfolio can be.

There’s tons of opportunities in crypto for pure alpha. I can compete in there and I will.

3) Short-side is insurance you pay

In traditional trend-following models, the short side tends to not make as much money as the long side. In many cases, it actually loses money over the long haul.

A trader could look at it and think about removing it from their portfolio. Why add a negative burden to the performance, just keep the money maker on right?

Well, if he traded through the 10/10, he’d have found out why it was there. It’s almost as insurance that you pay for each month, but when a black swan event happens, or markets simply collapse, it’s there to help absorb the blow.

Consider the example below. On a long-only trend model, on the 10/10/2025, the model experienced a -16% hit in a single day.

However if we test a simple long/short trend model, it would’ve not only protected the portfolio, but also provide a quite decent return on that day.

(obviously ADL is hard to simulate for because it’s dependent on liquidity at that time on the exchange, but many of the positions on the short side would’ve been closed out near the bottom of the wicks, which is far better than the backtested assumptions…)

So once again, reinforcing the idea that shorts add a necessary market protection, that I will be glad to pay for going forward.

3) Time to press the gas and when not to

Publicly I always say that traders should keep their risk constant, because they tend to overestimate their ability to size positions appropriately, to their perceived edge.

But privately, and in my trading friends group, I am not always 100% adhering to that rule.

There’s times in the market, that when you have some experience, you can see that its time to press size because it’s the most favorable time to do so. But that takes a lot of experience.

Now, I am not saying that I’d increase the risk of my models depending on how I felt about the market. I think that’s a pretty bad precedent. I leave systematic models to be systematic. However, I do carry a portion of my portfolio where I can swing for the fences, let’s say. A portion that if I lost, wouldn’t break me, but enough to scratch that itch when I see an almost too good of an opportunity to pass.

I’ve done that a lot this year, and it served me well.

I wouldn’t advise it to traders starting out, that don’t understand the game yet, and even less playing around with exponential betting size. But there comes a time, that you’re here every day, for years on end, that you can almost sense things that others can’t. And that is something I am taking more advantage off.

Conclusion

The 10/10 showed me risks and flaws I carried, that were hidden away underneath good performance, for far too long. When you get lazy in a high performance field, you will pay the cost for that, eventually. I paid it, and it could’ve been much lower had I stopped to think about it for a second. That won’t happen again.

Many people lost their entire portfolios in that day. More people than we realize because not everyone admits it. I was wounded, but not fatal. I am still here and a bit wiser for it.

This is how we learn. To be exposed to the game, to take the losses and confront them head on. Trading is a pure meritocratic endeavor. You either perform or you don’t, and your PnL reflects that. There’s no way around it, no office politics to carry you, no charisma, nothing. You either improve, or you get left behind.

It’s a ruthless game, but a game I love to play.

Subscribe now

Disclaimer: The content and information provided by the Trading Research Hub, including all other materials, are for educational and informational purposes only and should not be considered financial advice or a recommendation to buy or sell any type of security or investment. Always conduct your own research and consult with a licensed financial professional before making any investment decisions. Trading and investing can involve significant risk of loss, and you should understand these risks before making any financial decisions.

If you’re not a subscriber, here’s what you missed this past month so far:

If you’re not yet a part of our community, subscribe to stay updated with these more of these posts, and to access all our content.

I spend a lot of time on X (formerly known as twitter). Why? Well, there’s a lot of good stuff in there, if you curate your feed well enough. If I spent that time on pure entertainment, or doom scrolling the “for you” page, then I’d argue it’s time wasted, even though I do it sometimes.

But over the years, being deep in the niche of systematic trading, it has shaped a lot of my thinking to how I trade today. I’ve curated my feed so that I mostly get observations from actual practitioners, that provide me with useful information for my research process.

Even if you tried, you can’t observe every little aspect of market behavior yourself. To solve this, you can either hire an army of researchers, that are working on this stuff for you all day (expensive), or go into the places where people with experience participate in sharing their observations. That’s why I am on X most of the times and also why I write these articles.

I store all of the ideas I find in an organized notion database, where I rank them by different attributes, that I consider important of that particular piece of research. Right now, I have 177 ideas I want to research, and that pipeline is growing faster than I can close them.

Anyways I just wanted to share this because a lot of people ask me how I find ideas to work on. The answer is simple. Be online a lot, read a lot, curate the people that you want to research from, that you think are actual traders, and learn everything you can from them.

That’s what I’ve been doing as I’ve been spending a lot of time digging through the content of discretionary and quasi-systematic momentum traders.

I have this hypothesis that a lot of what these traders observe in the market, is actually important information, and if we can gather some data around it and find out if their observations yield value, we can improve our own models.

As I was exploring a different topic this week, I stumbled across the idea of how different stop-losses affect trading strategies in crypto, particularly trend-following strategies which is what I tend to focus my time on.

The performance you see below, is what we have achieved by the end of today’s article. The idea was once again obtained via the observation of good quasi-systematic traders I know on X.

Quite cool isn’t it? In trend/momentum world, this is a super neat equity curve. We’re used to much bumpier rides than this.

The idea that we will explore today is around how tighter stop-losses behave in comparison to looser stop losses with an interesting twist by the end of today’s article.

Read more

The charts don’t show the frustration of going through it on a day to day basis. From the outside, it might look like we had a really strong summer. And true, some coins did really explode. Coins like ETH, XRP, XLM, SOL and others.

But the reality is that in aggregate, the market has been quite stale. If you trade directionally like myself, you experienced the same frustration watching your returns go up a lot, and then, the next few days, give back a decent or even at times the entire returns you just made.

Another plot we can look at to show this behavior, is the % of coins above their daily moving averages.

It’s just a good way to visualize breadth of the market. You can see that when there’s been a spike above short-term SMA’s, the next few days it just retraces back down. However, the % of coins above their long-term SMA, shows more of a sustained uptrend. That gives a false sense that the market is strong, but when we zoom in, it has been chopping around on higher timeframes.

In essence, if you’ve done buy and hold in the top assets, you did well, if you traded short-term (swing trading), you probably chopped around most of the summer.

That’s the game we play right? And we can’t complain. I’ve been spending my time on doing a lot of research and just having fun. But I admit, I am a bit looking forward to the end of summer and start of the last quarter, which historically has been a good period for risk markets.

If you’ve been paying attention to my positions, you’ve noticed that since July I’ve been accumulating PUMP on my portfolio.

Since that first purchase , PUMP is up almost 80%.

I’ve been continuously adding into it, and right now, it is one of my largest concentrated positions. My average is lower than the chart below because Extended (this exchange) had a migration to StarkNet, which meant I had to close my original position and re-open it on the migrated platform.

I am mainly a systematic trader and every decision I take, I always want to factor EV (expected value) into it. But sometimes, there comes trades that are almost too good to pass, but that estimation of value, is very much up for interpretation. However, I don’t think PUMP is. Let’s dig into the reason behind the trade and where I think this could go.

Read more