The largest drawdown is the one about to come: Martí Castany on what a quant actually looks for

Ask ten people in trading what a quant is and you will get ten answers, most of them involving a CV. Martí's is simpler.

A quant, he says, is "someone who tries to find signal in the chaos."

Martí is part of Darwinex Labs, the three-person team that sits behind INDX alongside Javier, one of Darwinex's founders and its CIO. He started out as an electrical engineer, building audio circuits and programming the chips that filter sound. At 21 a friend introduced him to trading. He traded oil futures with two friends, a lawyer and a tennis player, built rudimentary systems and quit university to keep going. He was still going after both of them had burnt out. The bridge from engineering was obvious once he saw it. If a microphone turns a voice into a signal you can process, price is just another time series.

Today his work covers pricing models, high-frequency data pipelines handling billions of data points a day and the infrastructure that gets INDX trades executed at speed. This episode of Fly on the Fund Wall looks at trading from the side most retail traders never see: the people deciding which strategies get capital, and what they look for first.

Know why it makes money, or you will never know why it stops

Martí calls himself an idea-driven trader. He thinks about the strategy first and builds it second. Every set of trading rules, in his framing, is an attempt to reshape the return distribution of the asset you are trading towards something you want: trends, mean reversion, whatever your edge is supposed to capture.

If you understand that logic, you know when your strategy should be making money and when it should not. His own strategy is flat at the moment, and he knows exactly why. There have not been enough price extensions in his markets over the last couple of months. The strategy is behaving as expected.

The alternative is a strategy you found by letting an engine crunch data until something looked good.

"The dangerous thing about this is that when it stops making money, you also don't know why."

Is the edge gone? Was there ever an edge? You cannot answer either question. That is when traders start interfering, touching the code and changing the leverage, at exactly the wrong moment. As Martí puts it, "the largest drawdown is the one that is about to come."

A backtest is one version of reality

"Backtesting is a great tool, but it's like a double-edged sword."

There is only one S&P 500, and everyone is testing on it. The same underlying model could have produced a completely different price series, which means the curve in your backtest is a single draw from many possible ones.

"What you see is just one realisation of reality."

On top of that, it is very easy to overfit. A backtest with methodological faults will not tell you the truth. It will tell you the truth you want to see, and you end up modelling noise.

Kieran shared his own version of this. A gold strategy built in a strategy builder passed every backtest. On paper trading, it one day failed to calculate its stop loss, placed no stop at all and wrecked the account. Every strategy he runs live now carries a max-risk filter that checks P&L on every tick and closes the trade if it breaches a set limit.

Martí's point was that the strategy logic is only one layer. The code that executes it can have bugs. The data feed can drop or corrupt. The infrastructure has to process all of it in real time. He pointed to Knight Capital as the cautionary tale of untested code reaching production.

"You can have a gold bar, but if you cannot sell it or create jewels with it, you will have a very expensive paperweight."

Or, in the phrase Kieran is now sticking to his monitor: garbage in, garbage out.

What professionals do differently

There is no sacred knowledge inside the industry, Martí says. The difference is methodology. In a trading firm, nobody deploys a model the day after building it. There are steps, checks and firewalls before an idea reaches the live pipeline. You also work alongside people who have traded for 30 or 40 years and understand that the most important thing is to stay afloat so you can trade again tomorrow.

Retail traders do not get that network. Martí's advice is to use the tools that now exist. Ask Claude to act as the senior colleague you do not have, and have it design a research pipeline that accounts for overfitting and for how many times you have tested ideas on the same data set. He borrows his statistical approach from drug research.

"It will not be the same, but it's infinitely better than not having it."

Then trade it live, or paper trade it in real time. A backtest compresses five years into seconds. Watching a strategy trade at its own pace shows you things no backtest can.

What INDX watches for

Martí describes INDX as "the proof that what we've been telling the world for the last 12 years is true": that traders at home can beat the market systematically and consistently, and that investors can buy that idea while Darwinex handles the complexity.

His job means looking at track records all day, and he is candid about what he sees. A lot of promising traders have a lifespan of eight to 12 months, and then they blow up.

Once a record reaches 12 to 15 months, patterns start to show in how the curve behaves. He is less interested in how it looks. Does it keep the same character across volatility regimes? For martingale-style strategies, the signal is usually a sharp fall in equity alongside a sharp rise in leverage.

The subtler warning sign is negative skew. A record made of lots of small wins can have one large loss waiting in the tail that erases six of them at once.

"If we see that no big drop has happened yet, it's probably because the left tail has not expressed yet, and this is a warning sign."

A trader with that profile can still pass the INDX checklist, but they get a small allocation. The trader may not have seen the real danger of their own strategy yet, and INDX's job is to protect its investors from finding out the hard way.

This is the section every trader taking outside capital should hear.

"When you're a manager, your product is just not your strategy. It's how much of this strategy you can sell."

His analogy is a shop that makes the best microphone in the world but can only produce two a year. That business does not scale. The same applies to a strategy. If your target on a trade is 10 points and you lose five to slippage because of the capital following you, you have consumed 50% of your edge.

The practical lessons are clear. The longer your timeframe, the more capacity you can usually absorb. Negatively skewed strategies tend to have lower capacity than positively skewed ones. Martí wants traders to design for capacity from the start, which is another reason he favours idea-driven strategies over brute-forced combinations of indicators. On the Darwinex side, the team is working on open-source order-splitting code and on routing flow to futures to lift capacity further.

On the Darwinex risk engine, Martí's view is that VaR standardisation helps most traders. Darwins often look better than the underlying strategy. The only time it works against a trader is when they go "crazy". A martingale keeps adding size to recover a loss, and the risk engine keeps cutting it, so losses are realised before any recovery arrives. Kieran calls it a tug of war. Martí calls it protecting investors.

His own approach says a lot. He invests in his own strategy and lets the engine do its job.

"I know that the risk manager will manage risk even better than I can."

His advice to traders is to stop worrying about the risk engine. Trade your own thing and trade well.

What this means if you are building a record

Nearly everything Martí describes comes back to the same idea: understand what you are running before you trust it with capital. Know why your strategy makes money. Treat your backtest as one possible outcome. Test it live, in real time, before real money depends on it. Think about capacity before capacity becomes the problem. And be honest about your skew, because the people allocating capital will be.

That is exactly what trading on live servers with virtual capital is built for. You get real execution and real fills, and a verified record that shows how your strategy actually behaves.

The full episode of Fly on the Fund Wall with Martí is live and ready to watch:

Thanks for watching, 
Darwinex Zero

*Darwinex Zero and the domain www.darwinexzero.com are trade names used by Tradeslide Technologies, a company registered in the United Kingdom under number 14398381.

The contents of this blog post and video are for educational purposes only and should not be construed as financial and/or investment advice.