The Graveyard of My Cleverest Ideas

Early on, I had a folder full of trading strategies I had invented — a dozen of them, maybe more. Each one was a small theory about how the market worked, carefully turned into code, carefully scored. I was proud of that folder.

Today every strategy in it is dead. I buried the entire collection, and burying it turned out to be one of the more useful things I did in the whole project.

A zoo of theories

The work felt wonderful while I was doing it. I would take an intuition about the market — a sense that prices tend to behave a certain way in a certain situation — and encode it as an explicit strategy. One for when things seemed to be trending. One for quiet, range-bound stretches. One for sharp reversals. Then I would run them against the data and score them, keeping a kind of leaderboard, watching which of my ideas was currently winning.

It was, in its way, deeply satisfying. Each strategy was a piece of my own thinking made executable — a belief about the world that I could now test. And there was always a number next to each one, always a ranking, always a sense of standings to check.

Why building them felt like progress

This is the part worth being honest about, because the feeling was the trap. Producing strategies felt productive in the most literal sense: I was producing things, and the things had scores, and the scores moved. Every session ended with a fuller folder and a reshuffled leaderboard. By every signal my brain uses to detect work, I was working hard and getting somewhere.

What I was not doing was asking whether the entire activity meant anything.

What the leaderboard was really measuring

Because here is what the rankings actually reflected. Each strategy was being scored on the same fixed slice of history it had been shaped against. So the one that climbed to the top of my leaderboard was not necessarily the strategy that best understood the market. It was, more often, the strategy whose particular shape happened to line up most closely with the accidents of that specific past.

I had built, without quite realizing it, a machine that rewarded overfitting and handed out trophies for it. The leaderboard felt like research. It was closer to a contest to see which of my guesses had gotten luckiest against one stretch of data.

The deeper problem: every strategy was a frozen guess

Underneath the tournament was a more basic mistake. Every hand-built rule was me taking an intuition — the market tends to do this after that — and freezing it into the system as if it were a fact. But my intuitions about a noisy, crowded, adversarial market were mostly just that: guesses, and not especially reliable ones. And freezing a wrong guess into code does not make it less wrong. It mainly makes it harder to notice, because now it is buried in a strategy with a name and a score, wearing the costume of a tested idea.

A folder full of frozen guesses is not a body of knowledge. It is a collection of my own opinions, formatted to look like findings.

Burying the whole collection

Eventually I did the thing that felt, at the time, like giving up. I deleted them. Not refactored, not set aside for later — deleted, the whole zoo. The folder of my cleverest ideas became an empty folder. It was genuinely uncomfortable; those strategies were mine, each one a little artifact of my effort and my pride.

But keeping them was costing more than burying them. As long as they existed, they invited me to keep tuning them, keep re-running the tournament, keep mistaking the motion for progress. Deleting them closed that whole false avenue and pushed me toward a different question.

Why the graveyard itself was the lesson

The value, it turned out, was never in any individual strategy. It was in the shape of the whole graveyard: a dozen confident theories, each carefully built, and not one of them backed by validation I could trust. That was the finding. The pattern was telling me something I had been working hard not to hear — that my intuitions were not, by themselves, an edge, and that being clever about a market and being right about it are not the same thing.

One dead strategy is a disappointment. A graveyard of them, all built by the same hand, is information about the hand.

What I kept instead of strategies

So I stopped trying to be the source of the edge. Instead of encoding my guesses as rules and ranking them, I tried to build something that could look for structure on its own — to take my intuitions out of the part of the system that actually decides, and let out-of-sample evidence, rather than my opinion, do more of the deciding — while treating my own modeling choices as assumptions to be tested, not facts. I have written elsewhere about removing the hand-written rules from the core of the system that came next; this graveyard is where that conviction was born.

The strategies all died. The thing that survived was smaller and more valuable than any of them: a durable distrust of my own cleverness, and a preference for letting the data disagree with me. That has been worth more than the entire folder ever was.

— No signals, no returns, not investment advice.