david einhorn has run greenlight since 1996 and compounded at 12.7% net on more than $3 billion. he was asked what would surprise someone who sat beside him for a month watching every decision: "I think you'd be surprised how few decisions I actually make. my ratio between trading decisions and naps is much higher towards naps than people would expect." "when the data checks out but your gut says no I make the more conservative decision. [and in reverse] again, I'll choose the more conservative decision, whichever that is, if I'm conflicted."

The most expensive thing in systematic trading is often not a losing strategy. It's the two years you spent not being in the market. Everyone underestimates this because time doesn't show up anywhere in your P&L. A blown account is visible and painful and you learn from it fast. Three years spent building a custom machine learning backtest engine, a feature pipeline and an ML framework before you've placed a single live trade costs you *far* more, and the damage is completely invisible. And the cruel part is that the building genuinely feels like progress. You're learning, you're solving real problems, you're getting better at something. But you're getting feedback about your code, not about the market. You only get it by being on the field. Meanwhile the thing you were going to build usually already exists in a simpler form for almost nothing. If you want trend, start with a basic 12 month lookback and a monthly rebalance. That's your first sleeve done in an afternoon, live, compounding, teaching you things. Then every afternoon after that goes into the second sleeve, and the third. Five boring strategies stacked together will beat the one brilliant one you're still building, because their drawdowns land in different places. So before you build anything, ask what you can do *right now*, with the tools you already have, that is likely to make money. Do that this week. Complexity comes later, if the simple version genuinely can't harness the edge.

dan sundheim's fund returned 60% in 2020, one of the best years on wall street. one month later it lost $4 billion, about 20% of its capital, in january alone. it kept falling into 2022. this is what he told his investors at dinner three days after the bottom: "the trough of our drawdown was at the end of may 2022, and these dinners were scheduled for june 3rd. jeremy, the president of our firm, said to me, we can't do these dinners, this is going to be a bloodbath." "the message was that we were going to do things differently. not the stock selection, all of that was going to be the same, but the portfolio construction was going to be done in a way that was much less risk-prone. the analogy I gave was, we're going to hit singles and doubles. it might take us longer to get back to the high water mark." "even if the right positive npv thing would be to just keep taking a ton of risk, and obviously usually the best time to take a ton of risk is when you've lost a lot of money, emotionally I would not be able to go through this again. so we said, look, we're going to run the business differently."

Larry Williams turned $10k into $1.1M to win the 1987 World Cup trading championship. His single most important market belief: "I believe the current trade I am in will be a loser... a big loser at that."

imc trading made $3.1 billion in net trading revenue last year. they also run a global trading competition that 48,000 people have entered. this is what their own traders tell entrants: "one thing we sometimes said was: think twice, program once. not think once, program twice." "you want to think deeply about the things that you're doing. collaborate. and then coding up the thing that you find is actually quite simple usually."

this is what a semiconductor options desk actually does all day. a former imc trader explains how they made money off asian banks that had no choice but to trade: "in asia a lot of banks offer structured products, very complicated derivatives, to retail clients. people buy those products and now the banks are short them. to hedge, they trade on the options markets." "as a market maker you can buy data on the product issuance of banks, and based on their issuance and the current underlying stock, you can estimate what their risks are and what risks they would need to offload onto the general market." "I can make the judgment that it's probably good to trade against these people. not because they're stupid, but because they have to hedge away this risk. it's just good expectancy to buy based on the fact that you're trading against someone who has to." from "Odds on Open"

your backtest is probably the grey line. grey is the strategy with no fees and no slippage: up a ton. blue is the exact same trades paying actual costs: down 47%. nothing about the signal changed between them.

jim simons built the most profitable hedge fund in history. this is him at mit in december 2010, a year after he retired. 15 years on it is still the video I go back to most. he walks through why he stopped trading on judgment and went 100% models in 1988, and what he thought the actual secret was. the line that stuck with me: "if you're going to trade using models, you just slavishly use the models. you do whatever the hell it says, no matter how smart or dumb you might think it is at that moment." clipped my favourite parts here.

cyrus omar leads a lab at michigan studying how AI agents write code. speaking at jane street, he described what happens when you point a swarm of them at one repo with no coordination: "it's often not built for thousands of agents editing a codebase simultaneously. questions like, who else is editing this right now, are difficult to answer with git." "they end up in this loop of editing the same thing, then resolving conflicts in a conflicting way, and then having to resolve the conflicts between the conflict resolutions. you can actually end up spending 50k in tokens trying to build something because of that." "there was a project where they tried to build a verified compiler from javascript to web assembly. without a ton of coordination between the agents, they all just picked the lowest hanging fruit all the time. they all conflicted with one another in minor stylistic ways, and you ended up spending a lot of tokens with not a lot of output." parallel agents do not divide up work on their own, they often all reach for the same easy piece and then spend your budget arguing about it.

brian kelly, ex-crypto hedge fund manager, now running a trading firm where every employee is an AI agent. his prediction for what a hedge fund looks like next: "what I think wall street looks like is one single hedge fund manager working with maybe 1 or 2 people under them, and then a swarm of agents under that." "but my job as the human being is to be one step ahead of it, on creativity. so far AI agents haven't been able to do that. so they're much more of a tool in the whole process rather than a replacement of it." he says people at big funds told him they are building their own versions. no results published on his side yet. guess we will have to check back in the future.

the bloomberg terminal took three years to build before one customer paid for it. michael bloomberg, now worth $109 billion, says that gap is the reason nobody ever caught up: "there was a need. nobody really thought that having that computer in the middle was useful. and by the time it was done, which was three years before we really had a customer, it was too late for anybody to catch up." "since then there are plenty of people who have good products, but they only do one or two of the 10 or 20 things that we do." "we just focused on building the best thing we could, and not necessarily always listening to what the customers want. I've always asked, and customers can be very valuable, but that doesn't mean they are able to think in terms of what we could do with a computer." three years of no revenue looks identical to failure while you are inside it. takes a lot of grit to continue.

rob carver ran a multi-billion dollar portfolio at Man AHL and has written five books on systematic trading. his rule for anyone who wants to use discretion: you get one decision. "you have to limit the discretionary part of a trading strategy to one tiny part, which is the what should I buy and when decision. every other part should be based on a scientific, systematic, rule-based approach. how fast you trade, when to close positions, and how big positions should be." "if taking poor buy decisions was the only mistake retail traders made, you'd expect roughly half to lose money and half to make money, because worst case they're flipping a coin. but the real number is more like 95% of retail traders lose money." "that's because they're making errors with position and cost and risk management. and that's where there is no room for discretion. there's no art involved in that at all. that is science." entries are the part most people argue about. but sizing, costs and exits are where the 95% actually happens.

before the $109 billion, michael bloomberg spent his first summer on wall street sitting in his underwear in an unairconditioned bank vault, counting securities by hand: "we slaved in our underwear in an unairconditioned bank with an occasional six-pack of beer to make it more bearable." "my friends asked what I was doing for work. to save face, I told them I was studying methods and procedures to simplify workflows. my friends were research analysts and investment bankers with lush private offices, and I was what can only be called a clerk. why didn't I quit? I was too embarrassed." amazing episode on michael bloomberg from founders podcast.

cliff asness would take a drawdown 50% deeper if it ended in six months instead of three years. he runs $187 billion and has been in the seat since 1998: "I've never seen a model that looks at pain. the negative utility of losing money in terms of how long you've lost money for, not just magnitude." "in real life a drawdown that is one and a half times bigger but with six months instead of three years is ridiculously easier to live through." "you go back after six months. you get a lot of sympathy. you go back six months later, and six months later. eventually they just think maybe you're a dinosaur, maybe you just don't get this new wave."