Trading and Exchanges Chapter 22: Predicting Performance and Why Most Traders Fail (Part 2)
This is the part of the book where Harris basically tells you that almost everything you think you know about picking winning traders is wrong. If Part 1 was about how hard it is to evaluate past performance, Part 2 is about why predicting future performance is nearly hopeless. And honestly, it is one of the most important chapters in the entire book.
Is Warren Buffett Skilled or Lucky?
Harris asks this directly. Berkshire Hathaway appreciated 2,078-fold between 1965 and 2000. That is 23.6% compounded annually versus 11.8% for the S&P 500. Obvious skill, right?
Here is the catch. Buffett came to our attention only because he had amazing returns. If his performance had been average, nobody would write about him. So the proper question is not “what are the odds of this by chance?” but “out of 10,000 managers starting in 1965, what are the odds the best one would do this well purely by luck?”
Answer: about 0.5% for 10,000 managers, 5% for 100,000. Harris concludes Buffett very likely is skilled. But a standard t-test says probability of luck is 0.001%. The correct test says up to 5%. Massive difference just from asking the right question.
Then there is regression to the mean. In his first 26 years, Buffett beat the market by 13.2% annually. In the next 10 years, only 6.84%. Still great, but literally half. Harris suggests the earlier period included a lot of luck on top of genuine skill.
The Peso Problem: How to Fake Skill
Harris describes strategies that look brilliant until they blow up. The “peso problem” is named after Mexican currency. The government fixed the exchange rate, so investing in Mexican debt always beat US debt. Until the government devalued the peso and you lost everything overnight.
In trading, this shows up when managers sell out-of-the-money options on top of an index portfolio. Most years you collect small premiums and beat the market. Your t-statistics look great. Then the market crashes, options get exercised, catastrophic loss. Harris calls this an “informationless trading strategy.” Zero skill required. But until the crash, statistics say you are highly skilled.
There is also the doubling-down strategy. Lose a bet? Double it. Keep doubling until you win. You will almost always end up ahead. Until you go bankrupt. Harris says undisciplined managers love this because it hides bad decisions until the blowup.
Survivorship Bias: The Elephant in the Room
Survivorship bias happens when you only see the winners because the losers have been removed from your data.
Mutual fund companies do this constantly. They start many funds. Keep the winners. Kill the losers by merging them into better-performing funds. Then show you marketing materials featuring only “proven” track records. The average performance of all funds, including the dead ones, might be negative. But you never see the dead ones.
Harris has a brilliant example he calls “My Favorite Fraud.” Take 20,480 people. Send half a prediction the market goes up, half that it goes down. Send the correct recipients another split prediction. After 10 rounds, 20 people have watched you correctly predict the market 10 months in a row. They think you are a prophet. Pure math, zero skill.
The commodity pool data is even worse. Advisers who launched public funds had average monthly returns of 4.1% in the 36 months before going public. After going public? 0.23% per month. They were not skilled. They were selected for having been lucky.
Why Our Brains Fail Us
Harris makes an evolutionary argument. Our ancestors survived by connecting events to causes. Useful when a rustling bush might be a tiger. Terrible for financial markets. We are hardwired to believe good performance must have a cause, and that cause must be skill.
We also selectively remember. For others, we assume success is skill-based. For ourselves, wins are skill and losses are bad luck. Both habits lead to terrible investment decisions.
The Real Way to Predict Performance
Harris says forget past returns. The only reliable way to evaluate a trader is to assess their comparative advantage. Not absolute advantage. Comparative.
A marathoner who runs 2:20 is incredibly fast. But that time would place 36th at the Olympics. Being good is not enough. You need to be better than your competition. In a zero-sum game, this distinction is everything.
Harris lists factors that predict performance: intelligence, experience, discipline, access to data. But the most important factor? Whether the manager understands they need a comparative advantage at all. If a manager cannot explain why other traders will lose to them, run.
The Uncomfortable Bottom Line
Harris states it plainly: predicting future performance from past returns is “an essentially worthless activity.” Luck dominates skill over human time horizons. The luckiest managers will outperform almost all genuinely skilled ones. Sample selection bias means you only hear about winners. Your brain is wired to draw the wrong conclusions from all of this.
If you are going to trade actively, you better have a very clear reason why you will beat the other people at the table. “I am smart” is not enough. You need to be smarter, faster, or better-resourced than the people on the other side of your trades. If you cannot honestly say that, just buy the index.
That is the hardest sentence in the whole book for most people to accept.