Fast and Slow at the Trading Desk

Two systems for thinking, one of them is taking your trades, and a protocol for putting the other one back in charge.

---

There is a study, run by Daniel Kahneman and his colleagues in the 1970s, that I want to tell you about because it explains almost every bad trade taken by an experienced trader who "should have known better."

The setup is simple. Researchers gave a group of expert decision-makers (in the original study, doctors and statisticians) a series of medical case files and asked them to estimate the probability of a particular diagnosis. The cases were carefully constructed so that the right answer required Bayesian reasoning: weighing the base rate of the disease in the population against the accuracy of the test result.

The experts knew Bayes' theorem. They had used it in publications. If you asked them, in the abstract, "how should one combine a 1% base rate with a 90% accurate test?", they could write down the equation and get the right answer in fifteen seconds.

But when they were given the case file and asked for a quick clinical judgment, their answers diverged from the math by an enormous margin. They were three to ten times more confident in the diagnosis than the math justified. The same people. The same equation. Two completely different answers, depending on whether they were thinking or responding.

Kahneman and his collaborator Amos Tversky spent the next forty years documenting this kind of split, in domain after domain. Doctors. Lawyers. Pilots. Investors. Generals. The pattern was almost perfectly consistent: experts knew the right rule, could state the right rule, and routinely violated the right rule when they were operating in real time under any kind of pressure.