Thinking in Bets

A book about reasoning with probabilities and bets.

It did not really work for me. The idea is simple: we focus on the outcome, so if someone misses it is bad luck, and if they hit the target they are brilliant, but we are judging after the fact. In reality, at that moment they are making a bet, with probabilities depending on context, and making a choice. Everything is a bet, and we should not fool ourselves into thinking that anyone who gets results is always skilled. They reasoned well and placed their bet, and the rest is luck. So get used to losing and taking risks, iterating quickly. Much of the rest of the book did not engage me; I found it hard to stay with it, though the personal stories and deeper points made sense. Two and a half stars.

Notes

• Resulting: the basic mistake is judging a decision by the result; a decision can be good and turn out badly (bad luck) or bad and turn out well (good luck).
• “Poker > chess”: life runs on incomplete information and variance; there is almost never a 1:1 link between choice and outcome.
• Decision making = process, not verdict: separate “how I reasoned” from “how it went”.
• Ask: what alternatives did I have? what signals did I have at that moment? what probabilities would I have assigned ex ante? Example: a pitch in baseball that people call a bad choice, but that is easy to say after the fact!
• Train calibration: express confidence in percentages (e.g., 60/40) to force nuance and reduce the 0/100% mindset.
• Proper post-mortem: analyze outcomes without searching for a convenient story; a single case is often too little to really “learn”.
• Sample size: 4 coin flips tell you almost nothing; same for our experiences: you need to accumulate evidence. What do you know about that coin? Unless you can simulate two thousand flips (not feasible in reality).
• Bucket “skill vs luck”: attribute what is under control and what is not; avoid both self-absolution (“bad luck”) and self-celebration (“genius”) a priori. Example: politicians (I am skilled or unlucky; they are bad or lucky).
• Feedback loop: outcome -> update beliefs -> future decisions; but with an anti-hindsight-bias filter.
• Accountability: make your bet explicit before the outcome (write it / share it) to avoid rewriting history later.
• Decision group: engage with people who are not clones to reduce echo chambers and confirmatory thinking.
• CUDOS-like norms: bring data/context in full and ask questions to surface the missing info.
• “Alternative futures”: every choice is a bet among future versions of us; useful to think in opportunity costs.
• Practical application: when I evaluate an event (work/relationship/health), distinguish: (1) quality of reasoning, (2) execution, (3) luck.