Thinking in bets: what poker teaches about any decision

At the table, a call is a bet. Your opponent bets 50 into a pot of 100, so you pay 50 for a chance at 200. The equity you need is your call divided by the pot after you call, and the call's expected value is what it wins on average if you could make it many times.

Away from the table, take a shop offering a two-year protection plan for $120 on a new laptop. Suppose, for the sake of the sum, that a covered repair would cost $400 and that 10% of these laptops need one in the first two years. The numbers are invented, but the structure is the same as the call:

The call is worth making because 30% beats the 25% it needs. The plan would need a 30% failure rate to break even, three times the rate assumed here.

Expected value is not the whole story. If losing the laptop would mean missing a month of work you cannot afford, paying more than the average cost to remove that risk can be the right decision. Tournament players meet the same idea on the bubble, where losing a stack costs more than winning one gains.

The call above loses seven times in ten. Each loss feels like a mistake, and a player who judges calls by how they turned out will stop making a call that earns 10 chips on average each time it is made. Poker players call this habit resulting. Psychologists call it outcome bias: in a 1988 study, people rated a surgeon's decision to operate as worse when told the patient died than when told the patient lived, with everything else about the decision the same.

The trainer's lesson what a call is worth on average plays one call out many times, so the wins and losses average into its expected value in front of you.

When your opponent bets, you need to update your estimate of their hand. Bayes' rule gives the answer. Say that before the river you think 30% of your opponent's range is strong. They bet 90% of their strong hands and 20% of their weak ones. Then they bet.

The bet moves the estimate from 30% to 66%: a big shift, but not to certainty, since weak hands bet too. The same sum works for a job candidate who interviews well or a test that comes back positive. The trainer's lesson let the bet reshape the range shows a range narrowing this way, hand by hand.

Putting a number on your confidence lets you check it against what happens. Weather forecasters have been scored this way since 1950, when the meteorologist Glenn Brier proposed what is now the Brier score: for each forecast, square the gap between the chance you gave and what happened (1 if it did, 0 if not), then average. Lower is better.

Imagine 10 days on which it rained 7 times, and three forecasters who each gave the same chance of rain every day:

The forecaster who said 70% matched how often it rained and scores best. Saying 90% sounds more confident and scores worse. A longer record of forecasts can test whether events assigned a 70% chance happen about 70% of the time. Ten invented forecasts illustrate the score; they do not establish reliable calibration.