What Does Poker Know About Deciding Under Uncertainty?
From the Think Like Poker Players collection
When the scoreboard grades luck
A product launch beats its targets, and the team that approved it is celebrated. A second launch, planned with equal care, runs into a supply shock and misses, and its sponsors spend the next quarter explaining themselves. Both verdicts feel fair, because results are visible and reasoning is hard to see. Across many such verdicts, a team learns to repeat whatever happened to win, including the parts that won by accident.
The difficulty is structural. Most consequential decisions, from hiring and investing to launching and negotiating, are made with part of the picture missing, against other parties with plans of their own, and with outcomes that blend skill and luck in proportions no single result can reveal. Judgment under those conditions needs a discipline of its own.
Poker supplies one. The game combines hidden information, deliberate deception, money at stake, and feedback blunt enough that self-deception becomes expensive within months. John von Neumann and Oskar Morgenstern analyzed simplified versions of it in Theory of Games and Economic Behavior in 1944, because concealed cards and bluffing sit at its center.
Professional players then spent decades refining, under financial pressure, a set of habits for deciding well when certainty is unavailable. Those habits transfer intact to business, careers, investing, and negotiation, and they fit together into a single framework.
Decision quality and outcome quality
The foundation is a separation of two measures. Decision quality depends on the information, reasoning, and odds available at the moment of choice. Outcome quality adds luck. A sound decision can lose and a careless one can win, and one result reveals very little about which was which.
Poker players call the habit of collapsing the two "resulting," a term Annie Duke carried from the card room into business writing with Thinking in Bets. Naming the bias makes it inspectable.
A review that asks what was knowable at the time, and how well it was weighed, grades the decision; the scoreboard grades the luck. Skill shows up only in the pattern across many decisions, which players call the long run.
Ranges and expected value
Two ledgers raise an immediate question: if outcomes are noisy, what makes a decision good? The poker answer is a habit of reasoning in ranges. Facing an opponent's bet, a strong player asks which hands that opponent could hold and how likely each one is, then acts on the whole distribution.
The same move turns a sales forecast from a single figure into a spread: a one-in-five chance of falling short of plan, a three-in-five chance of landing near it, a one-in-five chance of beating it. Each new signal then reshapes the spread in proportion to how much it reveals.
Ranges make expected value computable. The question "if this choice were made a thousand times, would it come out ahead?" weighs each outcome by its probability alongside the best and worst cases.
A $50,000 pilot that could win a contract worth $1 million in profit pays for itself on average whenever the chance of winning exceeds about 5 percent. Framed this way, an argument about whether a project feels risky becomes a comparison of price against payoff that a team can examine, dispute, and improve.
Calibration and bet sizing
Expected value is only as good as the probabilities fed into it, which is why serious players work on calibration, the match between stated confidence and the observed rate of being right. A calibrated forecaster who says 70 percent turns out right about 70 percent of the time. Philip Tetlock's research on expert forecasting found that calibration varies widely and improves with tracked predictions and honest scoring. The precision pays directly, since being 60 percent sure and being 85 percent sure call for very different commitments.
A positive expected value still carries a hazard that poker players learn early and often painfully. A strategy with a genuine edge can go broke before the edge pays, if each bet is large relative to the reserves behind it. The probability of that outcome is risk of ruin, and bankroll management exists to hold it near zero.
In 1956, John Kelly, a researcher at Bell Labs, published a rule for the bet size that maximizes long-run growth: with a 55 percent chance of winning an even-money bet, the growth-maximizing stake is 10 percent of the bankroll. Larger stakes grow more slowly and court ruin. Many professionals bet a fraction of the Kelly amount, because their estimate of the edge is itself uncertain. The same principle governs a startup's runway, an investor's position sizes, and a career move: size each commitment so that losing it leaves the player in the game.
Tilt and the mental game
Sizing rules hold only as long as the person applying them holds steady, and here poker contributes its most practical vocabulary. Tilt is the state in which frustration, fear, or pride begins to rewrite the criteria for decisions. It rarely looks dramatic. Standards loosen, marginal bets start to feel justified, and the shift happens without any conscious choice to change direction.
Jared Tendler, a mental-game coach who works with professional players, describes performance as a spectrum from an A-game to a C-game and argues that raising the worst version of play often pays more than polishing the best. A manager who loses a major client at noon and then approves, at six in the evening, a deal rejected at nine that morning is playing a C-game. The remedy is procedural: decision rules written down in advance, live choices compared against them, and a narrower playbook or a pause whenever the gap opens.
Balance and exploitation
The final layer concerns opponents who are also thinking. Game theory offers balanced strategies, mixtures of strong and weak holdings behind the same actions, that no opponent can systematically exploit. In 2017, Libratus, a program from Carnegie Mellon University, beat four top professionals over 120,000 hands of heads-up no-limit hold'em with play built on that logic, and in 2019 Pluribus extended the result to six-player games.
Balance is the safe default against strong or unknown opponents. Against a specific, readable weakness, exploitative play earns more by deviating from balance, at the price of becoming exploitable in turn. A firm that discounts only when inventory is high has taught its rivals to read its prices; a firm that mixes its signals keeps them guessing; a firm facing one predictable competitor can price directly against that competitor's habits.
One framework for uncertainty
Taken together, these habits form one framework: separate the decision from the outcome, so that learning tracks judgment. Reason in ranges and expected value, with probabilities calibrated against a track record. Size every commitment for survival, so an edge has time to pay. Guard the decision criteria themselves against emotional drift, and choose between balance and exploitation according to what the other side reveals.
Each element is modest on its own. Combined, they describe a way of acting well when certainty never arrives, a condition that covers most decisions worth making. Poker earned these habits under unusually honest feedback, where a flawed method surfaces in the bankroll. Any field that runs on partial information and probabilistic results can adopt them without ever touching a card.