What We're Reading Now

Dicey Decisions

18 August 2026

Geof has been reading Thinking in Bets: Making Smarter Decisions When You Don't Have All the Facts by Annie Duke, a practical guide to better decision making when the future is uncertain and outcomes are shaped by both skill and luck.

Tags: geof read, learning, strategy

In her book Thinking in Bets: Making Smarter Decisions When You Don't Have All the Facts, Annie Duke described how every meaningful decision is a bet. We choose one future while giving up others, committing resources to what we believe is most likely to happen. That framing changes the way we think about judgment and decision-making, especially in complexity. The question is not whether we can eliminate uncertainty, but whether we can make better choices while uncertainty remains.

Duke parlayed (get it) her expertise as one of the top professional poker players to explore how we make decisions under pressure. In poker, like complex challenges, we act with incomplete information, hidden variables, and luck. A good decision can still lead to a bad outcome, and a bad decision can sometimes get rewarded.

When we stop judging ourselves only by outcomes, we can start improving the quality of our thinking. Better decisions come from better calibration of our beliefs by pausing to understand what we know, what we don’t know, what alternatives exist, and what risks we are accepting. Duke used poker and other games of chance to illustrate how our cognitive biases and beliefs affect decision-making, distinguishing between outcomes and process, as well as offering tools to help us calibrate our beliefs and make better decisions.

Beliefs Are Bets, Too

We make decisions based on what we believe about the world. The problem is that beliefs often form quickly and casually. We hear something, accept it, and only later (if ever) examine whether it is true. Once a belief takes hold, motivated reasoning protects it. We look for confirming evidence, discount information that contradicts us, and interpret ambiguity in ways that preserve our existing beliefs. Blind spot bias makes this even harder because we tend to see bias in others more clearly than we see it in ourselves.

One useful question that Duke often heard around the poker table cuts through our false certainty: “Wanna bet?” The point is not to gamble; it is to reveal that every belief already carries risk. If we had to put something meaningful on the line (bet), we would ask: How do I know this belief? What is the quality of my evidence? What alternative beliefs could be true?

Better belief calibration is not a one-time correction. It is a habit of treating beliefs as provisional assumptions, exposing them to disconfirming information, and noticing when our confidence runs ahead of evidence.

One of the most damaging bias traps in poker is “resulting,” or judging a decision solely by how it turned out. If the result was good, we assume the decision was good and we are expert decision makers. If the result was bad, we assume the decision was bad or luck was against us. But outcomes are shaped by both skill and luck.

Hindsight bias compounds the problem. Once we know what happened (the result of our decision), the outcome feels inevitable. Duke introduced backcasting and premortems as approaches to calibrate our decision making before the outcome.

Plan for Multiple Futures: Scenario Planning

Backcasting, a type of scenario planning, starts with success and works backward: if this decision went well, what must have happened? Imagining the steps to success helps us better plan for and create the conditions that increase the likelihood of success.

A premortem starts with failure and works backward: if this failed, what likely went wrong? This helps us look for disconfirming data and create a more realistic picture of both opportunity and risk. It gives us permission to safely surface assumptions in our plans.

We can also create Ulysses contracts: structured precommitments, like Odysseus having his men secure him to the ship’s mast as they sailed past the Sirens. Precommitments are designed by our more rational “past self” to protect our more impulsive present self. These safeguards create “stop and think” moments before emotional reactions, fatigue, or pressure take over.

In complexity, and in gambling, no single future is certain. Instead, there are multiple potential outcomes. We can identify those possible outcomes by mapping possible futures and asking what would have to be true for each one to happen. Scenario planning makes uncertainty explicit, which prepares us to respond instead of react.

Learning Means Separating Skill from Luck

The outcomes of our decisions, both favorable and unfavorable, are feedback, but feedback only helps when we interpret it carefully. Duke introduces a simple learning loop: belief, bet, outcome, updated belief. The hard part is deciding what the outcome teaches us. That requires sorting results into skill, luck, or some combination of the two. Good outcomes can hide bad reasoning. Bad outcomes can contain strong decisions. However, we tend to claim our wins as skill and our losses as luck. When we fail to sort results and look for feedback in our decision-making process, we reinforce our existing beliefs.

The goal is to care more about missing a learning opportunity than about admitting a mistake. That psychological safety makes us less defensive and more curious. It also allows us to find errors within success and randomness within failure.

The Traps That Keep Us from Learning

Duke introduced several bias traps that protect short-term emotional comfort at the expense of long-term learning. The traps can make decisions feel safer by creating the impression of certainty. Naming the trap creates distance from it and makes it easier to return to a more objective mindset.

• All-or-nothing thinking: Treating beliefs or decisions as either completely right or completely wrong, leaving no room for calibration.
• Motivated reasoning: Interpreting information to support existing beliefs rather than updating beliefs to fit new information.
• Self-serving bias: Crediting good outcomes to skill and bad outcomes to luck.
• Hindsight bias: Seeing what happened as inevitable once the outcome is known.
• Resulting: Judging decision quality solely by outcome quality.
• Tilt: Letting negative outcomes create emotional reactions that compromise future decision-making.

Communicate in Ways That Reduce Defensiveness

Duke also introduces “truth-seeking” communication, which works to combat our own biases and lower the reputational risks of recalibrating decisions as new information emerges.

• Express uncertainty: Stating confidence levels invites others to contribute without needing to confront us. Expressing uncertainty signals collaboration.
• Lead with assent: Start with what you agree with before adding dissenting information. Leading with assent shows that disagreement does not require rejection.
• Use “and” instead of “but”: “And” adds to the conversation; “but” can feel like denial or repudiation.
• Ask for agreement to truth-seek: Some people want to vent instead of analyze; invite the right kind of conversation.
• Focus on the future: Validate the past, then shift toward what could be done differently going forward.

Bring Your Future Self into the Decision

Even when we know better, current emotions can overwhelm future interests. We overreact to recent results, chase relief, or avoid discomfort. To mitigate this, Duke suggested tools that help widen the time horizon and make the future self more present.

One simple tool is 10-10-10: What are the consequences of this decision in ten minutes, ten months, and ten years? This question slows down emotional decision-making and helps us see whether our immediate feelings deserve as much influence as they are getting.

Future-self thinking also helps with regret. If we imagine potential regrets before deciding, we can use this as information rather than punishment. And if an outcome goes badly despite a sound process, we can respond with more compassion because we know we acted thoughtfully amid uncertainty.

The Bottom Line

Thinking in bets is not about becoming detached, robotic, or perfectly rational. It is about becoming more honest, or “truth-seeking,” as Duke called it. Honest about uncertainty. Honest about the limits of our knowledge. Honest about the fact that good decisions do not always win and bad decisions do not always lose. The payoff is better learning. When we separate process from outcome, express uncertainty, seek dissent, and plan for multiple futures, we improve the odds that our next decision will be wiser than our last. In complexity, we will never have all the facts. But we can still make better bets.

« Return to What We're Reading Now