Thinking in Bets: Better Decisions Under Uncertainty
Summary of the video “Thinking in Bets | Annie Duke | Talks at Google” by Talks at Google.
Annie Duke, a former professional poker player and cognitive scientist, explains how treating every decision as a probabilistic bet—rather than a certainty—helps overcome cognitive biases, tribal thinking, and the tendency to conflate decision quality with outcomes. She offers practical strategies like forming truth-seeking groups, deconstructing decisions before outcomes occur, and using Ulysses contracts to make better choices in life, business, and relationships.
Why We Make Bad Decisions
Uncertainty comes from two sources
Hidden information (like face-down poker cards or unknowns in life) and luck (randomness that affects outcomes even with perfect information). Most decisions are one-trial scenarios where we can't gather enough data to be certain, yet we must decide anyway.
Evolution built us to make false positives
Our brains are wired to connect patterns and see causality much more often than it actually exists. On the savanna, it was safer to run from rustling reeds (false positive) than to investigate and risk a lion attack (false negative).
We are built for tribes, which distorts information processing
Tribal membership gives us belonging, distinctiveness, and moral guidance, but also epistemic closure—the tribe tells us what knowledge is trustworthy. This causes us to dismiss information from outside the group.
We affirm our priors rather than disconfirm them
We process incoming information to make ourselves feel right about what we already believe. The uncertainty in data gives us room to interpret it in ways that confirm our existing views.
Motivated reasoning is larger than confirmation bias
We don't just notice confirming information and ignore disconfirming information; we actively work hard to discredit information that disagrees with us while accepting confirming information without scrutiny. A politically agreeable article feels smart; a disagreeing one feels flawed.
Echo chambers vs. epistemic bubbles
An epistemic bubble is lacking information; exposure to it helps. An echo chamber is actively distrusting sources outside the group regardless of information. Most people live in echo chambers, where showing someone facts doesn't change their mind because they don't trust the source.
Apply universalism to separate source from truth
Whether Copernicus, Mussolini, or your mom says the Earth is round, the roundness is independent of the source. Imagine information delivered by a trusted source, or absent of any source, to evaluate it objectively rather than dismissing it based on who said it.
Thinking in Bets: The Core Framework
Every decision is a bet
When you decide, you're making an implicit bet informed by your beliefs about possible futures. You're investing resources (time, money, health, happiness) betting that your choice will lead to the best outcomes compared to alternatives. Recognizing this makes decisions more explicit and open-minded.
John Hennigan's Des Moines bet illustrates everyday decisions
Poker players bet $30,000 that Hennigan couldn't stay in Des Moines for 30 days. He moved, called after two days claiming victory, offered to settle for $15,000, then returned to Las Vegas and lost $15,000. This mirrors any major life decision: weighing unknowns, comparing futures, forgoing alternatives. Your job choice, where you move, whom you marry—all are bets with uncertain outcomes.
Framing decisions as bets forces probabilistic thinking
When someone asks 'Do you want to bet on that?' about a claim like 'Democrats will win the House,' it shifts you from certainty into probabilistic thinking. You start considering time, intervening events, luck, and what the challenger knows that you don't. This bubbles up uncertainty and opens your mind.
Probabilistic thinking drives information hunger
Once you accept your beliefs and predictions are probabilistic, not certain, you need to refine them because you're betting limited resources on them. This makes you open-minded to disagreeing opinions—you already know why you're right, so you need to learn why you might be wrong.
Resulting: The Gravity Well of Outcomes
Resulting distorts ability to analyze decision quality
Once you know the outcome, it acts like a gravity well, pulling your judgment toward the result. A bad outcome feels like a bad decision; a good outcome feels brilliant. But one outcome tells you almost nothing about decision quality.
2015 Super Bowl: Seahawks pass play interception
Pete Carroll called a pass play on the 1-yard line instead of handing to Marshawn Lynch. The ball was intercepted. Media called it the worst call in Super Bowl or even football history. But the interception rate in that spot is only 1-2%, and if the pass is incomplete (40% chance), the Seahawks get three plays instead of two. The bad outcome retroactively made a reasonable decision look terrible.
Ignore the outcome to assess decision quality
The counterintuitive answer: unless you're running a Monte Carlo simulation (impossible for most decisions), you must ignore the outcome as much as possible and think about the decision in its absence. One outcome is just one outcome.
Strategies to Separate Outcomes from Decision Quality
Deconstruct decisions before outcomes occur
Write down possible outcomes and assign probabilities before the result is known. Memorialize your thinking. When the outcome arrives, you can point to it as one of the scenarios you considered, preventing overreaction. A trial lawyer can deconstruct strategy before a verdict; you can deconstruct career moves before results arrive.
Describe decisions only up to the point you need help
When asking others for advice, don't reveal the outcome. Describe a poker hand up to your decision point (e.g., 'I have K-Q, I raised, he re-raised, should I fold or re-raise?'), not what happened after. Once you reveal the outcome, others' bias becomes infectious—they'll rationalize the result and tell you a story that makes sense of what happened, not what was best.
Tell different people different outcomes
Tell one person a decision worked out well and another that it went poorly. Compare their analyses. You'll be amazed at how differently they assess the same decision based on the ending. Merging these perspectives gives you fidelity on actual decision quality.
Form a truth-seeking group
Other people spot your bias better than you spot it yourself. Assemble a group committed to accuracy, holding each other accountable, and remaining open to diverse viewpoints. This group can help you parse outcome from decision quality and refine your thinking over time.
Don't let people talk before giving opinions
If four people interview a job candidate, don't let them discuss before giving opinions. Make them write down their views separately. Once they talk, beliefs become infected—they'll converge and you lose independent signals. Same applies to asking for advice: don't reveal your own belief first, or you'll ruin the objectivity.
Phil Ivey vs. Phil Hellmuth: two cognitive styles
Phil Ivey won a huge tournament but spent dinner discussing all his mistakes. Phil Hellmuth says he'd be the best poker player ever if not for luck. Ivey separates decision quality from outcomes; Hellmuth attributes results to luck. Ivey's approach—looking for mistakes in wins and losses—is the path to improvement.
Self-Serving Bias and Tribal Comparison
Self-serving bias: we blame luck for losses, credit skill for wins
When I lose a poker hand, it's bad luck. When I win, I'm a genius. Over 90% of two-car accidents are reported as the other person's fault. We do this to maintain a positive self-narrative. But this bias is so strong that we miss learning opportunities.
Reference group bias: we reverse the pattern for peers
For people we compare ourselves to (our reference group), we flip the script. When they win, it's luck; when they lose, it's bad decision-making. This emerges from zero-sum thinking in poker: if I lost to you because of luck, you must have won because of luck (not skill), to preserve my self-image.
Eric Seidel's intervention: reason to be accurate, not right
When Annie complained about bad luck after losing, Eric told her: 'If you lost because of bad luck, there's nothing to learn. If you want to ask me strategy questions, I'm all ears.' He held her accountable to accuracy over self-protection. This shifted her from reasoning to be right to reasoning to be accurate—the mindset needed to win bets.
High IQ correlates with motivated reasoning
Smarter people are better at slicing and dicing information to convince themselves and others that their beliefs are true. Knowing about bias doesn't solve the problem; you still need external accountability and a group to catch what you miss.
Status Quo as a Decision
Not changing course is a decision, not a default
Staying with the status quo is a bet, not the absence of a bet. You should analyze it with the same rigor as any new decision. Most people don't, which means they stick with the status quo too much when they should switch.
Grocery store line selection as a bet
You calculate which checkout line will save you time—a clear bet. But once you choose and the line slows, you stay rather than switch, fearing your original line will speed up. Yet it's a new decision at every moment. Treating status quo as a decision means you'd switch when you learn new information.
Why Pete Carroll got demolished for switching
If he'd handed to Marshawn Lynch and they didn't score, no one would call it a bad decision—it's the consensus play, the status quo. Because he switched grocery lines (called a pass), he was judged harshly when the 1-2% outcome occurred. Progress requires switching, but it makes you vulnerable to criticism.
Poker players fold 80% of hands
Selective entry into decisions is crucial. Don't play every hand; vet each bet to ensure payoff is worth it. This applies to life: don't pursue every opportunity, but do regularly examine whether your current path (status quo) still makes sense.
Ulysses Contracts and Pre-Commitment
Ulysses contracts: bind your future self
Recognize in advance where you'll be biased and pre-commit to prevent irrational action. Odysseus asked his crew to tie him to the mast so he wouldn't steer toward the sirens' song. Modern examples: take a rideshare to a bar, sign up for exercise classes in advance, keep only healthy food at home.
Annie's book deadline strategy
Rather than let her editor set deadlines (feeling like external control), Annie set her own with confidence intervals: 60% Monday, 80% Tuesday, 98% Friday. By creating her own contract, she held herself accountable and delivered early. This made the commitment feel intrinsic.
Real-World Applications Beyond Poker
Parenting: focus on what's in your control
When a child says they did poorly on a test because the teacher was mean, don't challenge them harshly. Instead, ask: 'What are you going to do about it?' Focus on controllable factors: talk to the teacher before the next test, study differently. This teaches them to separate outcomes from decision quality.
Writing and editing: stay open to feedback
When writing the book, Annie had to separate her ego from her words and remain open to her editor's suggestions. This required the same accuracy-focused mindset as poker: the goal is the best book, not proving she was right about her original phrasing.
Career decisions: think probabilistically
Annie's career path (academics → poker → speaking → writing) emerged from a collision of luck (illness, referrals) and probabilistic thinking. Each time, she asked: what's the status quo, what's the new opportunity, what's my best bet? This openness to switching led to multiple fulfilling careers.
Tilt and emotional decision-making
In poker, tilt occurs when emotions shut down the frontal cortex and you make bad decisions. In life, a flat tire in the rain makes you feel your life is terrible, even if you got a promotion three days ago. The solution: zoom out to a longer timeframe (a year, a lifetime) to see the trend, not the moment.
Backcasting and pre-mortems for planning
Imagine you've reached your goal in three years. How did you do it? (Backcast.) Also imagine you failed. How did that happen? (Pre-mortem.) This reveals landscape better than standard planning and helps you anticipate what could go wrong.
Making Everything Quantifiable
All decisions are quantifiable as probabilities
Whether it's meeting the love of your life, liking a restaurant dish, or staying with a partner in 10 years—everything has an objective probability, even if you can't access it. You don't need perfect data; a range (20-80%) is better than 0-100%. Narrowing the range through research refines your bet.
Probabilistic thinking prevents overreaction
If you think life is chess (better play = guaranteed win), you'll be shocked and overreact when a bad outcome occurs. But life isn't chess; luck intervenes. Recognizing this means you plan ahead, stay proactive, and don't thrash when unlucky outcomes happen.
Overcoming Imposter Syndrome and Overconfidence
A good decision group checks both overconfidence and underconfidence
Imposter syndrome and overconfidence are both biases. A truth-seeking group can reflect back what they observe objectively: people keep hiring you, buying your book, referring you. This external perspective helps you escape your own head.
The group gets in your head in a good way
When you internalize the standards and feedback of your truth-seeking group (like Eric Seidel), you start processing situations differently even when alone. You imagine what they'd say, which helps you catch bias in real-time and refine your thinking.
Celebrate incremental improvement, not perfection
You'll never eliminate bias. The goal is to catch yourself more often and do it faster. Instead of berating yourself for all the times you screw up, celebrate that you did it one fewer time. This is a more self-compassionate and sustainable approach.
Annie's Career Path: Luck and Probabilistic Thinking
Illness led to poker, which led to speaking
A stomach illness landed Annie in the hospital and made her miss the academic job market. She started playing poker for money, discovered she was good at it, and eventually was asked to speak to options traders about poker and decision-making. Each turn looked like luck, but she was open to the opportunity.
Probabilistic thinking makes you less sticky to status quo
Because Annie thinks probabilistically, she's more willing to switch grocery lines. She sees opportunities on the side and is less likely to overreact to bad outcomes. This openness to change, combined with deliberative planning, has led to multiple fulfilling careers.
Notable quotes
I spent my life taking advantage of people who made decisions that weren't so good. Maybe I should spend some of my life trying to help people make better decisions. — Annie Duke
If you're not focusing on accuracy, I don't care. If you have a question about strategy, I'm all ears. — Eric Seidel (quoted by Annie Duke)
The result acts like a gravity well. You can't climb out of it to look at whether the decision process was good or not. — Annie Duke
Action items
- Form a truth-seeking group committed to accuracy, accountability, and openness to diverse viewpoints. Meet regularly to deconstruct decisions before outcomes occur.
- When asking others for advice, describe the decision only up to the point you need help—don't reveal the outcome. Write down your thinking in advance to prevent outcome bias.
- Practice separating outcomes from decision quality: zoom out to longer timeframes, ignore single results, and focus on process. Ask 'What did I do well?' and 'What could I improve?' regardless of outcome.
- Regularly examine your status quo as if it were a new decision. Set a calendar for strategic planning sessions to ask: Is there a better way? What are the goals? (Use backcasting and pre-mortems.)
- Create Ulysses contracts to pre-commit against known biases: sign up for classes, take rideshare to bars, keep only healthy food at home, set your own deadlines.
- When parenting or coaching, focus conversations on controllable factors: 'What are you going to do about it?' rather than attributing outcomes to luck or external factors.
- Practice probabilistic thinking: assign probability ranges (even 20-80%) to beliefs and decisions. Narrow the range through research and diverse perspectives.
- When you catch yourself in motivated reasoning or self-serving bias, pause and ask: 'Am I reasoning to be right, or reasoning to be accurate?' Commit to accuracy.