Thinking in Bets:
Better Decisions Under Uncertainty

Annie Duke

Author, Thinking in Bets & Quit
Special Partner, First Round Capital
World Series of Poker Champion

DECISION SCIENCE

Core Framework

Discover → Discuss → Decide

"People generally think the purpose of a meeting is for three things: discover, discuss, decide. The only thing that's ever supposed to happen in a meeting is the discussion part."

Opinions must be formed before group discussion to prevent anchoring
Independent brainstorming surfaces disagreement that groupthink buries
A senior person sharing first kills diversity of thought permanently

The Mechanics

Why Group Meetings Kill Good Thinking

Interrupting someone silences them — it's a dominance move, not discussion
Saying "You're wrong" triggers tribal defensiveness, not open minds
If a senior person speaks first, you will never hear the junior person's true estimate
Nominal group technique: write estimates independently, then share and discuss one by one

The Facilitator Move

Reflect back what each person said verbatim — no editorializing, no opinions. "What I heard you say is… is that what you meant?" This makes people feel heard even when the decision doesn't go their way.

The Result

People feel endowed to the decision — they see their input reflected in it. They also see the true spread of disagreement on the team, which makes accepting a non-preferred outcome much easier.

On the fly tactic When discussion breaks out spontaneously: "Stop. Everybody take out a piece of paper." You can get independent opinions at any moment — not just in pre-planned processes.

Playbook

Pre-Mortems Only Work If You Set Kill Criteria

"A pre-mortem is great only if you set up kill criteria. Commit to actions that you're going to take if you see those signals."

A pre-mortem without kill criteria is just brainstorming — it has no teeth
Kill criteria: specific, observable signals agreed upon before you start that will trigger a pivot or quit
The power is in the pre-commitment — it removes the emotional pull of sunk costs when the moment arrives
At First Round: structured forums elicit independent partner opinions before group discussion on any deal

The "No Long Feedback Loops" Insight

"There is no such thing as a long feedback loop. The way you choose to shorten the feedback loop is to say: what are the things that are correlated with the outcome that I eventually desire?"

Example in investing: you can't know in 10 years if a company will IPO, but you can measure leading indicators now — team quality, early retention, revenue shape.

Mental Time Travel

Ask future-you: "How do I really think I'm going to think about this in 10 years?" The focusing effect makes present feelings enormous. Time travel shrinks them to their true size.

From Kahneman

Making Intuition Explicit

"It's so incredibly necessary to take what's implicit and make it explicit. If you don't make it explicit, you don't get to find out when it's wrong."

Nothing is as important as it seems when you're thinking about it — Kahneman's most portable insight
Adversarial collaboration: find your sharpest critic and write a paper together — design the study that could resolve the dispute
The word "nevertheless": employees feel heard, leader commits to the path. "I hear you. Nevertheless, this is what we're going to do."
Resulting: judging decision quality by outcome is a fundamental error — a good decision can produce a bad outcome

Kahneman on his own work He was the first to say "I wish priming wasn't in the book — it doesn't replicate." Radical intellectual honesty about your own prior beliefs is the rarest and most valuable epistemic skill.

Contrarian

Annie Duke's Most Counterintuitive Takes

✗ Meetings are for brainstorming together INSTEAD → ✓ Meetings are only for discussion. Discovery must happen independently beforehand — groupthink starts the moment the first person speaks.

✗ Long-term decisions just have long feedback loops — you have to wait INSTEAD → ✓ There are no long feedback loops. Find leading indicators correlated with your eventual desired outcome and measure those now.

✗ A pre-mortem is enough to protect against bad decisions INSTEAD → ✓ A pre-mortem without pre-committed kill criteria is just theater. The power is in the binding trigger — not the worry list.

✗ Good outcomes mean good decisions; bad outcomes mean bad ones INSTEAD → ✓ "Resulting" is the root of most bad decision culture. Poker taught Annie: you can play a hand perfectly and lose, and terribly and win. Evaluate process, not outcome.