LLM Book Knowledgebase

Resulting and Probabilistic Thinking

9 min read

Resulting and Probabilistic Thinking

Annie Duke’s Thinking in Bets: Making Smarter Decisions When You Don’t Have All the Facts (2018) introduces resulting as a name for one of the most pervasive and damaging errors in decision-making — and then builds a systematic alternative framework around treating decisions as bets and thinking probabilistically.

Resulting: The Foundational Error

Resulting is the error of judging the quality of a decision by the quality of its outcome.

“Pete Carroll was a victim of our tendency to equate the quality of a decision with the quality of its outcome. Poker players have a word for this: ‘resulting.’ When I started playing poker, more experienced players warned me about the dangers of resulting, cautioning me to resist the temptation to change my strategy just because a few hands didn’t turn out well in the short run.” — Annie Duke, Thinking in Bets

The Pete Carroll case is Duke’s central example: Carroll’s call to pass (rather than run) at the goal line in Super Bowl XLIX is widely remembered as one of the worst decisions in NFL history. It was intercepted and the Seahawks lost. But the probability analysis of the specific situation — given the defensive alignment, the time remaining, the down and distance — supported the pass call as the higher-expected-value play. The decision was sound; the outcome was unlucky. Yet the entire sports commentary apparatus judged the decision by its result.

The inverse failure is equally common. A CEO who drives drunk and makes it home safely updates his belief that drunk driving is acceptable — despite the fact that the good outcome says nothing about the quality of the decision:

“He changed his behavior based on the quality of the result rather than the quality of the decision-making process. He decided he drove better when he was drunk.”

Resulting creates a systematic learning failure: bad decisions that happen to produce good outcomes get reinforced; good decisions that happen to produce bad outcomes get abandoned. Over time, this degrades both individual and organizational decision-making capacity.

The Two Sources of Outcomes

The corrective framework begins with a clear decomposition:

“Thinking in bets starts with recognizing that there are exactly two things that determine how our lives turn out: the quality of our decisions and luck. Learning to recognize the difference between the two is what thinking in bets is all about.”

Almost no outcome is 100% the product of decision quality or 100% the product of luck. Most outcomes sit on a spectrum. The error is treating outcomes as if they are exclusively one or the other:

“Off-loading the losses to luck and onboarding the wins to skill meant he persisted in overestimating the likelihood of winning with seven-deuce. He kept betting on a losing future.”

Self-serving bias produces this asymmetry: we attribute our own good outcomes to skill and our own bad outcomes to luck, while attributing others’ good outcomes to luck and their bad outcomes to skill. The pattern is predictable, documented across multiple research traditions, and deeply costly.

What Makes a Decision Great

The central definitional claim:

“What makes a decision great is not that it has a great outcome. A great decision is the result of a good process, and that process must include an attempt to accurately represent our own state of knowledge. That state of knowledge, in turn, is some variation of ‘I’m not sure.‘”

This is a radical reframe. A decision is not right or wrong based on what happened. It is good or bad based on whether the process — the gathering and weighting of evidence, the identification of alternatives, the assessment of probabilities — was done well given what was knowable at the time.

The implication: you can make a great decision and get a terrible outcome (the doctor who makes the best call given available information and the patient dies anyway). You can make a terrible decision and get a great outcome (the drunk driver who makes it home). Confusing the two — resulting — destroys your ability to learn from experience.

Decisions as Bets

The reframe Duke proposes:

“Decisions are bets on the future, and they aren’t ‘right’ or ‘wrong’ based on whether they turn out well on any particular iteration.”

Every decision is a bet: it commits resources (time, money, energy, reputation) toward a future that may or may not occur, against alternatives that could have produced different outcomes. Making this frame explicit has several benefits:

  1. It acknowledges uncertainty: a bet cannot be perfectly certain. Treating a decision as a bet forces acknowledgment that you might be wrong.
  2. It creates calibration norms: the person who wins bets over the long run is the one with more accurate beliefs, not the most confident one.
  3. It decouples process from outcome: a well-placed bet can lose; a poorly-placed bet can win. Neither outcome alone tells you the quality of the bet.

“Every decision commits us to some course of action that, by definition, eliminates acting on other alternatives. Not placing a bet on something is, itself, a bet.”

Beliefs as Probabilistic States

Duke extends the framework from decisions to beliefs:

“We bet based on what we believe about the world.”

Most people treat beliefs as binary: you either believe something or you don’t, you’re right or wrong. Duke argues that a more accurate — and more useful — representation treats beliefs as probabilistic states: how confident am I, on a scale from 0–100%, that this is true?

Expressing beliefs probabilistically has several advantages:

  • It makes calibration measurable and improvable
  • It creates openness to new evidence without requiring a full reversal (“I was 70%, now I’m 55%” is easier to accept than “I was wrong”)
  • It communicates honest uncertainty to others, reducing the contagion of overconfident claims
  • It triggers deeper vetting of the belief itself (the “Wanna bet?” trigger)

“Forcing ourselves to express how sure we are of our beliefs brings to plain sight the probabilistic nature of those beliefs, that what we believe is almost never 100% or 0% accurate but, rather, somewhere in between.”

Calibration Is Hard and Gets Corrupted

Our natural belief-formation process is not calibrated: “We hear something; We believe it to be true; Only sometimes, later, if we have the time or the inclination, we think about it and vet it.” Motivated reasoning then protects beliefs from disconfirmation. Even people who intellectually accept probabilistic thinking tend to slide back toward black-and-white certainty when emotions or identity are engaged. The practices Duke describes — truthseeking groups, explicit probability expression, “Wanna bet?” triggers — are corrective scaffolding against this natural drift.

Truthseeking Groups: The Social Architecture

Individual calibration is insufficient. We need social structures that reward truthseeking and penalize ego protection:

“A good decision group is a grown-up version of the buddy system.”

Duke’s model is the poker study group — a small pod of players who committed to evaluating each other’s decisions honestly, sharing information openly, and rewarding accuracy over ego comfort. The charter of such a group has four properties (drawn from Merton’s CUDOS norms for science):

  • Communism: share all relevant data — do not withhold information that would change the group’s assessment
  • Universalism: evaluate claims on their merits regardless of who made them
  • Disinterestedness: vigilance against conflicts of interest, including emotional ones
  • Organized Skepticism: create structured permission for dissent; reward the best critics, not the most enthusiastic cheerleaders

“‘You put individuals together in the right way, such that some individuals can use their reasoning powers to disconfirm the claims of others, and all individuals feel some common bond or shared fate that allows them to interact civilly, you can create a group that ends up producing good reasoning as an emergent property of the social system.‘”

Backcasting and Premortems: Thinking Backward from the Future

Duke’s most operationally distinctive contribution is the use of mental time travel to improve decision quality before committing:

Backcasting: imagine you have already achieved the goal. What happened? What decisions and events were necessary to get there? This reveals:

  • Which steps are genuinely necessary vs. merely assumed
  • Where low-probability events must occur for success
  • Whether the goal is achievable at all

Premortem: imagine you have already failed. What went wrong? The premortem harvests pessimistic intelligence that optimistic forward-planning suppresses:

“A premortem is where we check our positive attitude at the door and imagine not achieving our goals. Backcasting imagines a positive future; a premortem imagines a negative future. We can’t create a complete picture without representing both the positive space and the negative space.”

The premortem is organizational skepticism applied to your own plan — it gives permission to voice concerns that team dynamics and social norms would otherwise suppress.

Ulysses contracts: precommitments that bind future-self to present-self’s more rational choices. Named for Odysseus tying himself to the mast to resist the Sirens. Examples: automatic retirement contributions (prevent present-you from raiding the account), using a ride-share when going to a bar (prevent impaired you from driving), writing decision rules in advance for when market volatility triggers a sale.

Hindsight Bias and the Trunk-and-Branches Model

Duke introduces a striking metaphor for hindsight bias:

“Think about time as a tree. The tree has a trunk, branches at the top, and the place where the trunk meets the branches. The trunk is the past. The branches are the potential futures. As the future becomes the past, what happens to all those branches? The ever-advancing present acts like a chainsaw.”

Once an outcome occurs, we lose sight of all the alternative futures that could have occurred. The single outcome that happened expands retroactively to feel inevitable — as if there was never any alternative. This is hindsight bias: the feeling that “I knew it all along” or “that should have been obvious.”

Resulting is hindsight bias applied to decision evaluation: we judge the decision by the single branch that materialized and forget all the other branches that were equally possible at the time the decision was made.

Backlinks