This governs the gap between a decision and how it turned out, which is where almost all bad learning happens. Duke's claim is that exactly two things determine how a life goes, decision quality and luck, and that our default habit is to read the second as evidence about the first in whichever direction flatters us. Her method is to treat every decision as a bet on an uncertain future: state your confidence as a number, sort each result into luck or skill on purpose, and build a small group whose job is to check both.

Do

Separate your worst decision from your worst result, out loud. Duke's standing exercise with executives, asking for their best and worst decisions of the year, always returns best and worst results instead. A CEO named his firing of the company's president as his worst decision purely because the aftermath went badly; walking the room through his actual process, competitor benchmarking, coaching first, a rejected plan to split the role, produced agreement that it was a reasonable decision with a bad result. [Source: "Chapter 1: Life Is Poker, Not Chess"]

State your confidence as a number rather than a verdict. Duke's own example is replacing a flat claim that Citizen Kane won Best Picture with "I'm 60% that Citizen Kane won." This mirrors how scientists report findings, and the research she cites says calibrated uncertainty reads as credible rather than weak. It also converts the conversation: people volunteer what they know instead of defending themselves against being called wrong. [Source: "Chapter 2: Wanna Bet?"]

Ask "wanna bet?" of your own beliefs before someone else does. Beliefs arrive almost accidentally. Gilbert's research found we default to believing what we hear and only sometimes get around to vetting it later, which is how baldness from the maternal grandfather and one dog year equalling seven human years both became common knowledge with no real source behind either. New poker players adopt folklore about which hands are profitable that a look at their own results reliably disproves. [Source: "Chapter 2: Wanna Bet?"]

Sort each outcome into luck or skill on purpose, using a fixed test. If repeating the same decision would predictably repeat the outcome, it belongs to skill; if it depended on things outside your control, it belongs to luck. Duke's counterexample is a regular who believed the worst starting hand was secretly the best, played it constantly, credited every win to his own genius and wrote off every loss as bad luck, and so lost for years without ever updating. [Source: "Chapter 3: Bet to Learn: Fielding the Unfolding Future"]

Recruit a small group and give it an explicit charter. Duke's minimum is three, two to disagree and one to referee, with everyone having opted in to accuracy over feelings. Tetlock and Lerner's finding gives the conditions: open-minded reasoning rises when people expect to be accountable to an audience that is unknown to them, focused on accuracy, informed, and has a legitimate reason to ask. Her own poker group required her to discuss hands only in terms of her possible mistakes, never her bad luck. [Source: "Chapter 4: The Buddy System"]

Withhold the outcome when you ask someone to check your reasoning. Expert poker players describe a hand up to the decision point and deliberately do not say how it ended, so the listener judges the decision. Particle physics does the same thing at scale, increasingly hiding expected results from the analysts working the data to prevent unconscious steering. [Source: "Chapter 5: Dissent to Win"]

Run 10-10-10 before you commit. What are the consequences of this choice in ten minutes, in ten months, in ten years? Duke's extension of Suzy Welch's question is aimed at a specific failure: regret only helps if it arrives before the decision it warns against, and Nietzsche called after-the-fact remorse adding a second act of stupidity to the first. [Source: "Chapter 6: Adventures in Mental Time Travel"]

Precommit a rule that binds the version of you who is upset. The Ulysses contract is anything that constrains a future, less rational self: automatic deductions, a ride-share instead of driving, a walk-away rule triggered by named verbal and physical warning signs. Duke's own list of flagged phrases, "I knew it," "you're 100% wrong," "I can't believe how unlucky I got," works as a swear jar that triggers a deliberate pause. Her poker group imposed a preset loss limit on her early career precisely because she could not judge her own state in the moment. [Source: "Chapter 6: Adventures in Mental Time Travel"]

Do a backcast and a premortem for the same plan. Imagine the headline announcing the goal achieved and work backward to the decisions that got there, then imagine the headline announcing failure and work backward to the causes. Oettingen's research on mental contrasting says this beats pure positive visualization; in one study dieters who fantasized about slimming down lost 24 fewer pounds than those who pictured the obstacles. [Source: "Chapter 6: Adventures in Mental Time Travel"]

Don't

Don't grade the call by the card that came. Pete Carroll's decision to pass on the one-yard line was widely called the worst play call in Super Bowl history because it was intercepted; his own line, that the call would have been great if they catch it, is the fallacy stated perfectly. A 24% hand that wins is not a miscalculation, it is the 24% arriving. [Source: "Chapter 1: Life Is Poker, Not Chess"]

Don't say "I should have known." Once an outcome is known the mind treats it as having been inevitable. Duke cites Judge Easterbrook overturning a punitive-damages verdict specifically because the jury had judged a decision as though its rare later outcome had been obviously foreseeable, the same illusion that made a spectator's reflexive reach for a foul ball and a forecaster's 2016 election odds look in hindsight like certainties neither ever was. [Source: "Chapter 6: Adventures in Mental Time Travel"]

Don't credit your wins to yourself and your losses to luck. Self-serving bias is the predictable direction of the error. MacCoun found 91% of drivers in multi-vehicle crashes blamed someone else, and von Neumann, after wrecking his own car, blamed a tree that stepped into his path. The mirror image is worse: we field other people's wins as luck and their losses as their own fault. [Source: "Chapter 3: Bet to Learn: Fielding the Unfolding Future"]

Don't assume knowing about a bias protects you from it. Duke's anchor is Kahneman's Müller-Lyer illusion: knowing the two lines are the same length does not stop you seeing them as different. Worse, the bias blind spot grows with cognitive ability, and in Kahan's experiment the most numerate subjects were the most polarized in misreading identical data, because they were better at constructing a rationalization. [Source: "Chapter 2: Wanna Bet?"]

Don't hand someone a truthseeking challenge they never agreed to. Duke calls it Lettermanning, after an on-air suggestion to a guest that maybe she was the problem, accurate and badly received because no such exchange had been agreed. Her rule outside the group is to ask first whether the person wants to vent or wants advice. [Source: "Chapter 4: The Buddy System"]

Don't assemble a group that already agrees with you. Tetlock and Lerner separate confirmatory thought, rationalizing one point of view, from exploratory thought, considering alternatives even-handedly. Sunstein's study of over 6,000 federal appellate cases found a single judge from the opposing party had a large disciplining effect on how a panel voted, and the deliberate practice of hiring ideologically opposed clerks had nearly vanished by 2010. [Source: "Chapter 4: The Buddy System"]

Don't make someone defend a past decision. Asking why they did it triggers defense of the record. Asking what they could do differently going forward opens the same ground without it. Duke used this with her own son over poor grades, validating the past and then asking whether there was anything he could do to improve going forward. The smaller version is replacing "but" with "and" when you add contrary information. [Source: "Chapter 5: Dissent to Win"]

Don't let the source decide the merits. Mertonian universalism means judging an idea independently of who said it, in both directions: do not shoot the message because you dislike the messenger, and do not wave one through because you like them. Duke's own error was treating her brother's beginner hand list as scripture and dismissing anyone who deviated from it, which cost her a year of profitable strategies. [Source: "Chapter 5: Dissent to Win"]

Where they disagree

Can you actually put a number on the rare event? This is the real fault line between Duke and Taleb, and both corpora carry it explicitly. Duke's entire method runs on calibration: state beliefs as percentages, treat confidence as a spectrum, and use prediction markets, which she reports beating ordinary expert opinion at 71% accuracy versus 58%, because money on the line rewards accuracy over consensus. Taleb's position is that a probability attached to a rare event cannot come from frequency data, since the event is rare, so the number necessarily rests on an unverifiable prior theory; his example is a Lehman employee telling the Wall Street Journal that the August 2007 events should happen once every ten thousand years, followed by three similar events in three consecutive days. He also argues prediction markets specifically break down for open-ended exposures, because a binary bet on whether something happens does not hedge an unbounded loss if it does. [Source: "Chapter 2: Wanna Bet?"] [Source: "Chapter 4: The Buddy System"] [Source: "Postscript V: (Perhaps) the Most Useful Problem in the History of Modern Philosophy"] [Source: "Postscript VII: What to Do with the Fourth Quadrant"]

What separates them is the shape of the payoff, not the value of thinking clearly. Taleb's own map does the reconciling: he sorts decisions by whether the payoff is simple and bounded or open-ended, and by whether the randomness is mild or fat-tailed, and says that in three of those four cases finding the best available model and relying on it is entirely reasonable. Nearly every example Duke works, a poker hand, a hire, a job move, a session with a preset loss limit, is a repeated bet with a bounded downside, which is exactly where calibration earns its keep. Her method is not a claim that tail risk is estimable; it is a discipline for the bets you will make hundreds of times. Use the percentages where the loss is capped and you get to play again, and use exposure limits, not estimates, where it is not. [Source: "Postscript VI: The Fourth Quadrant, the Solution to that Most Useful of Problems"]

On the central point they do not disagree at all. Taleb states Duke's core claim independently and in the same terms: a decision cannot be judged by its outcome alone but by the cost of the alternative histories that could have unfolded instead, and only people who fail tend to say so, since winners credit their own decision quality. His illustration is a television host publicly humiliating an economist for a call the market had not yet vindicated, which is resulting under a different name. Both authors are attacking the same habit from opposite ends of the same problem. [Source: "Chapter Two: A Bizarre Accounting Method"]

The one line

Ask what you would have done knowing only what you knew then, put a number on it, and get three people whose job is to tell you when the number is wrong.