Six months in, the project is behind, the numbers are soft, and you keep finding reasons to give it one more quarter. The Annie Duke quit argument is built for that moment. Quitting on time usually feels like quitting too early, she argues, so the signals that tell you to stop should be written down before you start.

Duke makes the case in Quit: The Power of Knowing When to Walk Away. It reads as a reply to the persistence shelf. Our Grit summary found the largest meta-analysis put grit’s correlation with academic performance at .18, and our Carol Dweck mindset summary found .10 between mindset and achievement across 365,915 students.

We took 17 studies and real cases Quit relies on, as reviewers and readers’ notes record them, and traced each one to its original paper or record.

What We Found

Of 17 studies and cases from Annie Duke’s Quit that we traced to their sources on 20 November 2026, 11 hold up as the book uses them, and six need a caveat from later research or the record.

The core of the book holds. Losses do weigh about twice as much as gains, and Steven Levitt’s coin-flip study found that people who quit a job were happier six months later. The weak spots cluster in one place, the research the book uses to explain when people persist and when they stop.

Three of the six caveats sit there. The other three are details in two stories and one figure about Slack’s origin that we could not trace.

Quit by Annie Duke: What Is the Book About?

Quit by Annie Duke is a 2022 book from Portfolio, 336 pages long, arguing that quitting is a decision skill and that most people quit too late. Duke, a former professional poker player, says the urge to persist comes from predictable biases, and that you can counter them with expected value, advance kill criteria and an honest outside adviser.

In her 2022 interview with David Epstein, Duke said professionals play about 15 to 25 percent of starting hands, and amateurs more than half. Our piece on Tobey Maguire’s poker game traced 37 facts about one private game and tied 14 to court or government records.

The first section makes the case for quitting, with loss aversion and expected value. The second, In the Losses, covers escalation of commitment, sunk costs, and monkeys and pedestals. The third takes on the endowment effect, status quo bias, identity and the quitting coach. The fourth covers opportunity cost and the trouble with fixed goals.

Two ideas carry the practical weight. Monkeys and pedestals, from Astro Teller at X, the moonshot factory, means doing the hardest part of a project first, so you learn early whether it can work. The second is kill criteria, signals chosen in advance that tell you to stop.

The Annie Duke Quit Studies, Traced to Source

The table lists every Annie Duke quit study and case in our sample, what the book uses it for, the source we traced it to and our verdict. Holds means the source says what Quit uses it to say. Caveat means later research or the record adds something a reader should know.

Study or case What Quit uses it for Source we traced Sample Verdict
Levitt coin-flip study Quitters were happier on average Levitt, NBER 2016; Review of Economic Studies 2021 22,511 coin tosses Holds
Loss aversion Losses hurt about twice as much as gains Brown, Imai, Vieider and Camerer, 2024 meta-analysis 607 estimates, 150 articles Holds
New York cab drivers A case of quitting too early Camerer, Babcock, Loewenstein and Thaler, 1997 Three driver samples Caveat
Cab drivers, 2009 to 2013 Veterans judge when to stop better Farber, NBER 2014 Every NYC taxi trip, five years Holds
Portfolio managers Good at buying, poor at selling Akepanidtaworn, Di Mascio, Imas and Schmidt 783 portfolios Holds
Knee-deep in the big muddy People double down on a failing choice Staw, 1976 240 students Caveat
Sunk cost and endowment Paid tickets, the wine collector Thaler, 1980 Thought experiments Holds
Mug experiments Owners want double to sell Kahneman, Knetsch and Thaler, 1990 Mug owners and buyers Holds
Sunk costs in the NBA Draft order kept players on court Staw and Hoang, 1995 241 players Caveat
Minimum targets Targets set in advance cut escalation Simonson and Staw, 1992 193 students Holds
Jeffrey Rubin Entrapment scholar who died on a climb Spokesman-Review, 5 June 1995 One case Caveat
Status quo bias People stick with what they have Samuelson and Zeckhauser, 1988 Experiments, faculty plan choices Holds
IKEA effect People overvalue what they built Norton, Mochon and Ariely, 2011 Furniture, origami, Lego studies Holds
X and Foghorn Kill criteria, monkeys and pedestals Astro Teller, X blog, 2016 One project Holds
Everest, 1996 Turnaround times 1996 Mount Everest disaster One expedition Caveat
California high-speed rail Pedestals that look like progress Construction Dive, 2023 One project Holds
Glitch to Slack Quitting a game, keeping the tool Glitch (video game) One company Caveat
A founder talking with an investor in a venture firm's lounge, the stay-or-stop decision Annie Duke's Quit is about

Which Studies in Quit Need a Caveat?

Six of the 17 studies and cases in Quit need a caveat. Three are studies of persisting and stopping: Staw’s 1976 role-play, Staw and Hoang’s 1995 NBA paper and the 1997 New York taxi study. Two are stories, Jeffrey Rubin’s death and the 1996 Everest turnaround, where sources differ on details. One is a figure about Slack’s origin.

Staw’s experiment founded the escalation research. The 1976 paper put 240 business undergraduates at the University of Illinois through a role-played investment decision. Those personally responsible for a losing choice put the most money back into it.

The FORRT replication atlas lists a 2021 study by Białek and colleagues that found no evidence of escalation among 365 participants, and a 2022 study it rates as mixed.

Staw and Hoang followed players from the first two rounds of the 1980 to 1986 drafts, 241 of them in the playing-time analysis. A second-round pick meant 552 fewer minutes in the second season, and first-rounders stayed about 3.3 years longer, after controlling for performance. Quit reports both figures exactly.

Colin Camerer and Roberto Weber re-examined the same data in 1999. Once they tested rational explanations, the escalation effect was still there, but weaker in both duration and size.

The taxi research cuts against itself. Camerer, Babcock, Loewenstein and Thaler found in 1997 that New York drivers worked fewer hours when hourly earnings were high and more when they were low, across three samples. Henry Farber’s study of every New York taxi trip from 2009 to 2013 found drivers generally worked more when earnings were higher, which, Farber writes, does not support the reference-dependent preferences model.

Quit uses Farber’s paper for a different point, that veterans judge when to stop better than newcomers. That part holds: Farber found new drivers who responded poorly to earnings tended to leave, and the rest learned fast.

The Epstein interview tells the story of Jeffrey Rubin, a psychologist who studied entrapment and died climbing his 100th New England peak after a student turned back in heavy fog. He co-wrote Entrapment in Escalating Conflicts in 1985. The Spokesman-Review’s report of 5 June 1995 says Rubin, 53, a Tufts psychologist, was on the last climb of New England’s 100 highest peaks, and that after the pair reached a summit in high wind and heavy rain, his friend turned back and Rubin went on.

On Everest in 1996, three of Rob Hall’s clients turned back below the summit, and eight climbers died in the storm that followed. In the Epstein interview, Duke gives the Everest turnaround as 1 p.m. Wikipedia’s account of the disaster gives 2 p.m. and notes that some sources say 1 p.m.

Stewart Butterfield shut down the online game Glitch, and the team’s internal messaging tool became Slack. Glitch launched on 27 September 2011 and closed on 9 December 2012, for limited audience appeal, and Salesforce bought Slack in 2021. Strategy+business reports the book’s detail that devoted players were under 5 percent of Glitch’s following. We could not find that figure in Wikipedia’s account or in three 2012 reports of the shutdown we read.

Quit’s 11 of 17 sits close to the 12 of 21 stories that matched the record when we checked The Obstacle Is the Way. We traced the same way for the Chris Hemsworth diet, where eight of 27 claims were in his own words, and Winston Churchill’s daily routine, where 40 of 61 accounts were first-hand.

What Holds Up: Loss Aversion, the Coin Flip and the Selling Problem

Duke’s claim that losses hurt about twice as much as gains goes back to Kahneman and Tversky’s 1979 prospect theory paper, and the number has held. A 2024 meta-analysis in the Journal of Economic Literature pooled 607 estimates from 150 articles and found a mean loss aversion coefficient of 1.955. Loss aversion has critics, but the average sits close to two. Other findings from that tradition fared worse; our Thinking, Fast and Slow summary found that four of 16 headline studies held up and five failed replication.

In Levitt’s working paper, people stuck on a decision flipped a coin on a website, 22,511 usable tosses in all. Heads, meaning make the change, made people 26 percentage points more likely to change. At six months, those who had quit a job or ended a relationship reported being markedly happier.

Levitt lists the limits himself. The volunteers came through the Freakonomics site and are “far from representative,” the happiness is self-reported, and only 8,159 six-month surveys came back. The job-quitting effect was significant at six months and not at two.

Quit cites more than 700 institutional portfolio managers who were good at buying and poor at selling. We matched it to a study by Klakow Akepanidtaworn, Rick Di Mascio, Alex Imas and Lawrence Schmidt. Chicago Booth Review’s 2019 summary describes 783 portfolios averaging $573 million whose sells underperformed a random selling strategy by 70 basis points over a year.

Exits can stay opaque even at the top. Of the 13 deals behind Jay-Z’s net worth that we traced, the two drinks sales Forbes has called his biggest sources of wealth carry no price from either buyer.

The sunk cost and endowment material traces to Richard Thaler. His 1980 paper, as Nicholas Barberis describes it, set out both, with a colleague who would not sell a bottle of wine for $100 and people who drive through a snowstorm to a basketball game because they paid for the tickets. Quit’s paid-versus-free ticket question is the same thought experiment.

The mug experiments of Kahneman, Knetsch and Thaler in 1990 found sellers asked about twice what buyers would pay. In Arkes and Blumer’s 1985 field test at Ohio University’s theatre, the 18 patrons who paid the full $15 for season tickets used more tickets in the first half of the season than the 36 given discounts. The notes we read do not show Duke citing it.

Status quo bias comes from William Samuelson and Richard Zeckhauser’s 1988 paper, which found it in experiments and in faculty choices of health plans and retirement programmes. The IKEA effect comes from Michael Norton, Daniel Mochon and Dan Ariely, whose subjects overvalued furniture, origami and Lego they built themselves. The extra value appeared only when the task was finished.

What Are Kill Criteria?

Kill criteria are signals you choose before starting a project that tell you to stop. Duke recommends pairing a state with a date: if this measure is not reached by this time, I quit. Written in advance, they replace a decision made while losing, when sunk costs and loss aversion are strongest, with one made calmly.

Itamar Simonson and Barry Staw tested de-escalation methods in 1992 on 193 business students at Berkeley and Cal State Hayward. Setting minimum targets that would trigger a change of policy significantly cut escalation. Asking people to outline the advantages of each option had no effect.

The same study found that judging people on their decision process reduced escalation, while judging them on outcomes increased it. Our piece on Nick Saban’s Process found ten of 15 ideas attached to that coaching method trace to Saban’s own words.

X runs kill criteria at company scale. In a 2016 post, Astro Teller described Foghorn, a team turning seawater into fuel, which had five years to show the fuel could compete on cost. The team shut its own project down to free resources for better bets.

The Bryan Johnson Blueprint runs to 53 daily steps, and each one is a commitment that could carry a measure it must hit to stay on the list.

A chief executive listening to a product director on a steel bench in a headquarters atrium, weighing whether to end a project

How to Know When to Quit: What Does the Research Say?

To know when to quit, ask whether you would start this today, knowing what you know now, and compare the expected value of staying with your best alternative. The research says to set kill criteria in advance, because weighing pros and cons in the middle of a losing project changed nothing. Then ask someone who will tell you the truth.

Duke’s coaching example shows the expected-value step. A client unhappy in her job saw that staying meant near-certain unhappiness while switching only risked it, and the choice looked different.

The quitting coach is the third step. Duke credits Daniel Kahneman with the idea that everyone needs a friend who loves them but will not spare their feelings. She also cites the investor Ron Conway, asked whether founders owe it to staff to keep going, answering that life is too short for both.

This is planning ahead, the opposite habit from Eckhart Tolle’s. Our Power of Now summary found only seven of 16 summaries mention his carve-out for practical clock time, the room a quit date needs.

Anyone applying an Annie Duke quit rule to a career should expect the decision to feel wrong at first. That feeling is the book’s central prediction, and Levitt’s six-month data suggest it passes.

What Quitting Looks Like for a Company

Duke’s sharpest business example is about time. In a reader’s highlights, she scans the advertisers from Tom Brady’s first Super Bowl in 2002 and finds that by his seventh title, in 2021, he had outlasted many of them, Blockbuster, Circuit City and RadioShack among them.

Of the 11 companies in Jim Collins’s 2001 book, our Good to Great summary found seven still independent and listed on 15 November 2026.

California voters approved a $33 billion high-speed rail plan in 2008, and by March 2023 the estimate had reached $128 billion. Quit uses it to show how building the easy parts first looks like progress.

Companies quit ideas too. In our count of Netflix’s 2009 culture deck, 10 of 37 ideas are gone from today’s memo, including the slide calling hard work irrelevant. Our ranking of the best books for decision making sets Duke beside the rest of the shelf.

How We Did This

On 20 November 2026 our newsroom listed the studies and cases behind the Annie Duke quit framework from Graham Mann’s chapter notes, a reader’s Goodreads highlights, strategy+business’s review and Duke’s Epstein interview. We kept every item named clearly enough to trace, 12 studies and five cases, and followed each to the original paper, a scholarly record or a contemporary report.

A study holds if the source says what Quit uses it to say; it needs a caveat if later research or the record adds or changes something. Without the book’s endnotes, its wording comes from those pages. Concepts named without a source were traced to the paper that named them; unnamed anecdotes were left out. One more study, on optimists, came with no researcher or year; we could not identify it and did not count it. The full dataset is available on request.

Frequently Asked Questions

Which Annie Duke books should you read first?

Annie Duke’s books page lists her three decision books, Thinking in Bets, How to Decide and Quit, plus three other titles, two of them poker strategy guides. Start with Quit if a stay-or-go choice is in front of you.

Is the Quit book worth reading?

Yes, if you are weighing whether to stay in a job, a project or an investment. By our count, 11 of its 17 traceable studies and cases hold up as used. Read the escalation chapters knowing the classic experiments have weakened under replication.

Who publishes Quit, and how long is it?

Portfolio published Quit: The Power of Knowing When to Walk Away on 4 October 2022, and the publisher’s page lists it at 336 pages.

Quitting on time will still feel early. The research Duke gathered says that feeling is a poor guide, and a target written down before you start is a better one.

Latest From The Executive Insight