Nate Silver on Life’s Mixed Strategies

Statistician Nate Silver, author of On the Edge: The Art of Risking Everything, talks with Tyler Cowen about poker as applied game theory, how to reason under uncertainty, why he doubts a near-term AI intelligence explosion, the flaws in ranked-choice voting and prediction markets, the peculiarities of the Bay-Area rationalist subculture, and the politics of immigration and populism.

Key ideas

1. Life is a set of mixed strategies. Solving poker’s Nash equilibrium revealed that almost every hand is best played as a deliberate mix, not a single right move — Silver will literally randomise off the tournament clock. The moment you have any read on an opponent, though, you shift from indifference toward a dominant strategy, so local, specific, actionable information beats elegant theory every time.
2. Reason in calibrated probabilities and act on the tractable. Silver states beliefs as numbers you could bet on — Pascal’s wager at p = 0.001, Democrats 55–45 for 2028 — and deliberately ignores ‘infinite ethics’ edge cases in favour of the terrestrial and tractable. Apparently irrational habits like loss aversion are often, he argues, rational ‘on a higher plane’.
3. AI is further from superforecasting than the hype suggests. He holds to a 10–15-year timeline for AI to match human superforecasters and rates the past year’s progress at the 40th percentile of what he expected, distinguishing static problems (Math Olympiad, chess) from dynamic systems with shifting inputs, at which LLMs — very bad at poker — still fail. Cowen bets one to two years.
4. Prediction markets and polling both have edges and blind spots. Markets have grown more liquid but still misprice (a 7 per cent chance for a constitutionally ineligible candidate); a name and zip code predict a voter with ~90 per cent accuracy, yet elections turn on 1–2 per cent margins, so nothing yet beats polling’s calibration baseline stretching back to 1936.
5. The rationalist community is a proudly weird bubble. High in openness, high-variance in conscientiousness, and prone to gullibility (Silver thinks EA should have taken more damage over Sam Bankman-Fried), it leaps too quickly from rough AGI to superintelligence and underrates political constraints and human adaptability.

Content

Game theory, poker, and reading tells

Silver opens on what still feels fresh to him about expected-utility theory: not the expected-value part, which any economics-101 audience grasps, but solving for the equilibrium — what behaviour should emerge once everyone is playing their incentives optimally, a lens he applies even to why people act as they do on X. The deepest lesson from solving poker’s Nash equilibrium is how many mixed strategies it contains: nearly every hand is a mix, and Silver will randomise deliberately, reading a high number off the tournament clock as a cue to bluff or the rotation of his chips as a die. The flip side is that any genuine read on an opponent moves you off indifference toward a dominant play, so table image and tells matter enormously. Tells, though, are contextual and semantic — a visibly pounding carotid artery can mark a bluff in one player and a monster hand in another — and there is no substitute for building an implicit database by watching thousands of hands, most of them folded. Being right 60 per cent of the time against a 50 per cent baseline is a large edge; in gambling any 55/45 is enormous. Poker, he argues, has gone furthest of any domain in the literal real-world manifestation of game theory, with football’s unexpected third-and-long draw play a fainter cousin.

Academic influence and the death of the journal article

Asked for the last interesting thing he learned from academic political science, Silver laughs and cannot name one — an answer Cowen calls possibly correct. He reads Substacks, not journals, and thinks academics have lost influence: much of the conversation has migrated to a Bluesky he finds intolerable, and the reflexive, nuance-melting anti-Trump posture, combined with academia’s deliberately slow ‘slow-cooking’ method, is a poor fit for the news cycle. The right unit of work is no longer the paper but a multi-day newsletter post, a book that goes deeper than anyone has, or genuine investigative journalism — he cites the reporter who broke the Epstein story for the Miami Herald against a backdrop of lazy incuriosity. Ninety per cent of academic papers, he says, would work perfectly well as blog posts, shorn of the Greek symbols and the pretence; the value of a newsletter is precisely the tonal control to signal when he is speculating, joking, or presenting an original finding.

Expected value, Pascal’s wager, and the shape of risk

On whether to obsess over longevity like Bryan Johnson, Silver demurs: hyper-optimising every daily parameter tends to make you miserable, and he settles for serious but incremental improvements. He puts Pascal’s wager at p = 0.001 and declines to act on it, taking seriously only the profound uncertainty about the universe’s nature; he has drifted from atheism toward being ‘affirmatively agnostic’, helped along by AI raising questions about reality, but he keeps to the tractable terrestrial realm rather than infinite-ethics edge cases like shrimp welfare — glad, he adds, that a few people worry about them. His poker instinct dominates: what matters is the hand you hold and the read you have; local, specific, actionable information predominates over theory. Most people, he agrees, segregate risk rather than integrating it — Ezekiel Emanuel shunning restaurants over COVID while riding a motorcycle — and are not meta-rational about it. Behaviour that looks irrational, like loss aversion over a stolen pair of headphones, is often rational on a higher plane, disciplining sloppiness or serving an old evolutionary purpose. The danger of success is flying too close to the sun: feedback degrades, sycophants gather, and criticism shuts off — his diagnosis of Elon Musk, tempered by noting Musk did hold some lines on tariffs and skilled immigration.

Elections, voting systems, and prediction markets

Silver has turned mildly against ranked-choice voting on two grounds: it is not Condorcet-optimal — the order in which candidates are eliminated can knock out a compromise winner like Brad Lander, who might have beaten anyone head-to-head — and it takes weeks to count. He wants an American norm of counting inside 24–48 hours, as India does across extraordinary diversity, and the crisp one-question exit polls of Europe and Latin America over the US’s hundred-item instrument. On proportional representation he leans yes as a critic of both parties, while granting that America’s many veto points have probably served capitalism and growth. Prediction markets, on which he consults for Polymarket, have grown more liquid and accurate but still misprice: they gave the Uganda-born, constitutionally ineligible Mamdani a 7 per cent shot at the nomination, and were overconfident on a papal market where they held no inside knowledge. He rates a third Trump term nearer 2 per cent than the 6–7 per cent sometimes seen, and questions whether such a scenario even involves a real election. Given only a name and zip code he can predict a voter with ~90 per cent accuracy, but elections hinge on 1–2 per cent margins, so blunt AI-over-Facebook models would not beat polling and its calibration record back to 1936; his one free bet for the 2028 Democratic nominee goes to AOC, with no one above 15 per cent.

AI progress, superforecasting, and the limits of AGI

Revisiting a year-old prediction, Silver stands by his estimate that AI is 10–15 years from matching human superforecasters, and rates the last year’s progress at the 40th percentile of what he expected; Cowen, who a year ago said two to three years and now says one to two, calls him far too pessimistic and points to gold-medal Math Olympiad performance. Silver draws his central distinction: static problems differ from dynamic systems whose inputs change constantly. LLMs are very bad at poker because it demands exploitative strategies developed fast against an evolving game-theory dynamic — dedicated solvers take minutes where a human computes implicitly in seconds — and he worries Math Olympiad success reflects teaching to a prestigious test, with answers perhaps latent in training data. He separates rough AGI for desk jobs from superintelligence on one side and physical-labour AGI on the other, and thinks people leap between them far too readily. He now says an LLM ‘thinks’ without embarrassment and grants that they reason, while doubting an emergent intelligence explosion. Training a thought process is hard: he can school himself on his artist partner’s eye but caps out at a B+, and ‘a B+ trader is often on the losing side of a trade’.

Sport — the NBA, betting, and coming out

Nine years after both men predicted more NBA players would come out as gay, the count since Jason Collins in 2013 is zero — against roughly 44 out players in the WNBA. Silver finds the gap genuinely surprising and offers selection effects (elite sport tracks athletes narrowly and early, so any identity distraction knocks you off the path) and a recent conservative backlash in sports. On betting, injury data for women’s sports is far worse than for men’s — his team had to run one-by-one Google searches for the women’s NCAA model — so a diligent modeller can still find soft lines before markets adapt, though any consistent winner gets limited quickly. On the draft lottery he dislikes incentivising failure and likes the Mike Zarren ‘wheel’, where each team picks first once every 30 years on a known schedule; American sports, he notes, are ‘socialist’ and European ones ‘capitalist’, yet fans happily tolerate perpetual soccer dynasties. Watching this year’s higher-octane playoffs converted him toward trimming the NBA season to around 72 games and capping individual load — perhaps 75 games a player with guaranteed, pre-published rest — since continuous high-intensity basketball strains bodies more than football, hockey, or baseball. He thinks the ‘only a few teams can win’ thesis is overstated, with the Nuggets, Raptors, and a late-blooming Giannis as counter-examples, and that fans over-index on the Bulls and Warriors dynasties.

LeBron, mentors, and betting on flawed talent

Silver reads LeBron James as, by career metrics, the best NBA player ever, yet unable to win hearts short of three more titles; he would send him to a young team like the Spurs to contend and mentor, arguing a Wemby-led San Antonio could plausibly win now — a claim Cowen would give 4–5 per cent and love to bet against. The disagreement crystallises their difference on flawed talent: Silver wants high variance and would bet on a player with a fixable problem (a healthy, motivated Luka Dončić) over one already maximising, because the upside is larger; Cowen bets against low conscientiousness. On mentors, Silver names Bill James (good writing about technical subjects is still good writing; an 8/8 writer-statistician can beat a 10/2 specialist), Richard Thaler, and Cowen himself, while noting he has always blazed his own trail because the election-forecasting product he wanted simply did not exist. For following AI he leans on Zvi Mowshowitz as comprehensive and level-headed — even though, Cowen teases, Mowshowitz thinks it may kill us all.

The rationalist bubble and West Coast culture

Silver calls the AI-safety and rationalist community a bubble: proudly weird even by weird standards, experimental about everything from polyamory outward, but underrating political constraints and human adaptability, and leaping too fast from rough AGI to physical-realm AGI to ASI. His P(doom) rose in the new preface — written in February and March, amid Musk’s White House influence and the tariff push — driven less by an intelligence explosion than by humans using AI to make weapons, terrorism, or dangerous compounds easier, the asymmetric-technology worry he associates with Nick Bostrom. He characterises EAs and rationalists as high in openness and high-variance in conscientiousness, prone to a gullibility that should have cost effective altruism more reputationally over Sam Bankman-Fried; poker players share the phenotype but are far more suspicious and street-smart. The West Coast, he adds, is diverging culturally — a Bay-Area house party may have no wine, Seattle feels almost Canadian — and doom-belief correlates with polyamory through a mix of hedonism and disconnectedness, though Cowen finds the purely economistic account too tidy.

Populism, immigration, and thermostatic politics

Why can Canada not win a Stanley Cup — none since 1992? Silver points partly to tax law: no state income tax in Florida, inside a hard salary cap, effectively pays players 10–15 per cent more. On the wider politics he mostly believes in thermostatic effects, opinion swinging against the status quo, and suspects populist sentiment may have crested in Canada and Western Europe even as Cowen sees it still rising elsewhere. On the puzzle of why European centrists will not simply deliver the lower immigration voters demand, he notes Denmark and Trudeau’s Canada have moved, and that the left was historically more restrictionist over welfare-state cost; he predicts the fertility crisis and ageing populations will reverse the politics within a few years toward courting skilled young workers. US politics, he argues, is unusually efficient at producing consistent ~50–50 coalitions, and America is currently quite pro-immigrant by world standards. On polling, voters side with the left on itemised immigration questions yet trust Trump more overall, because they fear progressive excess — ‘give them an inch and they’ll take a mile’ — a worry he traces back to COVID-era public-health overreach. He closes on what comes next: an NFL model, the thriving Silver Bulletin newsletter, and, eventually, another book, perhaps about sport.

See also

• Nate Silver — speaker
• Tyler Cowen — host
• Deciding Under Uncertainty — the theme this conversation most directly feeds
• Kill Criteria — Annie Duke’s decision-under-uncertainty tool, a close neighbour
• Risk Posture — how decision-makers set their appetite for risk
• Effective Altruism — the movement Silver reads as a high-variance bubble