Hot Streaks Are a Lie. Your Trading App Is Built on Them.

In 1985, three psychologists asked basketball fans a simple question: if a player has made their last three shots, are they more likely to make the next one? Almost everyone said yes. Players, coaches, and broadcasters all agreed, a shooter who's "heating up" stays hot, at least for a little while. Then the psychologists pulled the actual shot-by-shot records of the Philadelphia 76ers and found something awkward: statistically, there was no such thing. A player who had just made three shots in a row was no more likely to make the next one than a player who had just missed three in a row.

That paper, Gilovich, Vallone, and Tversky's "The Hot Hand in Basketball," gave the bias a name...

the hot hand fallacy

And it became one of the most cited findings in behavioral science. Four decades later, it's having an unexpected second life, not in a lab, but on your phone. Prediction markets like Kalshi and Polymarket have turned the vague feeling of being "on a roll" into a literal, animated, push-notified streak counter, the same UI pattern that keeps people opening Duolingo, swiping on dating apps, and spinning slot machines. The question worth asking isn't just whether a hot hand exists. It's who benefits from you believing it does, right now, on this app.

Try it yourself

Below is a simulator with genuinely independent 50/50 outcomes: no hidden bias and no house edge baked in to make a point. Make predictions, build a streak, and watch the live panel track two numbers: your hit rate right after a win, and your hit rate right after a loss. Over enough trials, they converge. The streak will still feel meaningful while you're in it. That feeling is the whole article.

The fallacy, and its surprising rebuttal

The original hot hand study did three things. It surveyed fans and players, confirming the belief was close to universal. It analyzed the field-goal logs of the 76ers across a full season. And it ran a controlled experiment with Cornell's basketball team, shooting from a fixed distance under game-like conditions. In all three, the pattern was the same: the probability of making a shot after a hit was statistically indistinguishable from the probability of making a shot after a miss. Streaks existed, but they were the streaks you'd expect from any run of independent coin flips, not evidence of a hidden "hot" state.

For thirty years, that was the final word: humans are pattern-matching machines, wired by evolution to find signal in noise, and the hot hand was the textbook case of us finding a streak where only randomness seemingly lived.

Then, in 2018, two economists, Joshua Miller and Juan Sanjurjo, found a subtle bug in the method used to debunk the hot hand in the first place.

Here's the bug, in plain terms. If you flip a fair coin four times and look only at the flips that immediately follow a "heads," you are secretly biasing your sample. In short finite sequences, heads tend to be followed by tails more often than pure 50/50 odds would suggest, purely because of how conditional sampling works on small sets, it has nothing to do with the coin itself.

Gilovich, Vallone, and Tversky's method had exactly this blind spot, and it was pulling their estimates of "hit rate after a hit" systematically downward. When Miller and Sanjurjo corrected for it, a real, if modest, hot hand reappeared in the data, a few percentage points, not the dramatic heat-check narrative broadcasters love, but not nothing either.

This is the part most retellings of the hot hand story skip, and it's the most important part for anyone building an app, a trading strategy, or an argument around streaks: the "fallacy" wasn't that people believed in small effects it was actually that they believed in effects far larger than anything the data (even the corrected data) can support.

DiMaggio's 56 games, and the streak that "shouldn't" have happened

Basketball isn't the only sport that's been put under this microscope. In 1941, Joe DiMaggio hit safely in 56 consecutive games, a record that still stands, and one the biologist Stephen Jay Gould called the single most extraordinary statistical feat in the history of American sports. Gould's argument was elegant: given DiMaggio's career batting average, a streak that long should be a near-impossible outlier, evidence that something more than randomness was going on.

Later statisticians took Gould's challenge literally and simulated it. Using Monte Carlo models of the entire history of Major League Baseball, every player, every career batting average, and hundreds of thousands of simulated seasons. Researchers including Samuel Arbesman and Steven Strogatz found that a streak of 56 games or longer, by someone, at some point in baseball history, was close to a coin flip in likelihood. Not because DiMaggio wasn't exceptional (he was) but because when you give enough players enough chances, an outlier streak stops being surprising and starts being expected. It's the same logic behind why, in a room of just 23 people, there's better than even odds two of them share a birthday.

This is the clustering illusion's quieter cousin: it's not that streaks don't happen in random data. It's that random data is supposed to produce streaks, clusters, and droughts far more often than our intuition expects and when a streak finally lands on a real, nameable person, it feels meaningful in a way that a streak buried in a spreadsheet of coin flips never would.

Wall Street has the same problem, with real money attached

Nowhere does this matter more than in investing, where "hot hand" has a specific, expensive name: performance persistence. If a fund manager beats the market three years running, does that predict year four?

Mark Carhart's landmark 1997 study of mutual fund returns found almost no evidence that it does. Funds with hot streaks underperformed in the following period about as often as they outperformed, once you accounted for the handful of structural factors such as fees and momentum exposure, survivorship bias in which funds even stick around to be studied, that look like skill but aren't. The one truly persistent pattern Carhart found wasn't a hot hand at all, it was expense ratios: cheap funds tend to stay cheap, and expensive funds tend to stay expensive, and that predicts future returns far better than any winning streak does.

This hasn't stopped the entire fund marketing industry from printing five-star ratings and "top decile, 3-year returns" on brochures, a hot-hand pitch dressed up as "due diligence", with a regulatory disclaimer buried at the bottom doing the opposite job of the headline above it.

Enter the app that turns your hit rate into a cute badge

Kalshi and Polymarket are event contract exchanges, you're trading a "yes" or "no" position on whether something happens, ranging from an interest rate decision to who wins a soccer match, and the contract settles at $1 or $0. These systems have grown explosively: Kalshi's trading volume rose well over 1,000% year-over-year through 2025, its monthly active users climbed into the millions, and the company's valuation moved into double-digit billions of dollars. Polymarket, after regaining full access to U.S. users through a 2025 regulatory settlement, scaled from roughly 40,000 traders in early 2024 to well over two million today, and now distributes its odds directly through newsrooms and finance apps. Sports contracts alone drive a majority of Kalshi's volume, this is not a niche hobby for policy wonks anymore. It's a retail trading app with sportsbook-sized engagement numbers.

And like any retail trading app, it is designed, deliberately, to feel good to use. Both platforms lean on the exact interface patterns behavioral researchers have spent two decades flagging in gambling and social apps: win streak counters, leaderboards ranked by recent performance, push notifications timed around your open positions, and variable, unpredictable payouts that land on a schedule no different from a slot machine's. None of these features are illegal, and none of them are lying to you about the odds on any single contract. What they're doing is more subtle: taking a cognitive bias researchers spent forty years proving is mostly an illusion, and building a UI whose entire job is to make that illusion feel like insight.

A streak counter doesn't just show you information. It tells you a story about yourself, that you're reading the market correctly, that you've found a rhythm using the exact psychological hook that made Gilovich, Vallone, and Tversky's 1985 survey respondents so confident, and so wrong, about basketball players forty years earlier.

Why our brains do this

Psychologists trace the hot hand fallacy back to the representativeness heuristic, our tendency to judge whether a sequence is "random" by whether it looks the way we imagine randomness should look. True random sequences cluster far more than people expect; a genuinely random coin-flip sequence of twenty flips will very often contain a run of four or more heads in a row, but when people are asked to generate a "random-looking" sequence by hand, they systematically alternate too often and avoid exactly these clusters, because a cluster doesn't feel random to us even though it's a completely normal product of chance.

It's worth separating the hot hand fallacy from its closer cousin, the gambler's fallacy, because they point in opposite directions and people mix them up constantly. The gambler's fallacy is the belief that a streak is "due" to end and that after five reds in a row at roulette, black is overdue. The hot hand fallacy is the belief that a streak is "due" to continue and that the shooter who just hit three threes is about to hit a fourth. Both are wrong for the same reason (independent events have no memory), but they feel like opposite instincts, and which one takes hold often depends on whether the streak belongs to an impersonal machine (roulette, which feels "due" to correct) or a skilled agent (a shooter, a trader, a fund manager, who feels "due" to continue being good). Prediction market UX is built almost entirely around triggering the second one.

None of this means streaks are meaningless, or that skill doesn't exist. Shooters have better and worse nights for real physiological reasons and fatigue, rhythm, defensive attention. Traders and fund managers do sometimes have genuine, if modest, edges. The Miller-Sanjurjo correction is a reminder that the mathematically rigorous answer to "is the hot hand real" was never a flat no, it was "yes, a little, and far less than it feels like from the inside."

The honest reaction to a five-win streak on a prediction market isn't "I've cracked it," and it isn't "this is definitely about to end" either. It's closer to: this is what randomness looks like when it happens to you personally and the app showing you a flame emoji right now has a business model that benefits from you forgetting that.