July 13, 2026 · Polymarket Select

The Favorite-Longshot Bias: What 70 Years of Research Actually Says

One of the most durable findings in the economics of betting is also one of the least intuitive: across horse tracks, sportsbooks, and prediction markets, participants tend to overpay for longshots and underpay for favorites. It has been documented for over seventy years, across countries and eras, in markets with millions of dollars at stake. This post is a plain-English tour of that literature — what the pattern is, how strong the evidence is, and what economists think causes it. It is a review of public research, nothing more.

What the bias actually claims

Strip it to the core. If betting markets were perfectly efficient, then contracts (or horses, or teams) priced at a given probability would win at exactly that rate over the long run. A basket of 20-cent longshots should win 20% of the time; a basket of 90-cent favorites should win 90% of the time.

The favorite-longshot bias is the empirical observation that this does not hold symmetrically. Instead:

Longshots win less often than their prices imply. The crowd pays 20 cents for something that really happens, say, 13% of the time. Longshot bettors, as a group, systematically lose more than the odds alone would suggest.
Favorites win more often than their prices imply. The near-certain side is priced a touch too cheaply, because comparatively few people want the "boring" bet.

The gap is not enormous on any single wager, and it does not make favorites a free lunch — favorites still lose their share of the time. But as a statistical regularity across thousands of settlements, it is remarkably stubborn.

Seventy years of evidence

The pattern was noticed early. R. M. Griffith documented it in horse-race odds in 1949, and W. T. Weitzman followed in 1965. These were among the first careful looks at how implied probabilities from betting odds compared to realized win rates.

The modern treatment begins with Mukhtar Ali (1977), "Probability and Utility Estimates for Racetrack Bettors" in the Journal of Political Economy, which formalized the mismatch between subjective and objective odds. Richard Thaler and William Ziemba (1988), in the Journal of Economic Perspectives piece "Anomalies: Parimutuel Betting Markets," pulled the racetrack evidence together and put the bias squarely on the behavioral-economics map. Raymond Sauer's (1998) survey, "The Economics of Wagering Markets" in the Journal of Economic Literature, extended the review across sports betting and confirmed the regularity was not a quirk of one venue.

Crucially, the bias is not confined to horse racing. It shows up in sportsbook point-spread and moneyline markets, in lottery-style products, and — relevant for 2026 — in prediction markets, where the same overpricing of unlikely outcomes appears in the market for low-probability contracts. The through-line across all these venues is the same: the unlikely side is chronically too expensive.

The leading explanations

If the pattern is this robust, why does it persist? The literature offers several non-exclusive answers, and the honest summary is that it is probably a mix.

1. Risk-love (utility for gambling). The classic explanation, going back to Ali, is that some bettors simply enjoy risk. A longshot is a cheap ticket to a large, exciting payoff. If enough participants get utility from the possibility of a big win — beyond its expected monetary value — they will bid longshots above fair value, the way people knowingly overpay for lottery tickets.

2. Misperception of small probabilities. Behavioral work, including prospect theory (Kahneman & Tversky), shows people systematically overweight small probabilities. A 2% chance feels more like 8%. If bettors misjudge how unlikely a longshot really is, they will overpay for it — no love of risk required, just a predictable perceptual error.

3. Information and adverse selection. Ottaviani and Sørensen (2008) developed theoretical models in which the bias emerges from how privately-informed and uninformed traders interact and how late information arrives. Under certain structures, an equilibrium favorite-longshot pattern falls out of the trading process itself.

The landmark attempt to distinguish these was Erik Snowberg and Justin Wolfers (2010), "Explaining the Favorite-Longshot Bias: Is it Risk-Love or Misperceptions?" in the Journal of Political Economy. Using a very large dataset of horse-race bets, they tested which story better fit the data and concluded the evidence leaned toward misperceptions of probability rather than pure risk-love — though, as with most things in social science, the debate did not end there.

Why the bias survives arbitrage

A fair objection: if favorites are underpriced, why don't sharp bettors pile in and erase the edge? Several frictions keep it alive:

The edge is thin per bet. Underpayment on a favorite is small in percentage terms. It only compounds across many independent wagers, which requires patience and volume most casual bettors lack.
The unlikely side is where the fun is. The behavioral demand for longshots is persistent — new participants keep arriving with the same preferences, continuously re-supplying the mispricing.
Fees, spreads, and limits. Transaction costs and betting limits eat into the modest favorite edge, and structural frictions restrict how much any one participant can push.
It requires discipline, not genius. Harvesting a thin, frequent edge means sizing small, spreading across uncorrelated bets, and holding through variance — behavior that is psychologically hard even when the math is clear.

That last point is the quiet reason the anomaly has outlived its own discovery: knowing about a bias and having the temperament to exploit it responsibly are very different things.

What this means for a curious trader in 2026

The favorite-longshot bias is not a secret and not proprietary — it is in journals you can read for free, replicated across decades. What it offers a modern prediction-market participant is a lens, not a strategy: when you look at a market, ask whether the unlikely side is priced for excitement rather than reality, and whether the "boring" side is quietly cheap.

That lens is not a license to back every favorite — plenty of favorites are correctly priced or even overpriced, and any position below 100% loses sometimes. The literature describes a statistical tendency across large samples, not a guarantee on the next contract. Turning a documented bias into a survivable approach is entirely about sizing and diversification, which is a separate discipline we cover in How to Size Bets You Can Survive.

If you are new to the mechanics, What Is a Prediction Market? sets the stage, and Prediction Markets vs Polls covers why market prices are worth reading in the first place.