The favorite-longshot bias, explained

By the Cent Signals editorial desk. Reviewed June 9, 2026.

A price on a prediction market reads as an implied probability, and across most of the range it is a respectable estimate. But the deep tails carry a known, well-replicated distortion: long shots tend to trade above the frequency at which they actually happen, and heavy favorites slightly below. The pattern is called the favorite-longshot bias, and it is one of the oldest findings in the empirical study of market odds. This guide describes the evidence, the explanations, and what the pattern means when reading a one-cent market on Polymarket. It builds on how to read implied probability, and like everything on this site it is description, not direction.

What the pattern looks like

Take a large set of markets and group them by price. If prices were perfectly calibrated, outcomes priced at five cents would resolve YES about five percent of the time, outcomes at fifty cents about half the time, and outcomes at ninety-five cents about ninety-five percent of the time. The favorite-longshot bias is the systematic departure from that line at the edges: the five-cent group historically resolves YES less often than five percent, while the ninety-five-cent group resolves YES slightly more often than ninety-five percent.

The middle of the range is largely unaffected. Calibration studies of market odds repeatedly find prices between roughly twenty and eighty cents tracking realized frequencies closely; the distortion concentrates in the tails, and it grows as the price approaches the extremes.

A century of evidence

The pattern was first measured in racetrack odds, with academic work dating to the late 1940s, and it has since been replicated across roughly a hundred years of odds data, in sports markets, and in prediction markets. Its persistence is what makes it notable: most pricing quirks disappear once they are widely known, but this one has survived decades of publication.

The size of the effect is not constant. It is strongest in venues with high fees and casual participation, and weaker where costs are low and active participants can correct mispricing cheaply. Studies of low-fee exchanges find a measurably smaller bias than the racetrack literature, which matters for reading a platform like Polymarket: the tendency is real, but its magnitude on any given venue is an empirical question, not a constant.

Why it happens

Taste for lottery-like payoffs. Some participants accept a poor expected return for a small chance at a large multiple, which keeps demand for cheap tail outcomes higher than their frequency justifies.
Overweighting of small probabilities. Behavioral research finds people systematically treat a two percent chance as larger than it is, a result formalized in prospect theory, and prices set by such participants inherit the distortion.
The cost of correcting the tails. A participant who judges a two-cent contract to be worth one cent can gain at most a cent per share, must lock capital until resolution to collect it, and carries the risk of the long shot landing anyway. Small distortions in the tails are expensive to iron out, so they persist.

What it means for reading Polymarket tails

A YES price of one or two cents reads, mechanically, as a one or two percent implied probability. The favorite-longshot literature adds a quiet footnote to that reading: across large samples, prices that deep in the tail have historically overstated how often the outcome arrives. A two-cent market is therefore better read as "the market prices this around two percent, and tail prices of this kind have tended to run above realized frequency" than as an exact two percent.

That footnote cuts both ways and stays statistical. It does not make any particular cheap market wrong, and it says nothing about which long shot is the exception; some resolve YES. On Cent Signals the pattern is part of why the heuristic behind markets worth a second look treats deep-tail prices with extra care, and why a tail price is flagged only when other public facts, such as heavy volume close to a resolution date, sit oddly beside it. For what those activity figures measure, see volume vs liquidity on Polymarket.

Frequently asked questions

What is the favorite-longshot bias?

It is a documented statistical pattern in wagering and prediction markets: outcomes priced as long shots tend to happen less often than their price implies, while strong favorites tend to win slightly more often than their price implies. In prediction-market terms, a contract at two cents may correspond to a true frequency below two percent, and one at ninety-five cents to a frequency a little above ninety-five percent.

Where has the favorite-longshot bias been documented?

First in racetrack wagering data, in research going back to studies of the 1940s, and since then across roughly a century of odds data, sports markets, and prediction markets. The effect is one of the most replicated findings in the empirical literature on market odds, though its size varies by venue and era, and it is measurably weaker in markets with low fees and active arbitrage.

Why do long shots trade above their true frequency?

The leading explanations fall into two families. Preference-based accounts hold that some participants value lottery-like payoffs, accepting a worse expected return for a small chance of a large win, which keeps tail prices elevated. Friction-based accounts note that correcting a one-cent overpricing ties up capital until resolution for at most one cent of gain, so small distortions in the tails are not worth the cost of fixing. Both forces can operate at once.

Does the bias mean every cheap market is overpriced?

No. The bias is a statistical tendency across many markets, not a verdict on any single one. Plenty of long shots are priced exactly as cheaply as they deserve, and some resolve YES. The pattern only says that, averaged over a large set of tail-priced outcomes, the realized frequency has historically come in below the implied one. Any individual market still has to be read on its own facts.