The Favourite-Longshot Bias, and Where It Shows Up

If you take every contract that ever traded at two cents and check how often those events actually happened, the answer across most studied markets is: less than two per cent of the time. The same exercise at the other end finds that contracts trading at 95 cents resolve Yes slightly more than 95% of the time.

This is the favourite-longshot bias. It has been documented in horse racing, sports betting, and event markets, across decades and jurisdictions. It is one of the few genuinely durable regularities in this field.

What it looks like

Plot implied probability on one axis and realised frequency on the other. A perfectly calibrated market produces a straight diagonal line. Real markets produce a line that sits slightly below the diagonal at the low end — things priced at 5% happen maybe 3% of the time — and slightly above it at the high end.

The size of the effect varies by venue and category. In deep, heavily traded markets it is small. In thin markets and in the long tail of exotic questions it can be large.

Why it happens

Several explanations, all partially true.

Risk-seeking at the bottom. A two-cent contract is a lottery ticket. Some people buy lottery tickets at prices that are bad in expectation because the shape of the payoff is what they want. That demand props up the price.

Nobody wants to be the seller. Selling a two-cent contract means committing 98 cents of capital to collect two. Annualised over months, that is an unattractive return with a small chance of total loss. Even participants who think the contract is worth one cent may not bother. The correction does not happen because correcting it is not worth anybody's time. Capital lockup is the underlying reason.

Attention concentrates on the exciting side. Long shots are where the interesting stories are. The flow arrives asymmetrically.

Fees bite hardest at the extremes. A fixed or near-fixed cost per contract is enormous relative to a two-cent position and trivial relative to a ninety-cent one. That asymmetry discourages exactly the trades that would close the gap.

Rounding. The minimum tick is one cent. A contract truly worth 0.4 cents cannot trade there; it trades at one cent, which is 150% of its value. At the low end the tick itself creates a floor.

Where it shows up in event markets

The long tail of candidate fields. In a market on which of twenty people gets some appointment, the fifteen people priced between one and three cents are, collectively, priced too high. The sum across all outcomes exceeding 100% is partly spread and partly this effect.

"Will X happen by date Y" markets far from the date. These sit low and stay low, with persistent buying from people who want the optionality.

Extreme threshold markets. The tails of a distribution of economic outcomes are typically a touch expensive.

Where it does not show up much

Deep, short-dated, heavily traded markets on scheduled events. Where there is real institutional participation and the horizon is short, the capital cost of taking the other side is small and the bias largely disappears. This is consistent with the explanation: the bias is strongest where correcting it is least worth anybody's time.

What to do with this

The honest answer is: mostly, use it as a reading correction rather than a strategy.

As a reading correction. When you see a market at 3%, read it as "probably a bit under 3%". When you see one at 96%, read it as "probably a bit over 96%". Do not quote tail probabilities from thin markets as though they were precise.

As a strategy, with care. The obvious move — systematically sell the tail — is exactly the trade the bias exists because nobody wants to do. You would be committing large capital for small, slow returns, with occasional complete losses, and paying fees that are proportionally largest on precisely these contracts. Several of those costs are the reason the mispricing survives. That is not a reason it cannot be done; it is a reason to run the arithmetic on fees, capital cost and variance before assuming a documented bias is an available return. The limits of any measured edge is the sober version of this.

The cross-venue version

The bias is not identical across venues, because the participants and the fee structures are not identical. Two venues can both exhibit it to different degrees, which produces a systematic difference in tail pricing that looks like an opportunity and is partly just two different populations of buyers.

This site grades cross-venue pairs and reports the difference as a theoretical cross-market edge precisely because that phrase is accurate: it is what the price difference implies before anything real happens to it. What a cross-market gap is actually worth works through what comes off the top.

Next

Why prices near 1c and 99c behave differently covers the mechanical reasons the extremes are their own world.