- Contracts under 10¢ lose 60%+ of capital on average — Bürgi, Deng & Whelan (2026), 313,972 Kalshi contracts
- 5¢ contracts have a true payoff rate of only 4% despite 5% implied probability — negative EV before fees
- Inverse FLB: contracts at 88–95¢ are mispriced by ~4¢ — true probability is systematically higher than price
- Ali (1977) first documented FLB in horse racing. Snowberg & Wolfers (2010) confirmed across 5 million races
- Manski (2006) proved mathematically that binary payoffs + budget constraints inevitably produce longshot overpricing
- NO buyers earn +0.83% dollar-weighted returns. YES buyers lose −1.02% (Polymarket empirical data)
What Is the Favorite-Longshot Bias?
The Favorite-Longshot Bias (FLB) is the single most documented inefficiency in prediction markets. It has one implication you can trade on immediately: cheap contracts are systematically overpriced, and expensive contracts are systematically underpriced.
In practical terms: contracts priced under 10¢ are consistently worse bets than their price suggests, and contracts priced above 88¢ are consistently better bets than their price suggests. This is not an opinion — it is supported by analysis of over 300,000 Kalshi contracts (Bürgi, Deng & Whelan, 2026) and confirmed across decades of horse racing data.
Retail investors purchasing contracts priced below 10¢ lose more than 60% of their invested capital on average. A 5¢ contract has a historical payoff rate of only 4% despite its 5% implied probability — systematic negative ROI before fees. (Bürgi, Deng & Whelan, 2026, 313,972 Kalshi contracts)
Academic Origins: From Horse Racing to Prediction Markets
The FLB was first documented by Ali (1977) in pari-mutuel horse racing, where bettors consistently overpaid for long-shot horses relative to their true win rates. The finding was confirmed across 5 million horse races by Snowberg & Wolfers (2010), who traced the bias to probability misperception as defined in Kahneman & Tversky’s Prospect Theory — humans overweight small probabilities when a large gain is possible.
Manski (2006) provided a mathematical proof that the combination of binary payoffs and budget-constrained traders inevitably produces longshot overpricing in equilibrium. The bias is not a quirk — it is a structural feature of how humans interact with lottery-like bets.
Bürgi, Deng & Whelan (2026) brought this to prediction markets specifically, analyzing 313,972 Kalshi contracts and confirming the same pattern with greater precision than the horse racing literature. This is the most comprehensive prediction market microstructure study to date.
The Full Probability Spectrum
| Price Range | Implied Probability | True Resolution Rate | EV Signal | Trade Action |
|---|---|---|---|---|
| Under 5¢ | <5% | ~2–3% | Strongly negative | Never buy YES |
| 5¢–10¢ | 5–10% | ~4–7% | Negative (FLB peak) | Avoid |
| 10¢–50¢ | 10–50% | Near implied | Mixed — fee-sensitive | Only with model edge |
| 50¢–85¢ | 50–85% | Near implied | Neutral to positive | Situational |
| 88¢–95¢ | 88–95% | ~4¢ above implied | Positive (inverse FLB) | Core NO bond zone |
The Inverse FLB: Favorites Are Underpriced
The FLB works in both directions. While longshots are overpriced, favorites are underpriced. Research confirms that contracts priced above 88¢ are systematically mispriced by approximately 4 cents pessimistically — meaning their true resolution probability is consistently higher than the market price implies.
An 88¢ contract often has a true resolution probability of ~92%. That is a +$0.04 EV edge per contract before fees — the structural basis of the 88-Cent Rule strategy covered in Chapter 4.
Why Traders Overpay for Longshots: The Psychology
1. Probability Misperception (Prospect Theory)
Humans overweight small probabilities when a large gain is on offer. A 5% chance of winning 19× your money feels more valuable than it mathematically is. The perceived excitement of the longshot exceeds its mathematical expectation every time.
2. The Lottery Ticket Effect
Cheap contracts are psychologically appealing: small downside, large upside. This is identical to why people buy lottery tickets despite deeply negative expected value. The asymmetric return profile overrides rational probability assessment.
3. Narrative Overreaction
When sensationalized news breaks — a shock weather forecast, a dramatic political development — retail traders buy YES on dramatic outcomes, pushing prices above any rational probability. The crowd is pricing the narrative, not the math. This creates tradeable spikes.
4. Partition Dependence (the 1/N Heuristic)
In multi-bracket markets (like weather temperature bands), traders anchor to an equal-probability baseline: “there are 6 options, so each is about 16.7% likely.” This ignores physics and base rates entirely, pulling rare tail outcomes up to the average. Tail brackets worth 2–3% get priced at 5–10¢. The center brackets — the genuinely probable outcomes — get underpriced by 5–13 percentage points.
Dollar-weighted returns on Polymarket show: NO buyers earn +0.83% on average. YES buyers lose −1.02%. The FLB creates a structural, persistent edge for systematic NO sellers — the basis of the 72% harvesting strategy.
Fee Interaction: The Compounded Disadvantage
The FLB does not operate in isolation. On Kalshi, taker fees on a 15¢ contract consume roughly 6.6% of invested capital. The combination of FLB (overpriced longshots) and taker fees creates a compounded negative EV for retail longshot buyers:
- True probability: 10% (vs. 15¢ implied probability)
- Base EV before fees: −$0.05
- Taker fee: −$0.007 additional
- Total EV: −$0.057 per contract
This is why the 87.3% loss rate among retail participants is not surprising. The most natural and instinctive trade — buying the cheap exciting contract — is the worst trade by a wide margin.
Frequently Asked Questions
Yes — but only when you have specific private information suggesting the true probability significantly exceeds the market price. The FLB means the bar is high: if a contract trades at 5¢, your model needs to estimate at least 7–8% true probability before fees make it neutral EV. Without specific information advantage, buying under 10¢ is a losing strategy by the data.
Yes. The 72% NO resolution rate and the underpricing of favorites has been confirmed across both platforms. The magnitude of the FLB may vary by market category — sports markets often show less FLB than political or weather markets — but the directional bias is consistent.
The inverse FLB creates a structural edge in the 88–95¢ range, but the recovery rate math still matters. At 97¢, one loss wipes out 33 wins. The 88¢ sweet spot balances positive EV with a sustainable recovery rate. Higher prices offer less edge, not more.