- 87.3% of Polymarket wallets are net unprofitable (112,000 wallet analysis)
- 60%+ of invested capital lost on contracts priced under 10¢ (Bürgi, Deng & Whelan 2026)
- 5¢ contracts have a historical payoff rate of only 4% — negative EV before fees
- Takers lose 32% on average vs. Makers lose only 10% (same study)
- Taker fees consume up to 56% of gross profit on 15¢ contracts
- Zero-sum math: total fees collected = guaranteed aggregate negative P&L for all participants
The Profitability Numbers Are Brutal
Let's start with the data. Three independent analyses of prediction market participant performance converge on a consistent, uncomfortable picture:
The range across studies (12.7% to 30% profitable) reflects different methodology, time periods, and how "profitable" is defined. The consistent signal: the vast majority of participants lose money, and the structural reasons are well-documented and avoidable.
The Favorite-Longshot Bias: The Primary Culprit
The Favorite-Longshot Bias (FLB) is the single most studied and most consequential inefficiency in prediction markets. It was first documented in pari-mutuel horse racing by Ali (1977) and has since been confirmed across decades of sports betting data and, now, in granular prediction market data.
The documented tendency for market participants to systematically overprice low-probability contracts (longshots) and underprice high-probability contracts (favorites). In prediction markets, this means cheap contracts are consistently the worst bets per dollar invested — and expensive contracts are systematically underpriced by the crowd.
The landmark 2026 study by Bürgi, Deng & Whelan analyzed over 300,000 Kalshi contracts and quantified the FLB with precision:
Retail investors purchasing "cheap" longshot contracts priced under 10¢ lose more than 60% of their invested capital. A 5¢ contract has a historical payoff rate of only 4% despite its implied 5% probability. That's systematic negative ROI before accounting for any fees.
The Inverse: Favorites Are Underpriced
The FLB works in both directions. The same research found that contracts priced above 88–95¢ are systematically mispriced by approximately 4 cents in the pessimistic direction — the market assigns them lower probability than historical base rates justify.
This is the empirical foundation of Beatpoly's bond harvesting strategy: the crowd's bias toward dramatic outcomes means they undervalue boring, high-probability, "nothing-happens" contracts. We buy those.
| Price Range | Implied Probability | Actual Resolution Rate | EV Signal |
|---|---|---|---|
| Under 10¢ | Under 10% | ~4% (for 5¢ contracts) | Strongly Negative |
| 10¢ – 50¢ | 10% – 50% | Approximately accurate | Mixed — fees dominate |
| 50¢ – 85¢ | 50% – 85% | Near accurate | Neutral to slightly positive |
| 88¢ – 95¢ | 88% – 95% | ~4¢ higher than implied | Positive (underpriced) |
Why Traders Overpay for Longshots: The Psychology
The FLB is not random — it emerges from specific, predictable cognitive failures:
1. Probability Misperception (Prospect Theory)
Humans are hardwired to overweight small probabilities when a large gain is on the table. A 5% chance of winning 20× your money feels more exciting than a 95% chance of winning 5¢. The perceived value of the lottery ticket outweighs its mathematical value — every time.
2. Partition Dependence (The Ignorance Prior)
When a market has multiple outcome brackets, traders anchor to an equal-probability baseline: "there are 6 options, so each must be roughly 16.7% likely." This ignores actual physics, base rates, and all relevant information. It pulls rare outcomes up to the average — systematically overpricing them.
3. Salience Overreaction
Dramatic, sensationalized news triggers disproportionate price spikes in related markets. When a news headline screams about an event, retail traders buy YES on dramatic-sounding contracts, pushing prices above any rational probability assessment. The crowd is buying the narrative, not the math.
4. Hot/Cold Bias
Recent dramatic events cause traders to over-extrapolate. After a surprise political event, the market over-prices similar surprise events for weeks. After a calm streak, the market under-prices the possibility of surprises. Neither is rational — both are exploitable.
The Execution Layer: How Taker Fees Amplify Every Mistake
The FLB is damaging enough on its own. Combined with taker fee exposure, cheap-contract buyers face a compounding disadvantage that can consume more than half of their theoretical gross profit.
On a 15¢ Kalshi contract, taker fees consume approximately 6.6% of invested capital and up to 56% of a winning trade's gross profit. This means a theoretically 50/50 trade at 15¢ is not 50/50 after execution — it is heavily negative EV the moment the order fills.
The Bürgi, Deng & Whelan (2026) study captured this starkly: Takers lose 32% on average. Makers lose only 10%. The 22-percentage-point gap between these groups is almost entirely attributable to fee exposure and adverse selection — not to differences in information or ability.
| Trader Type | Order Type | Average Capital Loss | Primary Cost Driver |
|---|---|---|---|
| Retail Taker | Market orders | −32% | Taker fees + adverse selection |
| Retail Maker | Limit orders | −10% | Mispricings, no fee drag |
| Systematic Maker | Limit orders, Kelly-sized | Positive (top 12.7%) | Structural edges exploited |
The Mathematical Proof: Why the Average Must Lose
Regardless of skill, the mathematical structure of prediction markets guarantees that the average participant loses money. Here's the proof in three steps:
- Zero-sum base: In any binary market, the sum of YES and NO probabilities equals exactly 100%. Every dollar won by a YES holder is exactly balanced by a dollar lost by a NO holder. Total aggregate P&L before fees = $0.
- The fee tax: Platforms collect taker fees on executed orders. This creates a net negative cash flow out of the participant pool.
- Conclusion: Because aggregate P&L = $0 before fees, and fees > $0, aggregate P&L after fees must be negative. The average participant is mathematically guaranteed to lose money.
This is not a statement about skill. It's a statement about the structure. The fee extraction means that for every dollar paid out in profits, less than a dollar was paid in. The market is a slightly negative-sum game — and retail Takers absorb the majority of the fee burden.
The Path Out: What the 12.7% Do Differently
The profitable minority is not random. The research identifies consistent behavioral differences:
- They use limit orders (Maker execution). Zero taker fee exposure. On some categories, they collect rebates. This alone moves them from the 32% average loss bucket to the 10% bucket.
- They avoid low-price contracts. No 5¢ contracts. No longshot lottery tickets. They understand that the FLB makes cheap contracts the worst trades per dollar invested.
- They size positions correctly. Kelly Criterion or fractional Kelly. No flat betting, no gut-sizing. The math of survival is non-negotiable.
- They trade structural inefficiencies, not opinions. The Free Donut, the 88-Cent Rule, mean reversion after panic pricing — these are exploiting systematic crowd errors, not trying to predict outcomes better than the market.
The 87.3% who lose money are not losing to a house. They are losing to the 12.7% who operate systematically. The money flows from emotional, narrative-driven, market-order Takers to disciplined, data-anchored, limit-order Makers. Our job is to be in the right group — consistently, repeatably, without exception.
Frequently Asked Questions
Not blindly. The 72% NO baseline means that randomly selected NO contracts have a structural edge over YES, but the price of the NO contract must reflect the correct probability to give you positive EV. If a market has already priced NO at 75¢ (implying 75% probability of NO), but the true probability is 72%, that's a slightly negative EV trade. The edge comes from finding NO contracts that are mispriced — priced below their true probability — not from buying every NO contract indiscriminately.
Because market orders are immediate. A limit order might not fill — the price can move away before your order executes. Retail traders choose certainty of execution over fee efficiency, which is a classic short-term/long-term tradeoff error. Over 100+ trades, the taker fee drag ($70 per 100 trades at $10/trade vs. $0 for Makers) is the margin between profit and loss for borderline-edge strategies.
Adverse selection means that when you execute a market order, you are most likely to get filled when someone with superior information is on the other side of your trade. If a bot just received a model update showing the probability shifted dramatically, it immediately posts limit orders at the old price — and your market order hits those orders. You bought at the wrong price, the informed party sold at an advantageous price. This is the "Winner's Curse" in market microstructure theory.
The primary academic source is Bürgi, Deng & Whelan (2026) — a study of 300,000+ Kalshi contracts examining Maker/Taker performance gaps and the Favorite-Longshot Bias. Secondary sources include a wallet-level analysis of 112,000 Polymarket accounts, Reichenbach & Walther (2025) analyzing 124 million trades, and the foundational FLB paper by Ali (1977) from the horse racing context.