Underdog Betting on Polymarket: When 10 to 25 Cent Contracts Are Actually Cheap
The break-even math for cheap underdogs, why public teams and political longshots are different bets, and how to stake a tail that usually loses.

Most underdog buying on Polymarket is negative expected value. The favorite-longshot bias is one of the most replicated patterns in betting markets, and the 2026 Polymarket evidence points the same way: cheap contracts have tended to lose more than their price implies. That does not make every 10 to 25 cent contract bad. It makes the default bad and puts the burden on you to show why this underdog is different.
The exceptions are structural: a public team that retail money overbets, a political longshot with a broken resolution rule, or a sports underdog in a market priced by a narrow book. "It looks cheap" is not a reason.
The break-even math
Polymarket charges takers a fee of share count times the rate times price times one minus price. Politics and finance use 0.04, sports 0.05, crypto 0.07, and geopolitics is fee-free. Makers pay nothing.
At 15 cents, the politics fee is 0.04 times 0.15 times 0.85, or 0.51 cents. The all-in cost is 15.51 cents. A winning share pays $1, so the net gain is 84.49 cents and the break-even probability is 15.51 percent. At 10 cents the fee is 0.36 cents and break-even is 10.36 percent; at 25 cents the fee is 0.75 cents and break-even is 25.75 percent.
The spread is worse. A one cent spread on a 15 cent contract is 6.7 percent of the stake; two cents is 13.3 percent. If you buy at 16 when the midpoint is 15, you need a true probability above 16.54 percent. On a 25 cent contract a one cent spread is 4 percent, which is manageable. On a 5 cent contract it is 20 percent, which is not.
Where the bias comes from
Thaler and Ziemba documented the pattern in parimutuel racetracks and state lotteries in 1988: longshots were overbet relative to their chance of winning, and favorites were underbet. Snowberg and Wolfers later tested the two main explanations on a large horse-racing dataset and found more support for misperception of small probabilities than for pure risk-love.
The prediction-market version has the same shape. The 2026 study by Cardozo and Rivero-Wildemauwe looked at 588 million trades and found purchases below 10 cents lost 19.3 cents per dollar. The loss was 6.3 cents when every contract was weighted equally, and the sign flipped when related contracts were grouped by parent event. The pattern was robust in crypto and politics but absent in sports.
That is why most underdog buying loses: the price already contains the bias, and the fee and spread sit on top. You need a specific reason the market is wrong.
Public-team bias in sports
In sports, the underdog is often a team rather than a number. Fan money tends to back the popular team, the national brand, or the side with the louder story. If a venue's flow is retail-heavy, the popular side can trade above fair and the unpopular underdog below fair. That is the public-team bias, and it is one reason two sportsbooks can disagree.
On Polymarket the effect is weaker than folklore suggests. The 2026 study found no aggregate favorite-longshot bias in sports, consistent with prices set by high-volume participants. A public team at 20 cents may be inflated by a cent or two, but do not assume a 5 cent discount because the opponent is boring. The screen is one-sided flow plus a price that disagrees with a sharper venue.
Sports underdogs versus political longshots
A sports underdog has a fixed rulebook, a scheduled start, a short settlement window, and a result that is usually unambiguous. The game repeats hundreds of times per season, so the market has a deep history; the main risks are injuries, weather, and a line that moves before kickoff.
A political longshot has a long lockup, a source that may be slow or ambiguous, and a payoff that depends on a single event. It can sit at 15 cents for months, then gap to zero or 80 cents in a day, and a 50/50 resolution is a real cost because a contract that can be voided is not a binary bet.
That changes the hurdle. A one cent edge can be enough on a sports underdog if you can repeat it across many games. A political longshot needs a larger edge for the lockup, the resolution risk, and the correlation among contracts in the same election.
Correlated tails in the same event
The most common mistake is buying a basket of cheap contracts in one event and calling it diversification. Suppose three mutually exclusive outcomes trade at 15, 12, and 10 cents. The basket costs 37 cents. If true probabilities are 18, 13, and 9 percent, the sum is 40 percent and the gross edge is 3 cents, or 8.1 percent. Fees are about 1.29 cents, leaving 1.71 cents, or 4.6 percent. That is thin for a bet that pays nothing in the 60 percent of states where none wins.
If the outcomes are not mutually exclusive, the arithmetic is worse. The chance at least one wins is not the sum of probabilities; it is one minus the product of the complements. Ten independent 15 cent underdogs have an 80 percent chance that at least one wins, but the expected number of winners is 1.5 and the basket costs 1.50 per dollar of payoff. Before fees it is fair; after fees it is negative. That is why "buy every underdog" fails.
Cap exposure by event, not by ticket. Three tails in one election are one bet.
Staking a bet that usually loses
A 15 cent underdog wins about one time in seven if the market is fair; a 25 cent underdog wins about one time in four.
Use a flat 0.25 to 0.5 percent of bankroll per ticket. On $10,000 that is $25 to $50. A 20-ticket basket at the top of that range is $1,000, or 10 percent of the account, which is too much. Cap the basket at 2 to 3 percent of bankroll and accept that most baskets go to zero.
Kelly is a poor guide because q is uncertain and the payoff is skewed. Full Kelly on a 15 cent contract with a 20 percent true probability is (0.20 minus 0.15) divided by 0.85, or 5.9 percent of bankroll; a quarter is 1.5 percent. If the true probability is 16 percent, the one point edge is a donation. Size to the low end. Never martingale; doubling after a loss turns independent bets into one bet on your bankroll.
How to execute
- Write the thesis in one sentence, including the true probability and why the market is wrong. If you cannot, pass.
- Compare the price with an independent reference: a sportsbook line, a poll average, a base rate, or a related market.
- Check the spread and depth at your size. Use a limit order at or below fair value; do not cross a wide spread.
- Cap the ticket at 0.5 percent of bankroll and the event at 2 percent. Count correlated contracts as one position.
- Set a time stop. If the catalyst passes and the price has not moved, exit or accept that you are holding a lottery ticket.
- Record entry, closing price, and outcome, and compare your hit rate with the break-even rate for the price bucket.
Risks
The obvious risk is that the underdog loses, which is the normal case. The subtler risk is that it wins big and you size up. A 15 cent ticket pays 5.7 to 1, so one hit can make a month look profitable; the next ten tickets can give it back.
Resolution risk is larger in politics. A 50/50 resolution returns half the stake, which looks like a partial win in your records.
Checklist
- Is the all-in cost, including fee and spread, below my true probability by a clear margin?
- Can I name the specific reason the market is wrong, not just the price?
- Is this a sports underdog with a clean rule, or a political longshot with a long lockup?
- Are the other tickets in this basket correlated with this one?
- Is the stake at or below 0.5 percent of bankroll, with the event capped at 2 percent?
- Do I have a time stop and a record of the closing price?
Underdogs are not cheap because they are unpopular. They are cheap because they usually lose. The occasional mispricing is real, but it is a narrow exception that requires a reference price, a clean rule, and a stake small enough to survive being wrong.
Sources
- 1.The Favorite-Longshot Bias in Prediction Markets: Evidence from Polymarket— arXiv
- 2.Explaining the Favorite-Longshot Bias: Is it Risk-Love or Misperceptions?— National Bureau of Economic Research
- 3.Anomalies: Parimutuel Betting Markets: Racetracks and Lotteries— Journal of Economic Perspectives
- 4.Fees— Polymarket Documentation
- 5.Prices & Orderbook— Polymarket Documentation