The 90-Cent Favorite Strategy on Polymarket: Yield, Fee Drag and Ruin Math
How buying 85 to 95 cent contracts actually pays, why a 99 cent favorite is a different trade, and the sizing that survives one bad resolution.

Buying 85 to 95 cent contracts on Polymarket is a yield trade, not a forecast. You are lending capital against a high-probability outcome and collecting the last few cents before resolution. The question is whether those cents cover the fee, the spread, resolution risk, and the cost of locked capital. Often they do not cover all four.
At 90 cents the arithmetic looks generous. At 99 cents it looks safe. Neither impression survives one loss at size. The break-even math, the fee drag, and the recovery ratio decide whether a favorite is cheap or just expensive safety.
Break-even is not the quoted price
Polymarket charges takers a fee of share count times the category 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 90 cents, one share costs 0.04 times 0.90 times 0.10, or 0.36 cents. The all-in cost is 90.36 cents. A winning share pays $1, so the net gain is 9.64 cents, a 10.7 percent return. The break-even probability is 90.36 percent.
At 95 cents the fee is 0.19 cents, the net gain is 4.81 cents, and the break-even probability is 95.19 percent. At 99 cents the fee is 0.04 cents, the net gain is 0.96 cents, and the break-even probability is 99.04 percent. These are hold-to-resolution numbers; an early exit adds a second fee and a second spread.
90 cents versus 99 cents
The two prices look like the same trade with a different haircut. The recovery math is the difference.
At 90 cents, a loss costs 90 cents and a win makes 9.64 cents, so you need 9.3 wins to recover one loss. At 95 cents you need 19.8 wins. At 99 cents you need 103 wins. The higher price has a higher hit rate, but the loss-to-win ratio grows faster. A 99 cent favorite at 99 percent true probability is roughly break-even after fees. At 98.5 percent it is a losing business with a pleasant win rate.
With fixed dollar stakes the bankroll effect is clearer. On a $10,000 account, twenty $500 positions at 95 cents is the whole account exposed. One loss costs 5 percent; three losses cost 15 percent and require about 60 winning positions to recover. The cents are compensation for a low-probability, high-severity event, not passive income.
Fee drag and spread drag
At 90 cents the 0.36 cent fee is 3.6 percent of the 10 cent gross edge. At 95 cents the 0.19 cent fee is 3.8 percent of the 5 cent edge. At 99 cents the 0.04 cent fee is 4 percent of the 1 cent edge. The percentage is stable, but the absolute edge shrinks, so fixed costs matter more.
The spread is bigger. A one cent spread on a 99 cent contract means you buy at 99 and sell at 98. The half-spread is half the 0.96 cent net gain. A two cent spread makes the trade negative before the event starts. Polymarket shows the midpoint unless the spread is wider than $0.10, when it shows the last trade, so the screen can understate the cost.
Category rates change the answer. A crypto favorite at 95 cents pays 0.33 cents versus 0.19 cents in politics. If your edge is half a cent, the category rate decides whether the trade exists.
Where favorites are genuinely cheap
The 2026 study of 588 million Polymarket trades by Cardozo and Rivero-Wildemauwe found purchases at or above 90 cents earned 0.83 cents per dollar, while purchases below 10 cents lost 19.3 cents. That supports the average favorite, not every favorite.
The pattern was robust in crypto and politics but absent in sports. Sports closing lines are deep and competitive, so a 90 cent favorite is usually fair. Politics can embed a resolution discount: a long lockup, an ambiguous rule, a slow source. Crypto carries a higher fee and gap risk, so a 90 cent threshold contract is a different asset from a 90 cent election contract.
The test is not the price bucket. A 90 cent contract with a 93 percent true probability has a 3 cent edge before fees and 2.64 cents after. At 91 percent the edge is 0.64 cents, which a one cent spread erases. The price is the market's view, not the size of your advantage.
Why buying every favorite fails
Selection comes first. The favorites that screen as cheap often carry a hidden defect: a rule that can resolve 50/50, a deadline that can slip, a wide spread, or a counterparty who knows more. Clean favorites are priced accordingly.
Correlation is second. Ten 95 cent contracts on one election, storm, or crypto threshold are one bet. If the shared factor fails, all ten lose together.
Negative skew is third. The strategy wins most months and gives back a year in one event. That is tolerable only when the size is set in advance and the loss does not change your behavior.
Lockup is fourth. A 95 cent contract that resolves in six months earns 5.05 percent, about 10 percent annualized if capital could be redeployed instantly. It cannot. Compare the yield with the next best risky use of the capital, not with cash.
Sizing a favorite book
The Kelly fraction for a binary buy is (q minus p) divided by (1 minus p). At p equal to 0.95 and q equal to 0.97, that is 0.02 divided by 0.05, or 40 percent of bankroll. The number is unusable because q is uncertain, bets are correlated, and you cannot repeat without limit.
Use a fraction of Kelly and a hard cap: 1 to 2 percent per position, 5 to 6 percent per event cluster, and at least 30 percent in cash. Size to the low end of your probability range. If you think the true probability is 96 to 98 percent, use 96 in the formula. The midpoint is how a good process produces a bad quarter.
How to execute
- Compute the all-in cost at the size you can fill, including fee and half-spread. Never use the midpoint as your entry price.
- Check depth on both sides. A 200 share ask with no bid is not a 95 cent contract for a $5,000 order.
- Prefer a maker order when the spread is wide, but assume adverse selection. Your bid fills when the market moves against you.
- Read the resolution rule and the named source. A 95 cent contract that can resolve 50/50 is not a 95 cent contract.
- Record entry, closing price, fee, and outcome so you can tell skill from a lucky streak.
Risks
The main risk is a large loss after you size up because the win rate felt safe. Resolution risk, correlated events, a widening spread, and a fee change can arrive together. The exit bid can also vanish when you need it.
A strategy that earns 5 percent per trade with a 95 percent win rate can look excellent for months and still be negative after one 50/50. The record does not identify the outlier until it happens, which is why the cap matters more than the forecast. The 0.83 cent estimate is also a sample from one platform and one period; it can shrink as capital arrives.
Checklist
- Is the all-in cost below my probability estimate by at least one cent?
- Is the resolution rule unambiguous, with a named source and a firm deadline?
- Is this position correlated with others in the same event or category?
- Is the size at or below 2 percent of bankroll and 6 percent per event cluster?
- Can I hold to resolution if the market becomes illiquid?
Favorites are not free money. They are a yield trade with a tail. The traders who do well buy a price that is too low for the risk and size it so one bad resolution is an inconvenience, not an ending.
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.Fees— Polymarket Documentation
- 4.Prices & Orderbook— Polymarket Documentation
- 5.Resolution— Polymarket Documentation