Most people lose money on prediction markets for a reason that has nothing to do with picking winners. They buy contracts that are fairly priced, pay a fee for the privilege, and repeat until the fees have eaten the account.
Expected value is the test that catches this before it costs you. On a prediction market the maths is unusually simple, and once you have seen it you cannot unsee how many trades fail it.
The formula, and why it collapses
A contract settles at $1 if the event happens and $0 if it does not. Buy at price c and two things can happen: you are right and make (1 − c), or you are wrong and lose c.
If your probability is p, expected value per contract is:
EV = p × (1 − c) − (1 − p) × c
Multiply that out and almost everything cancels:
EV = p − c
That is the whole thing. The expected value of a prediction market contract is the gap between your probability and the price. Nothing else enters it.
A contract at 52¢ that you believe has a 60% chance carries 8¢ of edge per contract. On 100 contracts that is $8 of expected value on $52 staked — about 15% expected return. Our expected value calculator runs this for any combination.
The number that should worry you
The formula is easy. Getting p right is not.
The entire calculation rests on your probability being better than the market's. Not different — better. And prediction markets are well calibrated: Polymarket's markets have posted Brier scores around 0.09, which corresponds to roughly 94% accuracy.
That is a hard bar. When your calculator shows positive EV, it is not telling you that you found a mispriced market. It is telling you that if your probability is more accurate than a market that is right 94% of the time, the trade is profitable. Those are very different claims, and confusing them is the most expensive mistake in this category.
The honest use of an EV calculation is as a filter, not a green light. It tells you when a trade is definitely not worth making — when your own number agrees with the market, or is worse. That is most trades.
Break-even is just your probability
One consequence of EV = p − c is elegant: your break-even price is exactly your probability.
Think you have a 60% read? Then 60¢ is where the trade stops being worth making. At 59¢ you have a penny of edge; at 61¢ you are paying a penny for the privilege. There is no cushion, no zone of "close enough", and no room for the fee.
Which brings us to the thing the formula leaves out.
Fees break small edges
EV = p − c ignores costs, and costs are not small relative to typical edges.
Kalshi's fee is 0.07 × contracts × price × (1 − price), which peaks on a coin-flip contract. On 100 contracts at 50¢ that is about $1.75 against $50 staked — 3.5%. An edge of two cents per contract is $2 of expected value, so the fee consumes most of it. An edge of one cent is a losing trade after fees, despite the calculator showing positive EV.
Polymarket takers pay 0.75–1.80% of the position, which is smaller on a coin-flip but applies at every price. Makers pay nothing, which is a real argument for posting rather than taking if you have the patience.
The practical rule: a one-cent edge is noise, a two-cent edge is marginal, and anything worth acting on usually needs three or more. Run the exact numbers through our fee calculator before deciding.
Positive EV does not mean you win
Expected value is an average over many repetitions. It says nothing about any single trade.
A positive-EV bet on a 60% contract still loses 40% of the time. Four losses in a row happens roughly 2.5% of the time — which, if you place a hundred trades, you should expect to experience more than once. The edge only shows up across enough repetitions for variance to wash out, and "enough" is more than most people have the bankroll or patience for.
This is why sizing matters more than selection for most traders. A genuine 5% edge staked at 40% of your account will still bankrupt you on a bad run. The edge is real and the outcome is still ruin.
Where edges actually come from
If markets are 94% accurate, where is there room?
Speed. News moves one venue before the other. The market is efficient eventually, not instantly. This is the same gap our arbitrage scanner exploits across Polymarket and Kalshi.
Thin markets. A market nobody is watching can sit mispriced for a long time, because the people who would correct it are not looking. The catch is that thin markets are thin in both directions: you may not be able to exit.
Genuine domain knowledge. If you actually know something specific about a niche — a league, a regulatory process, an industry — you can occasionally price it better than a market of generalists.
Structural mispricing. Long shots are systematically overpriced across nearly every prediction market, the same favourite-longshot bias that shows up in horse racing. Selling long shots rather than buying them is a real strategy, though it caps your upside and tests your nerve.
What is not an edge: a strong feeling about a game, a headline you read, or a narrative that seems obvious. The market has read the headline too.
Putting it together
Before a trade, write down your probability before you look at the price. Anchoring is real and looking first will drag your estimate toward the market's.
Then compare. If your number and the price are within two cents, pass — the fee eats it. If the gap is larger, ask honestly why you think you know better than a market that is right 94% of the time. If you have an answer that is not "I just think so", size the position small enough that being wrong five times running does not matter.
That process filters out most trades, which is the point. Our guide to making money on prediction markets covers what to do with the ones that survive it.
Frequently Asked Questions
How do you calculate expected value on a prediction market?
Subtract the contract price from your estimated probability. A 52¢ contract you rate at 60% has 8¢ of expected value per contract, or about 15% on the money staked. The formula collapses to p − c because contracts settle at exactly $1.
What is a good expected value?
Anything genuinely positive after fees is worth taking, but "after fees" does most of the work. A one-cent edge does not survive Kalshi's fee on a coin-flip contract. Three cents or more is where a trade starts being clearly worth making.
Does positive EV guarantee a profit?
No. Expected value is an average across many trades. A positive-EV bet at 60% loses four times in ten, and runs of losses are normal rather than evidence you were wrong.
Why is my edge always negative?
Usually because the market is right and you are anchoring on its price. Write your probability down before looking at the quote — if your estimate keeps landing within a cent or two of the market, that is the market being efficient, not you being unlucky.
Do prediction markets beat sportsbooks on expected value?
Generally yes, because the margin is smaller — under 1% on a liquid prediction market against 4–5% on a sportsbook line. The comparison is in prediction markets vs sports betting, and you can measure any market yourself with the vig calculator.
How accurate are prediction markets?
Polymarket has posted Brier scores around 0.09, roughly 94% accuracy. That is the bar your own estimate has to beat for an edge to be real.
