University · Lesson 11 · Prediction Market Mechanics
You Won 90% of Your Bets. Why Are You Still Losing?
A 90% win rate can lose money. Calculate prediction market expected value, break-even prices, and how entry costs change the economics of a likely winner.
Lychee University · Lesson 11
You buy 100 contracts across different markets, one contract at a time. Each costs 92¢. Ninety win.
A 90% win rate sounds impressive. Your bill was $92. Your winning contracts pay $90.
What win rate sees
90%
90 winners · 10 losers
What the cash says
Net before fees
−$2.00
You're down $2 before fees. Winning often and earning a profit are different measurements. To understand the difference before an outcome is known, calculate expected value.
This example assumes standard binary contracts paying $1 for a winning outcome and $0 for a losing one, all held to settlement. Polymarket documents binary resolution and redemption.
What does win rate leave out?
Win rate counts winners relative to the observations being counted. It leaves out the dollars paid and received.
In our hypothetical record:
| Result | Contracts | Profit or loss per contract | Total |
|---|---|---|---|
| Win | 90 | +8¢ | +$7.20 |
| Lose | 10 | −92¢ | −$9.20 |
| Net | 100 | −$2.00 |
At 92¢, a 90% win rate finishes behind before fees. Win rate never saw those dollar weights.
The winners earned a little each. Each loser cost much more. Different position sizes, early exits, and fees would change the calculation further.
Also check what a tracker calls a “win”: a fill, a closed position, a resolved market, or a whole event. Those counts need not describe the same record.
This “win” means
A resolved market scores the binary outcome. Size and entry still decide whether dollars follow.
What is expected value in a prediction market?
Expected value weights each possible profit or loss by its probability. It describes a probability-weighted average, not the result promised on your next trade.
Suppose you estimate a YES contract has a 90% chance of paying $1, and you can buy it at 92¢.
Your expected profit per contract, before fees, is:
That's −2¢ per contract under your probability estimate, even though YES is very likely to win. The price asks you to pay more than the $0.90 expected payout.
Expected payout
0.90
You pay
0.92
EV / contract
−0.02
EV = 0.90(1 − 0.92) + 0.10(0 − 0.92) = −0.02
How do you calculate the break-even price?
For a binary share paying $1 or $0, held to settlement:
Here, p is your estimated probability, c is the actual purchase price in dollars, and f is a known per-contract entry fee in dollars. This simple version assumes no other costs or credits.
With a 90% probability and no fees, the break-even purchase price is 90¢. Paying 88¢ gives +2¢ expected profit; paying 92¢ gives −2¢. Any applicable entry fee lowers what you can pay while retaining positive expected value.
Break-even price
90¢
You pay
88¢
Expected profit
+2¢
Paying below 90¢ leaves positive expected profit under this p and f — if your estimate and fills hold.
That doesn't make an 88¢ purchase automatically good. Your 90% estimate might be wrong, and the quoted 88¢ might not be available for your full order.
Does positive expected value guarantee a profit?
No. A positive-EV trade can lose, and a negative-EV trade can win.
+EV setup
Buy at 88¢ with p = 90%
Expected +2¢ before fees
This trade
Contract loses
You still lose the 88¢
Takeaway
EV ≠ guaranteed profit
Limited samples can run cold
Shared risks and large sizes change how damaging a cold streak becomes. The math is only as good as p, fills, and costs.
Even repeated favorable opportunities can produce losses over a limited record. Shared risks and large position sizes also affect how damaging those losses become. A calculation is only as useful as its probabilities, fill assumptions, and cost accounting.
For the broader task of judging a wallet's track record, see what makes a good Polymarket or Kalshi trader. That lesson examines risk, entry prices, and what happens after you start watching.
What should you record alongside win rate?
Record the actual cost, quantity, fees, and final sale proceeds or settlement payout. Separate realized results from open positions valued at a stated price and time.
- 1Actual cost and quantity
- 2Fees
- 3Final sale or settlement payout
- 4Open positions (price + time)
- 5Pre-trade EV estimate (optional)
A likely winner can still be expensive. How often you win matters only alongside what those wins pay — and what the losses cost.
Then compare total money earned with total money lost. If you estimate EV before entry, save that estimate too, so later research doesn't quietly replace your original belief with hindsight.
A likely winner can still be expensive. How often you win matters only alongside what those wins pay—and what the losses cost.
Get started with Lychee.
Free to explore here · Jarvis included in every paid plan at launch
More reading
A 90% Prediction Lost. Was It Wrong?
misterrpink · 3 min read
· Research
When Do Prediction Markets Become Accurate? A Kalshi Political Market Lifecycle Analysis
misterrpink · 18 min read
· Research
Are 90% Prediction Markets Reliable? A Kalshi Political Calibration Study
misterrpink · 15 min read
· Research
Political Prediction Markets: What Historical Data Reveals About Election Forecasting, Polls, and Market Accuracy
misterrpink · 12 min read
Kalshi Shows 40¢. Why Did Your Order Average 42.8¢?
misterrpink · 3 min read
Polymarket Shows 65¢. Why Does Buying Cost 67¢?
misterrpink · 3 min read