Ask traders how prediction market arbitrage works and you will hear about buying Yes on one exchange and No on another. Almost nobody talks about the quieter trade hiding in plain sight: selling your winning contract at 99 cents instead of holding it for the final dollar. It looks like leaving money on the table. Run the numbers and it is closer to buying insurance at a price the market mislabels as a loss.
Every contract on Kalshi and Polymarket settles at exactly one dollar or zero. When your side trades at 99 cents, the market is telling you the outcome is about 99 percent certain. Holding means risking the 99 cents you could bank right now to chase one more cent. Selling means the last cent, and every risk attached to it, belongs to someone else.
The math of the last cent
The clean way to see it: at a price of 99 cents, the market itself estimates a 1 percent chance that something still goes wrong. If that estimate is even roughly right, holding to settlement has an expected value of 99 cents, exactly what selling pays you today, minus nothing. The two choices are equal on expected value. They are not equal on risk, timing, or fees.
| Choice | You receive | When | What can still hurt you |
|---|---|---|---|
| Sell at 99 cents | 990 dollars, minus a taker fee of about 70 cents on Kalshi | Now | Nothing. The trade is finished. |
| Hold and it settles Yes | 1,000 dollars | At settlement, hours to weeks later | Nothing, if it settles as expected |
| Hold and something goes wrong | 0 dollars | At settlement | A reversal, a disputed ruling, a rules technicality |
That third row is the whole argument. A 1 percent chance of losing 990 dollars to gain 10 is a bad trade at almost any bankroll, and it gets worse the larger the position. Selling at 99 converts a probability into a certainty, and certainty is what the penny buys.
Settlement risk is not theoretical
Contracts resolve against a named source and a written set of rules, and both can bite. An election market waits for official calls. A sports market can hinge on a stat correction or a game finished under protest. Polymarket resolutions run through an oracle process with a dispute window that occasionally produces outcomes traders did not expect, and Kalshi sometimes pauses or settles markets early under its rulebook. None of this is common, which is exactly the point: it is roughly a 1 percent class of event, and the 99 cent price is already charging you for it.
Your capital is stuck while you wait
The second cost is invisible: time. A contract that resolves in three weeks parks your money for three weeks. Sold at 99, the same dollars can work in the next market today, and traders who roll capital through many markets a month give up real returns holding every position to the final cent. The same logic runs through earnings season, when a contract tied to one report can park money past the next name on the earnings calendar. There is also a fee detail in selling's favor: Kalshi's fee formula shrinks toward the extremes, so exiting at 99 cents costs about 7 cents per 100 contracts, the cheapest exit on the schedule. If you want to lock in profit, Kalshi practically discounts the door.
When holding to 100 still makes sense
Sometimes the penny is worth chasing. If the book at 99 is thin, your sell order may only fill in part, or push the price down as it fills. If settlement is minutes away and the outcome is already final in everything but paperwork, the residual risk is close to zero. And small positions change the arithmetic emotionally if not mathematically: on ten contracts the argument is about ten cents. The rule most disciplined traders land on is boring: the bigger the position and the longer the wait, the more the certain 99 beats the hopeful 100. It sits next to its cross-venue cousin, classic prediction market arbitrage, as one of the few genuinely structural edges retail traders can use. For the data behind prices like these, see the APIs and research feeds both exchanges publish, and for the markets themselves, how sports event contracts compare with sportsbooks.