Learn / Strategy
Seven prediction market strategies, and how each one fails.
Short answer
Every prediction market strategy is a claim that the price is wrong for a specific reason. Write down the reason, the costs and the way it fails, then test it on data the strategy never saw. Strategies without a stated reason usually lose to fees.
Every trade in a prediction market says “this price is wrong.” A strategy is a repeatable reason why it is wrong. Here are the seven reasons traders use most often, what each one needs to work, and how each one usually fails.
| Strategy | The edge it claims | Main way it fails |
|---|---|---|
| Model vs market | Your probability is better than the price | Model is not better after fees |
| Cross-platform arbitrage | Same event, cheaper combined price | Rules differ; fees eat the gap |
| Market making | Earn the spread by quoting both sides | Informed traders pick you off |
| News speed | React before the price does | Others are faster; headlines mislead |
| Rules reading | The market misreads the settlement terms | Your reading is the wrong one |
| Fading longshots | Cheap contracts are overpriced | Rare big losses; bias absent in that market |
| Near-certain contracts | Collect the last few cents | One surprise wipes out dozens of wins |
1. Model versus market
Build an independent probability, compare it with the price, and trade when they disagree by more than costs. Sports are the natural home for this: team ratings, injuries, rest and venue can be modeled. Kelix Lines does the comparison step, putting a simulation-based probability beside available market prices.
How it fails: a disagreement is not an edge. Markets are often well informed, so your model must be measurably better calibrated than the price on events it never trained on. A 2-point gap at 50 cents is mostly fees.
2. Cross-platform arbitrage
Buy Yes on one venue and No on another when the pair costs less than $1. How it fails: the two contracts settle on different rules, fees on two legs erase small gaps, or only one leg fills. Full walk-through in Kalshi–Polymarket arbitrage, honestly.
3. Market making
Post a bid and an ask and earn the difference when both fill. On venues where makers pay no fees or receive rebates, this can be attractive. How it fails: adverse selection. Your orders fill most often right before the price moves against you, because the traders hitting them know something. Market makers survive by widening quotes or pulling them around news and by capping inventory.
4. News speed
Trade the gap between when information appears and when the price absorbs it. How it fails: you are competing with automated traders and with people closer to the source. Headlines are also often ambiguous relative to the contract’s exact wording. Measure your actual reaction time against the price’s before assuming you are early.
5. Reading the rules
Prices sometimes reflect the headline question while settlement depends on fine print: a specific data source, a deadline time zone, or how revisions are treated. How it fails: the market may already understand the rules, and you are the one misreading them. When in doubt, the platform’s rule text and its resolution history are the only authority.
6. Fading longshots
The favorite-longshot bias is one of the oldest findings in betting research: low-probability outcomes tend to be priced above how often they win (see Thaler and Ziemba, “Parimutuel Betting Markets,” Journal of Economic Perspectives, 1988). Selling cheap Yes contracts, or buying the expensive No, tries to harvest it. How it fails: the payoff is many small wins and occasional large losses, and the bias must be measured in the specific market you trade, after fees, rather than assumed.
7. Buying near-certain contracts
Buying Yes at 97 cents on something “definitely happening” feels safe. You risk 97 cents to make 3, so one surprise erases about 32 wins. After a Kalshi fee of about 0.2 cents per contract at that price, you need the event to happen more than roughly 97.2% of the time. That can be true, but it is a precise claim that needs evidence, not a feeling of safety.
How to test any of these
- Write the hypothesis. “Prices on X underreact to Y,” not “buy when it looks cheap.”
- Use point-in-time data. Only prices and information that existed at each decision.
- Pay realistic costs. Executable asks and bids, current fees and size limits.
- Hold out a later period and evaluate on it once.
- Count every trade the rules would have made, including the losers.
Our prediction market backtesting guide covers each step. Kelix Lab runs strategies against historical Kalshi markets with quoted fills, taker fees and a separately reported held-out sample. To automate a tested strategy, see building a prediction market bot.
Frequently asked questions
What is the best prediction market strategy?
There is no universal best. The useful question is which edge you can actually measure: a better model, faster information, cheaper execution or a closer reading of settlement rules.
Is buying 97-cent contracts a safe strategy?
No. You risk 97 cents to make 3, so one surprise erases about 32 wins. It only pays if the true probability is above 97% plus fees.
What is the favorite-longshot bias?
A long-documented pattern in betting markets where low-probability outcomes tend to be overpriced relative to how often they happen. Whether it holds in a given prediction market must be tested on that market’s own history.
How should I test a strategy?
Use point-in-time prices, include fees and spread, hold out a later time period, and record every trade the rules would have made, not only the winners.
Educational content, not financial, legal or tax advice. Prediction markets involve risk, including loss of capital. Examples are illustrative calculations, not live quotes or trading results. Kelix is independent of Kalshi, Polymarket, Robinhood and sportsbooks.