Prediction market arbitrage: how it works and what can go wrong
Overview
Prediction market arbitrage seeks a combined position whose payout exceeds its total cost across equivalent contracts. One approach buys Yes on one venue and No on another. It only works on paper when both payout rules align—and real trades still face fees, failed fills, settlement differences and capital lockup.
The calculation: buy both outcomes for less than the payout
For one equivalent binary pair with a $1 total payout, the conditional net surplus is $1 minus the Yes buy cost, the No buy cost and all applicable costs. Use executable buy quotes for the same quantity. Two displayed Yes probabilities do not tell you the cost of buying Yes and No.
| Component | Assumption | Amount |
|---|---|---|
| Yes position on venue A | 100 shares bought at 42¢ | $42 |
| No position on venue B | 100 shares bought at 53¢ | $53 |
| Total entry cost | $42 + $53 | $95 |
| Conditional payout | Exactly one outcome pays $1 per pair under aligned settlement rules | $100 |
| Gross surplus | $100 − $95 | $5 |
| Assumed total costs | Illustrative allocation of fees, funding and other costs | $2 |
| Conditional net surplus | $100 − $95 − $2 | $3 |
The $3 is a scenario result, not a guaranteed return. If total costs reach $5, the net surplus is zero. If the second leg fills at 58¢ instead of 53¢, the entry cost reaches $100 and the assumed $2 costs produce a $2 loss. Partial fills or conflicting settlement rules can create larger losses.
A price gap is not the same as arbitrage
A Yes price of 42¢ on one platform and 50¢ on another shows an 8¢ price difference, not an 8¢ hedged return. The hedge needs the other platform's actual No buy quote. A 50¢ Yes buy quote does not prove No is available at 50¢; the bid–ask spread matters.
On Kalshi, PP uses the Yes ask for buying Yes and one minus the Yes bid for buying No. Missing supported quotes stay unavailable. Never replace them with a midpoint, the last trade, zero, or an assumed complementary ask.
Before placing either leg
- Confirm both accounts can legally and contractually trade; pre-funding does not remove access restrictions.
- Verify identical outcome conditions, threshold, observation window, time zone and resolution-source treatment.
- Check live buy quotes and cumulative depth for the same number of shares on both sides.
- Calculate contract-specific trading fees plus funding, network, withdrawal and conversion costs.
- Plan what you will do if one leg fills and the other does not; no cross-venue atomic fill is promised.
- Assess settlement delays, voiding scenarios and how long capital may be committed.
PP is a snapshot-based research starting point. LIVE means effective age is at most 5 minutes; DELAYED is above 5 and at most 10 minutes; STALE is above 10 minutes. Comparisons use the snapshot age and older required-leg quote, whichever is older. Even a LIVE snapshot is not a guaranteed current fill.
Settlement risk is the big one
The whole trade depends on both legs paying out on the same event. If the two markets resolve on different sources, different dates, or different definitions, you are not hedged — you are holding two separate directional positions that can both lose.
Read both sets of resolution criteria before assuming a pair is equivalent. Automated matching, including ours, is a starting point, not a guarantee.
Execution and liquidity risk
- Other sources may display a midpoint rather than a buy quote. PP uses supported source buy prices, but depth, fees and price changes can still make your actual fill worse.
- Order books in prediction markets are frequently thin, so the second leg may not fill at all.
- Prices move while you are moving funds or clicking through two interfaces, and the gap can close before both legs are on.
Capital, cost, and timing risk
- Capital may be committed until settlement unless you can close both positions. An early exit can incur spreads and fees; closing only one leg leaves directional exposure.
- Applicable deposit, withdrawal, network and conversion costs can erase the spread. Not every method charges every cost; check your actual funding route.
- Use each venue's current market-specific fee schedule and order preview. Do not substitute an assumed flat percentage or treat illustrative costs as actual venue fees.
- You need funded accounts on both platforms in advance; funding reactively is usually too slow.
Platform and regulatory risk
Platform availability differs by country and can change. Accounts can be restricted, markets can be voided or re-settled, and access rules for a given jurisdiction may shift while your capital is committed. Custody risk also differs: a regulated exchange and an on-chain venue carry different failure modes.
Prediction Pantheon is an information and comparison service. Nothing here is investment advice, and gaps flagged in the feed are labelled beta because market matching is imperfect.
Frequently asked questions
Is prediction market arbitrage risk-free?
No. Even a perfectly matched pair carries execution, fee, capital-lockup, platform, and regulatory risk. Mismatched resolution rules can turn a supposed hedge into two losing positions.
Why do arbitrage opportunities persist if they are real?
Usually because they are not fully real: costs, thin liquidity, regional access limits, or subtly different settlement rules stop them being captured cheaply.
Can I practise before risking money?
Yes. Prediction Pantheon includes a free virtual practice portfolio so you can place simulated trades and follow markets to resolution without depositing funds.