Scalping prediction markets — when speed beats edge
How short-timeframe scalping works on Polymarket and Kalshi, when it makes money, and where it falls apart.
Definition
Scalping in prediction markets means competing for fills on short-duration binary markets where the settlement source is itself an aggregator (e.g., Chainlink RTDS). The edge is informational latency: the bot reads the source exchanges directly while the oracle aggregates them on a slower tick.
When it works
Liquid 5-minute and 15-minute crypto binaries. Tight spreads. Active book on both sides. Calm-to-moderate volatility.
When it fails
Choppy markets where price oscillates inside the no-trade zone. Adverse selection from latency-arb desks. Fee drag at extreme strike prices.
Fee math
5-minute crypto binary taker fee:
fee = shares * 0.25 * price * (price * (1 - price))^2
The fee peaks at p=0.50 (~1.56% notional) and decays toward the wings. Strategies that enter at p=0.45-0.55 pay the most fee per dollar; wing strategies pay less but need higher hit rates.