How arbitrage bots exploit stale odds between books

Why odds go stale

Odds are a consensus that updates as information arrives: lineups, weather, sharp money, injuries. Every book updates on its own schedule, and the schedules disagree. A book with a manual trader is slower than a book with an automated feed. A book that suspends in-play markets during every dangerous attack is slower than one that keeps them open. Each gap is a window where two books disagree about the same event, and disagreement is arbitrage.

Feed latency is the structural cause. Even fully automated books ingest the same data at different speeds, and the slowest book in the comparison is always the one the bots target. There is no version of this business where every book updates at the same instant.

How the bots work the window

An arbitrage bot holds accounts at many books and watches the same markets on all of them. When the combined odds across two books imply a guaranteed profit, it stakes both sides within seconds, sizing the bets so the payout is identical whichever outcome lands. The profit per arb is small, often one to three percent, so the operation runs on volume: thousands of small arbs across every sport and market.

The sophisticated operations hedge across books in real time and recycle the bankroll continuously. The crude ones just hammer the slow book's stale price until the limit or the account dies. Both cost the slow book the same way: paying out risk-free profit to customers who never take a position.

Telling arbs apart from sharp action

Arbitrage betting has a fingerprint. Stakes sized to round-number payouts rather than round-number stakes, bets placed within seconds of a line move elsewhere, action concentrated on markets where your price disagrees with the consensus, and accounts that never bet your opinion, only your mistakes. A genuine sharp bettor has views; an arb account has a scanner.

The distinction matters because the response differs. Sharp action is information you want to price in. Arbitrage is latency you want to eliminate. Treating arbs as sharps just teaches your traders to respect prices that were never real.

Closing the window

The durable fixes are operational, not punitive. Synchronize your pricing feed so the lag behind the fastest book shrinks to seconds. Add a short bet delay on in-play markets so stale prices cannot be hit during the moments they are most wrong. And set stake limits that bind on arb-pattern accounts while leaving genuine customers alone: the goal is to make the window too small and too illiquid to be worth scanning for.

Account limits remain the backstop, but they work best as the last step after the operational fixes. A book that only limits accounts is admitting its prices cannot be trusted; a book that fixes the latency keeps the customers and loses the bots.