In-Play Odds History: Where to Get It and How to Use It

In-play history records prices for markets such as match result, totals, and handicaps throughout a match. Each record includes a timestamp, market, selection, price, and often volume or liquidity for exchange data. Because prices update constantly, a single match can produce thousands of records per market.

There are three main ways to get in-play history, each with different costs and coverage.

Odds data providers

Betting exchanges

Your own collection

In-play prices only make sense alongside what happened in the match: goals, cards, injuries, substitutions, and the clock. Join odds history with event data using accurate timestamps. Watch for time zone differences and feed latency, since a price change may appear seconds after the event that caused it.

Books suspend markets when significant events occur, such as a goal or a penalty decision, and reopen with new prices. Exchange data may show thin liquidity during these moments. Treat suspensions as meaningful events rather than missing data, and mark them clearly in your dataset.

In-play data grows quickly. Store it in a format suited to time series, partition by sport and date, and compress older data. Keep raw snapshots alongside cleaned, normalised tables so you can reprocess when your methods improve.

In-play history lets analysts study how quickly markets react to events, how prices move after goals or red cards, whether certain markets overreact or underreact, and how closing in-play prices relate to outcomes. It also provides training data for live win-probability models.

Backtesting in-play strategies requires realistic assumptions: bet acceptance delays, the price actually available at the time, liquidity, and suspension rules. Historical last-traded prices can overstate what was achievable, so include slippage and delays in any simulation.

Different books and exchanges label markets, teams, and times differently. Normalise identifiers and convert prices to a single format before comparing. Parlay Ledger's focus on normalising multi-sport odds across sources applies directly to in-play history, where inconsistencies multiply.

Respect the terms of any provider or exchange whose data you use, particularly regarding redistribution. Avoid scraping sites that prohibit it.

Pick one sport and one market, collect or buy a season of in-play history, align it with event data, and start with a simple question, such as how much the match odds move after a goal at different minutes.

Typical questions include how much match odds move after a goal at different minutes, whether markets overreact to red cards, how total goals markets decay as time passes without goals, and how quickly different books update after events. Each requires precise event times and prices captured at sufficient frequency.

Decide how often to capture prices. Every few seconds captures fast reactions but produces large datasets; once a minute is smaller but misses short-lived prices. Many analysts capture at high frequency around events and lower frequency otherwise.

In-play history provides training data for models that estimate win probability from the score, time remaining, and team strength. Comparing model output with in-play prices shows where markets agree, disagree, and how quickly they converge.

Check for duplicated timestamps, prices outside plausible ranges, missing periods, and markets that never reopened after suspension. Compare a sample of matches against a second source. Record every cleaning rule so analyses can be reproduced.

Linking in-play history to opening and closing pre-match prices shows how pre-match expectations evolve once play begins. It also lets you test whether pre-match mispricing persists into the early minutes of matches.

In-play history is large, detailed, and demanding to clean. Align it with events, respect suspensions, and backtest with realistic execution assumptions.

Trading teams, sportsbook risk departments, academic researchers studying market efficiency, and modellers building live win-probability systems all rely on in-play history.

This page describes how in-play price history is collected and used, and is not betting advice.