Guides · 8 June 2026 · 7 min read
How to build and backtest a horse racing scoring system
Start with a rule you can write down
A scoring system is just a rule you commit to before you see the result. 'Back every pick scoring 450 or more at odds between 2.0 and 8.0, one point level stakes' is a system. 'Back the ones that look right' is not. If the rule cannot be written in one sentence, it cannot be tested, and if it cannot be tested you will never know whether it works or you have just been lucky.
Good raw materials for rules are measurable and available before the off: a ratings threshold, a price band, a race type, a going preference, a field-size limit. Bad raw materials are anything you only know afterwards, which is where most homemade systems quietly cheat.
Backtest against everything, not the good bits
Backtesting means running your rule over past races as if you had bet it, and the golden rule is: every qualifying race counts. The classic mistake is trimming the sample, starting the test after a bad run, or excusing losers as 'unlucky'. A real backtest over a season of UK and Irish racing means thousands of races, and the result you care about is not the best month but the whole curve.
Read the numbers together, never one at a time. Win rate tells you how often the rule lands. Place rate tells you how violent the losing runs will feel. Neither means much without the prices those winners went off at, and that is the number most homemade systems never record. A 40% win rate at short prices and a 20% win rate at big ones are describing completely different rules, which is why a win rate quoted on its own fools so many people.
Tools that do the heavy lifting
Doing this in a spreadsheet is possible and painful: you need a season of results, starting prices, and patience. A purpose-built backtesting engine collapses the work to minutes. The Racing Bot's simulator holds a season of settled results and re-runs it live as you change the rules: reweight any of the 23 scoring signals - the 15 scored on every runner, plus the eight more Pro adds - filter by race type, going, field size or price band, choose singles through six-folds, and watch win rate and place rate update as you drag.
Two honest warnings before you trust any backtest, including ours. Past results are a record of what happened, not a forecast of what will, and the more rules you stack, the more likely you have fitted the past rather than found an edge. Test one change at a time, demand a decent sample, and start small with real money. 18+, please gamble responsibly.