If a back-test reads a horse's rating as it stands today, it already knows which horses went on to improve. The results look better than anything you could have bet at the time. The video shows the problem on one real race, and how the System Builder avoids it.
One race, two sets of ratings
Newmarket, 17 April 2024: a nine-runner Class 3 handicap. On the racecard for a race that has been run, the H column shows each horse's RaceMetrics Rating as it stood before the off, not as it stands now. That morning, Kalpana was only the fourth-best rated horse in the race. She won.
Rank the same race on today's ratings and the picture changes. Kalpana has improved so much since that she goes straight to the top, and Bust A Move, second-best rated that morning, has gone backwards and drops to the bottom. A back-test built on today's figures pulls in the horses that went on to improve and pushes out the ones that went the other way.
What the System Builder reads
Every RaceMetrics rating rule in the System Builder reads the figure as it stood on the morning of each race. We started keeping those morning ratings at the start of 2023, so when a system uses one and starts earlier, the Builder moves its start date to 1 January 2023 and shows a note saying so. That is not an arbitrary limit: before 2023 there is no morning figure to read, and we do not fill the gap with one worked out afterwards. Systems that do not use these ratings still go back to 2018.
Green and amber
Every field in the rule list carries a coloured dot. Green means the figure is known before the off, so the rule can find today's qualifiers. Amber means the figure only exists after the race, such as the Betfair starting price. You can back-test with an amber field, but it never feeds your qualifiers, and a saved system that uses one is labelled a research back-test.
The System Builder is included with Proform Premium and RaceMetrics Pro. See the plans.
