How it works · 12 September 2026 · 4 min read

What goes into a score

A rating is built from the same information printed in a racecard, just weighed differently. Recent form, the trainer's record, the jockey booked, how a horse handles the course, the distance, the ground, the size of the field it's stepping into. None of this is secret. What changes is how consistently it gets applied, run after run, horse after horse, without a name in the paper or a hunch from last season nudging the answer.

Fifteen signals sounds like a lot until you watch a race and realise how many things actually decide it. A horse can be well-related on pedigree, fit on form, dropped in class, ridden by a jockey in form, and still get beat by traffic in running. A model can't see traffic. It can see the pattern of everything measurable, applied the same way every time, which is where the value sits.

Form on the racecard

Recent form is the obvious starting point but it's easy to misread. A string of fourth places can look unremarkable next to a single win buried three runs back, yet the horse finishing fourth in stronger races might be working harder for less reward. Form has to be read against the grade it was earned in, not just the finishing position on the page.

Class moves matter just as much. A horse dropping from handicap company into a weaker race, or stepping up after an easy win, changes what its recent figures actually mean. Days since its last run counts too. A horse fresh off a break can turn up sharp or rusty, and trainers vary in how they use time off, some with a strong record of first-time-out improvement, others needing the run to strip fitness. None of these signals mean much alone. Stacked together, they start to sketch a shape.

The people running the horse

Trainer and jockey strike rates get quoted constantly and for good reason. A yard in form is sending out fit, well-primed horses, and that shows up in the numbers over weeks, not just on the day. A jockey booking can matter too, especially when a stable's number one rider takes over from whoever sat on the horse last time.

These figures move slowly and reward patience rather than a single card's data. A trainer having a purple patch through the summer is a different signal to one buried in a winter slump, and it's the kind of pattern that's easy to miss reading a card by eye but straightforward to track when it's simply being counted.

Course, distance and the conditions on the day

Course and distance form tells you whether a horse has actually done it before, on this track, over this trip, under these conditions, rather than just looking the part on paper. Some courses are quirky enough, sharp turns, undulations, a stiff finish, that horses either handle them or don't, and it shows up in their record there specifically.

Going preference works the same way. A horse that's bred for soft ground and has only ever won on it is a different proposition on quick summer turf, whatever its official rating says. Field size adds another layer again, since a horse that thrives in small fields with a clear passage can find life much harder buried in a maximum-runner handicap with nowhere to go.

Why the sum beats the eye test

Reading a racecard by eye is quick and it works, up to a point. The trouble is consistency. A human reader will weigh the same signal differently race to race depending on mood, tiredness, or what caught the eye last week. A model applies the same weighting every time, across every runner, which is the entire argument for using one alongside the racecard rather than instead of it.

The Racing Bot's public record currently stands at 15,750 settled picks across 436 days, with 25% of those selections winning and 50% placing. That's the output of applying the same fifteen signals consistently rather than picking favourites on gut feel, and it's published openly rather than quoted selectively.

The variance that never goes away

None of this changes what horse racing is. Fields of ten or more runners, jumps and interference, ground that shifts through the afternoon, a horse that simply doesn't fire on the day for no obvious reason. Favourites get beaten constantly, and outsiders win races they had no business winning by the figures. That's the sport, not a flaw in how it's measured.

A rating is a considered starting point, not a forecast. It tells you which horses the numbers rate highly and why, and leaves the racing itself, the bit nobody can model, to happen on the track.