Two trusted ability numbers · One scale · Built for the Saturday handicap
Two numbers. One mark. Backed by maths.
The Short Answer
SR is the average of two trusted ability numbers — the Racing Post's recent-form RPR and the BHA handicapper's settled OR — then nudged by market, form and weight.
— It's what you'd get if a sharp form student and a handicapping jury had to agree on one mark.
Why we don't trust either one alone
RPR and OR are both excellent, and both biased. RPR rewards recent form and overreacts to a single peak run; OR is settled by committee and lags behind the truth. Either alone is noisy. Blending them nets out both biases — and matches how UK handicapping juries actually triangulate the two when assigning weights for Saturday's handicaps.
Exhibit A — RPR alone
RPR rewards what the horse did last time. One peak run on heavy ground at Cheltenham, one sticky opponent who fell at the second-last, and the number jumps. For weeks after, the horse looks elite on paper.
"5/1 winner on heavy at Cheltenham. RPR jumps 14 lbs. Was the trip a one-off? RPR can't tell you."Exhibit B — OR alone
OR is the BHA handicapper's settled mark. A jury panel reviews each run and raises the figure after the fact. A young improver winning three in a row is already 6 lbs better than his OR — but the number hasn't caught up.
"Three wins in a month. OR raised 6 lbs. The horse has actually improved 12. OR alone says he's still beatable."The Blend
RPR
170
Recent peak form
OR
140
Handicapper's mark
SR Anchor
150
The honest middle
A 50/50 mean — deliberately even until backtest data tells us one side out-predicts the other on Saturday handicaps. The ability anchor then becomes the foundation of the final SR, with market, form and weight layered on top.
A worked example
Fictional horse, real maths. Every line below comes from the
live formula in ai/rating.py.
Won a 0-55 handicap at Ripon last month by 7 lengths — RPR jumped to 84. The handicapper saw that race and raised OR by 4 lbs, to 71.
The ability anchor
Then the layers
Top of the field — backed up by the market, the form line and the weight. Not just one big run.
For the curious
The anchor. Blends two ability signals into one, with graceful fall-backs whenever one is missing. Carries the weight of the final number — everything else nudges this.
Shorter price, positive nudge; long-shot price, negative. The market is information — but it never overwhelms the ability anchor.
A steamer (10/1 → 4/1) and a drifter (3/1 → 4/1) settle at the same spot price but tell opposite stories. This component reads the full price lifecycle per bookmaker and rewards the steamer, penalises the drifter — even when their final price is identical.
Real form analysis — not digit counting. Each of the last four runs scores from +5 (clear winner) down to −3 (faller), then multiplies by a class weight. A 2nd beaten ½L scores like a winner; a 2nd beaten 20L barely registers. A Group 1 win outscores a Class 6 seller by ×1.9.
The handicapper's leveller. The lighter you are, the harder you are to stop — and the higher your SR climbs.
Reading the number
Most Saturday handicap runners sit between 60 and 95 on the BHA-anchored side; the bands below cover the elite end where Group and Pattern races live.
Why it wins
A horse priced 2/1 on the back of one career-best RPR often blends down to a more honest SR. When SR disagrees with the market, that's an each-way play against the favourite — or a fade.
When OR is stale and RPR is climbing, SR sits just below RPR but still flags the horse as ahead of the handicapper. Classic "blot on the handicap" territory — the market often hasn't caught up yet.
Pure RPR varies wildly between codes. SR's 0–200 scale is the same on a Class 6 Catterick handicap as on a Group 1 — one number, every Saturday, same meaning.
ai/rating.py. Want the
picks? They're already on every racecard, ready to read.