It’s the wrong question. Not because stamina doesn’t exist, but because “will it stay?” forces a binary frame onto something that isn’t binary. Stays or doesn’t stay. Gets the trip or doesn’t.
The reality is a curve, not a cliff.
Every horse has a range of distances it can physically complete. Within that range is a narrower band where it performs at its peak — where biomechanics are most efficient and stride mechanics best match the energy demands of the race.
That narrower band is what the model predicts — not whether a horse will stay, but where it will be most effective. That is what horse racing distance prediction should be.
How StridePredictor Works
The StridePredictor model approaches horse racing distance prediction differently — it predicts peak efficiency. The distance at which a horse’s stride mechanics are best matched to the demands of the race. It anchors on a horse’s best performance. Not the most recent run. Not an average. The race where conditions allowed the horse to perform closest to its true capacity.
Every prediction carries a confidence rating. A high confidence reading from a genuinely run race carries more weight than a moderate reading from a slowly run race. When a horse produces a compromised reading — a slowly run race, conditions that distort stride mechanics — that reading is noted and weighted accordingly. It does not override the body of clean data that preceded it.
The Double Standard
Here is what nobody challenges in racing. A horse runs two races, three weeks apart, and finishes 20 lengths closer to the winner on the second occasion. Form analysts adjust. They contextualise. They talk about unsuitable ground, a bad draw, a pipe opener, traffic problems. They accept, without question, that the first performance was not representative of the horse’s true ability.
That is standard practice. Accepted. Nobody questions it.
Now apply the same logic to stride data.
A horse records a below-par stride reading in a compromised race. The reading differs from its best performance readings. And immediately — the stride data has changed. The horse is a different horse now. New conclusions drawn.
One approach is given infinite latitude to account for variation. The other is held to a standard of perfect consistency that form analysis would never meet.
A Live Example: Action
Action’s best two-year-old performances returned predicted distances of 8.2f at Newmarket in the Royal Lodge and 8.4f in the Futurity at Doncaster. Consistent. High confidence reads from genuinely run races.
For his 3-year-old season debut, he ran in the 10-furlong Group 3 Classic trial at Sandown. He went off at 4/6. He finished fourth, well beaten. The stride reading from the race returned 10.0f.
His optimal distance prediction remains around 8.3f — anchored on his two best performances. A 10-furlong trip asked him to race beyond the distance where his stride mechanics are most efficient. He wasn’t at his best because he wasn’t at his best distance.
The market backed him as though 10 furlongs was well within his range. The stride data from his best performances pointed clearly at a miler’s profile. He was well beaten in a 10-furlong trial.
Not a failure of the model. A confirmation of it.
The Curve, Not The Cliff
A 400 metre runner can complete a mile. That doesn’t make them a miler. They will finish. They will not be at their most effective. The distinction matters — to the athlete, to the coach, and to anyone trying to predict what happens next.
Racing asks “will it stay?” because that is the question form provides an answer to. A horse either got the trip or it didn’t. The result is the evidence.
Biomechanics asks a different question. Not whether the horse got the trip. Whether the trip got the best out of the horse.
What Predicted Distance Means
Predicted distance is not a stamina limit. It is the point where a horse is most likely to perform at its peak. Not the furthest trip it can complete — the trip most likely to get the best out of it.
That is where the edge lies.