Originally published December 2025. Updated July 2026.
The market gets distance wrong.
Talented horses are regularly campaigned over trips that don’t suit their biomechanics. Form tells us what a horse has already done. Pedigree tells us what horses with similar breeding have tended to do.
StridePredictor asks a different question:
Where is this individual horse most effective?
That distinction matters.
Predicted distance is not a stamina limit. It’s an efficiency peak.
A horse predicted to be most effective at 10 furlongs will stay 12 (eventually). A horse predicted at 12 furlongs can still win over 10 — the model doesn’t measure the ability of the horses it races against, only where its own biomechanics are working most efficiently. The prediction isn’t a cutoff point. It’s the point where the horse’s biomechanics suggest it can perform most efficiently.
The market asks, will it stay?
StridePredictor asks, where is it most effective?
That is the gap horse racing stride analysis is designed to close.
StridePredictor was built to identify horses racing at the wrong distance before form proves it and prices move.
The technology was already in place. Total Performance Data pioneered the stride capture system. Simon Rowlands brought stride analysis into the mainstream. Kevin Blake correctly predicted the first three home in the 2025 Epsom Derby using stride data.
The question was whether stride data could reliably identify where an individual horse would be most effective.
SPS And Distance: The Foundation
Lower stride frequency, more stamina. Higher stride frequency, less.
Stride frequency is measured in strides per second — SPS. The variable that matters most is minimum SPS: the slowest rate at which a horse turns its legs over during a race.
We analysed 1,225 horses across all distance categories, each from a truly run race. The pattern is clear:
| Distance Range | Typical Min SPS | Median |
|---|---|---|
| Sprint (5f→7f) | 2.27–2.37 | 2.31 |
| Miler Plus (7.1f→10f) | 2.16–2.26 | 2.21 |
| Middle Distance (10.1f→13f) | 2.09–2.18 | 2.13 |
| Staying (13f+) | 2.06–2.13 | 2.09 |
For a horse to be most effective at 12 furlongs, you want minimum SPS below 2.18 — ideally 2.13 or lower.
Above 2.20, the stride profile is pointing towards miler-plus territory.
Simple rule. Powerful filter.
But stride frequency alone wasn’t enough.
Why Early Models Failed
The first attempts to model distance from stride data alone produced an R² below 20%.
Stride frequency and stride length explained less than a fifth of the variation in optimal distance.
Something was missing.
The breakthrough came from introducing pace into the analysis.
Race Pace Changes The Reading
After raw ability, race pace has one of the greatest influences on race outcomes.
Efficiently run races — where energy is distributed logically across the trip — produce results that reflect ability rather than tactics.
Stride data behaves the same way.
In an inefficiently run race, stride cadence and stride length can distort. The data starts describing the race rather than the horse.
Restricting the training set to efficiently run races transformed the models. R² moved from below 20% to above 80% when predicting optimal distance. The same inputs — stride frequency and stride length — but used only from races where the horse was genuinely tested. The noise cleared. The signal emerged.
Case Study: Delacroix and The Cashel Palace Trial
Delacroix lined up in the Cashel Palace Hotel Derby Trial Stakes at Leopardstown as his final preparation before the 2025 Derby.
The race returned a predicted optimal of 11.5f.
On the cusp of the Derby trip.
Here’s the full race career:
| Date | Course | Distance | Predicted Optimal at 3yo |
|---|---|---|---|
| 25 Jul 2024 | Leopardstown | 8f | 9.6f |
| 10 Aug 2024 | Curragh | 7f | 8.3f |
| 14 Sep 2024 | Leopardstown | 8f | 9.1f |
| 12 Oct 2024 | Newmarket | 8f | 9.8f |
| 26 Oct 2024 | Doncaster | 8f | 8.9f |
| 30 Mar 2025 | Leopardstown | 10f | 9.5f |
| 11 May 2025 | Leopardstown | 10f | 11.5f |
Six runs between 8.3f and 9.8f. One outlier at 11.5f.
Watch the Cashel Palace trial back. They crawled through the early stages and sprinted home from three furlongs out. The 11.5f reading wasn’t biomechanics. It was tactics.
He went off 2/1 at Epsom and was beaten 16 lengths.
He subsequently won the Eclipse and the Irish Champion Stakes — both slowly-run middle-distance races where his late acceleration was decisive. A brilliant horse, most effective around 10 furlongs. The stride profile never suggested the Derby was where he’d be most effective. The trial misled.
Quality Data Only
Treat inefficiently run races with scepticism.
Horse racing stride analysis needs more than the raw stride numbers. It needs the context in which those numbers were produced.
The Two-Year-Old Problem
Two-year-olds are harder to model.
The reason is structural. Physically immature 2yos often race at distances that don’t reflect their natural stride characteristics. Few 2yo races exist beyond a mile. Trainers chase three quick runs for a handicap mark, often at suboptimal trips. Stamina horses aren’t yet ready for longer distances.
But stride data can still reveal direction. A future 10f horse running at 6f as a two-year-old may not yet be racing over its optimal distance — but the stride profile can already carry the signal. Lower cadence — markers of aerobic development already visible in a sprint context.
The challenge is reading that signal against the noise of immaturity and race context. That’s where the 7-furlong breakpoint becomes important.
The 7-Furlong Breakpoint
Seven furlongs is a meaningful dividing line in how horses develop.
Using the 2yo model to predict where each horse will be most effective as a three-year-old, two broad groups emerge — those predicted below 7f and those predicted above.
The split matters because the two groups develop differently.
92% of two-year-olds predicted below 7f remain sprinters as three-year-olds. They get faster, not further. Maturity translates into speed — sharper anaerobic systems, more explosive over shorter trips. The development path is narrow and predictable.
Two-year-olds predicted above 7f have a wider range of outcomes. They tend to develop further as they mature — the upside is on the stamina side.
The 7-furlong mark sits around the transition between anaerobic and aerobic dominance. Some horses divide closer to 6.5f, others around 7.5f, but the pattern centres there.
That’s what makes the two-year-old season valuable for stride analysis. A horse winning over six furlongs may look like a sprinter — but if its stride profile is predicted above 7f, the data is pointing towards middle-distance development. The market usually needs the horse to prove that on the track. Stride analysis can identify the direction earlier.
The full breakdown is in Racehorse Development: The Critical 7-Furlong Breakpoint.
Why Minimum SPS Tracks Stamina
The relationship between minimum SPS and distance isn’t accidental.
Horses produce energy through two systems.
Anaerobic — speed. Explosive, powerful and quickly depleted.
Aerobic — stamina. Sustainable over distance and more efficient over prolonged effort.
The balance between those systems changes with the demands of the race.
Minimum SPS provides a useful window into that balance.
Horses capable of dropping to a lower stride frequency during a race can operate at a lower energy demand than those maintaining a higher cadence throughout. For stamina horses, that ability to rev down through the middle of a race is important.
Speed horses are optimised for a different system. The engine needs to match the trip.
Putting It Into Practice
When assessing a potential stayer, minimum SPS is a useful place to start.
- Above 2.20 — miler-plus territory. Treat staying claims with scepticism.
- Between 2.13 and 2.18 — borderline. Middle-distance effectiveness becomes more plausible.
- Below 2.13 — increasingly consistent with a staying profile.
These are guidelines, not rigid cut-offs. Race context matters and so does the individual horse. But the signal is consistent enough to be useful as a filter — a way to eliminate horses racing at unsuitable trips before committing time to deeper work.
Form, pedigree and trainer comments all have value. Minimum SPS adds another question: does the horse’s physical profile fit the trip?
Track Geometry And Stride
Take a typical sprint stride profile — stride frequency of 2.40 and stride length of 7.60. On a course with conventional geometry like York, a general prediction model handles it well. Run the same profile through Epsom’s gradients or Pontefract’s turns and stiff incline and the prediction shifts — sometimes by several furlongs.
The horse hasn’t changed. The track has.
Inclines compress stride. Declines extend it. Bends compress it further. Every British racecourse combines these factors differently.
A general prediction model — trained across multiple tracks — handles neutral courses well. But at venues with pronounced geometry, the same model can be pulled out of shape. The track distorts the stride profile before the model reads it.
StridePredictor now runs track-specific predictions across more than 20 British racecourses. Each model is built from race data at that venue, learning how stride profiles translate to distance on that specific course. The advantage over a single general model is precision — the track’s geometry is accounted for rather than averaged away. The strongest track model returns an R² of 97%. Across the portfolio the average sits in the mid-80s. These are tested against horses the models have never seen — predicted blind, then compared to their actual optimal distance.
At some venues, splitting by course configuration improved prediction accuracy further. Ascot, Newbury, and Newcastle each have distinct straight and round course layouts. Horses stride differently through bends than on the straight. Separate predictions for each configuration produce cleaner results.
Mean prediction error at track level typically falls below one furlong. At some courses it sits closer to half a furlong.
The prediction should describe the horse. Not the horse and the track.
The detail is covered in Standardise the Track, Measure the Horse.
Going And Stride
Going affects stride mechanics. Softer ground shortens stride length — by an average of 1.2% per single step of going change. Stride frequency barely moves. The speed loss comes almost entirely from shortened stride.
That matters for prediction because stride length is one of the model’s inputs. When a horse runs on softer ground, the shortened stride enters the model directly.
But there’s a subtlety. The relationship between stride and distance changes on different ground. Good and Good to Firm produce similar stride-to-distance patterns — at most tracks they can be treated interchangeably. Good to Soft starts to diverge. On Soft ground, the stride compression is severe enough that the patterns shift meaningfully.
A model trained on mixed going conditions struggles because the same stride profile means different things on different surfaces. A horse striding at 7.30m on Good is producing a different signal to a horse striding at 7.30m on Soft.
So each track model is built from data on a confirmed range of going conditions, and separate going-specific models handle conditions outside that range. The model matched to the going reads the stride data in the right context.
Predictions are most reliable from stride data captured on Good to Firm through Good to Soft. On Soft and Heavy ground, the risk of distortion increases. Predictions from extreme ground should be treated with more caution.
The full research is in How Going Affects Racehorse Stride.
When Biomechanics Disagree With Pedigree
Pedigree remains useful. It tells us about inherited probability — what horses with similar breeding have tended to do. Stride analysis measures the individual. That’s where the two can diverge.
A horse from sprinting bloodlines can show a biomechanical profile consistent with middle-distance performance. A well-bred staying horse can produce a stride profile pointing towards a shorter optimum. When that happens, the disagreement is interesting — and it can be where the opportunity lies.
Pedigree tells you what the horse’s family tends to do. Stride analysis tells you what this horse is actually doing. When the two disagree, that’s where the individual horse becomes most interesting.
What Horse Racing Stride Analysis Means In Practice
Stride analysis offers predictive information the market still undervalues.
For punters, it can identify horses racing at unsuitable distances — genuine stayers at value prices, or horses whose apparent stamina claims are based more on pedigree and assumption than biomechanics.
For trainers and owners, it can inform campaign planning, race targeting and realistic distance expectations.
For analysts, it provides another layer of information beyond traditional form metrics.
The opportunity is not necessarily in finding horses that can’t stay.
It’s in finding horses that are being asked to race at a distance that doesn’t allow them to show their best.
A horse can win over the wrong trip. A horse can run well over the wrong trip. The question is whether that performance represents the horse’s peak.
That’s where predicted distance matters.
The Bottom Line
The model will miss.
Horses surprise.
But the question remains useful:
Where is this individual horse most effective?
All horses stay eventually. The question is where they are most effective. That’s what StridePredictor measures.
We’re measuring what the market largely ignores.
See the methodology in action: Epsom Derby 2026 ante-post, Epsom Oaks 2026 ante-post, Derby 2026: Trainer Intent. Or read why “will it stay?” is the wrong question. For weekly biomechanical analysis, the Stride Watch series runs through the season.