Race Shape Before Runners

Six runners on a galloping track over a mile. Two of the six are confirmed front-runners — they have made the running in their last three races and they want to do it again today. Three of the six are confirmed hold-up horses, ridden quietly and brought late. The remaining horse is prominent, sitting just off the pace.

This is a race shape. It is not a result. It is not a forecast. It is a statement about what the race is going to look like when it is run, and it can be made before any of the runners leave the stalls.

Two front-runners on a galloping mile is a contested early lead. Neither rider will be happy to take a soft second, because at this trip and at this track a soft second from the front is a winnable race for the horse on the lead. So they push. The early fractions go quicker than they otherwise would. The two leaders are running each other's races for the first five furlongs, and by the time the race turns for home both of them are deeper into their reserves than a horse going at its own pace would be.

The three hold-up horses, sitting last, are now in the race. They are being given a softer first five furlongs than usual, in the sense that the field is coming back to them rather than getting away. The strongest finisher among them — the one with the most stamina and the best turn of foot — has a race set up for it that it would not have had if there were one front-runner instead of two.

That is a pace scenario. It is also a rating prediction. The hold-up horses' ratings, for today, are higher than their figures alone would say. The front-runners' ratings, for today, are lower. The prominent horse has a near-neutral adjustment, slightly positive — its running style suits this shape better than the styles of the horses being pulled to extremes.

None of this is colour commentary written afterward. It is calculated before the race runs, from the form of the runners declared, and it changes the numbers attached to each horse.

Race shape is a rating factor

The standard treatment of pace is post-hoc. The race had a strong gallop. The pace collapsed at halfway. The leader got an easy lead. These are sentences written after the race has run, used to explain why a strong horse on figures finished poorly or why an outsider ran well. They are descriptions, not predictions.

This is the wrong order. The race shape that explains a result was, in nearly every case, the same race shape that could be read off the runners before the race. The horses that win in a slowly run race are not different from race to race. They are hold-up horses with a turn of foot in races where the gallop is so soft that the pace cannot collapse — and prominent-running horses in races where the front-runners take each other on. These are profiles. They can be matched against scenarios.

PaceMap's engine does the matching. It classifies every runner's running style from the comments and positions of their recent races — front-runner, prominent, midfield, hold-up — and counts how many of each are in today's field. Two confirmed front-runners is a contested-pace scenario. One front-runner with a history of going off too hard is a soft-lead scenario. No identified front-runner is a slowly-run scenario, in which the race will likely be won by whoever moves first off whatever pace there is. Four or more front-runners is a collapse-likely scenario, in which the race breaks up early and the closers benefit.

Each scenario has a different shape of rating adjustment. A hold-up horse in a slowly-run race is being asked to do something its style does not equip it for — the race may already be over before its kick matters. The engine takes percentage points off its rating. A hold-up horse in a collapse-likely scenario is set up — the engine adds them. The adjustments are not large. A few pounds in either direction, weighted by how confident the scenario read is. But they are direct multipliers on the rating, applied before the value calculation runs, before any of the downstream logic touches the number.

This is what pace as a rating factor means. The race shape is not a thing that affects the result. It is a thing that affects the rating. The result, when it comes, is downstream of the rating.

The order of operations

The thing most worth saying about pace is how often it gets read in the wrong order.

Most form reading starts by finding the best horse on figures and then asking whether the race shape suits it. This is backwards. It puts the rating in front of the shape, and treats the shape as a check on the rating rather than an input to it. When the shape suits, the horse is backed; when the shape does not suit, the horse is downgraded by intuition; either way, the figure-derived rating sits at the centre of the read.

The right order is the other way round. Read the race shape first, from the runners declared. Then see which horses' profiles fit the shape. Then rate the horses, with the shape already accounted for.

This is not a subtle reorganisation. It changes which horses come out on top, sometimes substantially. A horse rated three pounds above the next best on figures, in a scenario whose shape costs it five pounds, is not the best horse in the race today. A horse rated six pounds below the favourite, in a scenario that adds four to its rating and takes three from the favourite, is closer than the figures alone suggest. These cases are not unusual. They are routine. They are also routinely missed by reads that build the rating first and check the shape after.

The engine does it in the right order because it cannot do it any other way. It does not see the favourite. It sees the runners, classifies them, builds the scenario, applies the modifiers, and only then produces the projected ratings that get compared to market prices. The shape is part of the rating because the rating is built downstream of the shape.

When the shape is wrong

The honest part of any pace model is what it does when the shape is not what it expected.

A race scenario is a probability, not a certainty. A field with two confirmed front-runners might still produce a slowly-run race if both riders sit back and dare each other not to make the running. A field with no obvious leader might be turned into a strongly-run race by a hold-up horse with new tactics. These are real outcomes, and a model that prices them out is making a confidence error.

PaceMap's pace classifier produces a scenario and a confidence number. A contested-pace read with two clear front-runners is held more firmly than a read built on horses whose running style is itself uncertain — a horse with three runs and inconsistent comments does not have a settled style yet. The rating modifier scales with the confidence: a strongly-held scenario produces the full adjustment, a weakly-held one produces a smaller fraction of it. The dampening engine then adds its own band around the result, because a rating built on an uncertain shape is itself uncertain.

This is the connection back to the layer the series opened with. Pace is a rating factor; the strength of the conviction in the pace read is itself a thing that has to be measured. Confidence about the shape gets translated into confidence about the projected rating, and the projected rating is then held with the band that confidence implies.

The race shape and the rating it produces are not separate readings. They are the same reading, taken in the order that lets each correct the other. The next part of this series turns to the foundation underneath all of it — the anchored private rating, what the rating is actually measuring, and where the engine's number diverges from the Official Rating most.