Every score carries the provenance it arrived with, and the national board counts every comparable score in full — scratched rides, unridden scheduled shows, and off-scale classes count not at all. Verification is a federation-origin trust state on a rider's profile that gates claiming and planning surfaces; it plays no part in the public board's ranking arithmetic. Every ranking is a transparent model rather than an opinion, and every prediction is a calibrated range — never a single false-precise figure. Here is exactly how it is built.
| Tier | Description | Counts |
|---|---|---|
| Official | Result mirrored from a federation-anchored source of record. | IN FULL |
| Reviewed | Held row approved by human review against its source. | IN FULL |
| Self-reported | Rider-submitted record — staged for verification, never in ranked metrics. Not currently displayed. | NOT SHOWN |
| Held | Ambiguous identity or source conflict — parked in the review queue, not canonical. | EXCLUDED |
| Off-scale | Freestyle, young-horse, equitation and in-hand classes — not comparable on the standard scale, so they add zero ranked points. | EXCLUDED |
| Abstain | Evidence incomplete — we show “unknown / needs review” instead of guessing. | NO ANSWER |
Every counted score sits inside a rolling four-month window — inside it a score counts in full, outside it it drops out entirely. Long enough to absorb one bad test, short enough that current form drives the table.
Your last counted rides produce a 50% and 80% confidence interval around a predicted next score — drawn against the line you have to clear, never a single deterministic number.
Each official is a slow-moving signal. We split a judge's scored rides chronologically and compare the early half against the recent half — average score, and average deviation from the rest of the panel on multi-judge classes — surfaced by level.
Placements add a fixed bonus on top of raw percentages — 10 points for a win, stepping down to 1 for fifth — and only in classes scored on the standard scale. Off-scale classes add nothing, so the table stays score-dominated.
The leaderboard covers a rolling four-month window: a score counts in full inside the window and drops out entirely when it leaves.
Each result that is score-comparable contributes its percentage score plus a placing bonus: 1st +10, 2nd +7, 3rd +5, 4th +3, 5th +1. A 68% ride that wins its class contributes 78 points.
Results that are not score-comparable — freestyle/Kür, young-horse, equitation, and breed/in-hand classes — contribute zero: neither their score nor a placing bonus. Their scales aren't comparable to standard tests, so they don't belong in a standard-scale ranking.
A rider's total is the sum over all their comparable results in the window. More quality rides = more points; the board rewards both form and activity.
The first published version of the DPI methodology. It describes what the shipped code actually computes — nothing aspirational. Scores count in full or not at all; the next-score forecast is published as a calibrated range with its measured coverage, never a single false-precise number; and where the data is too thin to answer honestly, we abstain instead of guessing.
Calibration cycle 04. Index covers USEF and USDF feeds. Rank is computed over a rolling 4-month window.