Data quality & coverage
Everything below is counted from the database when this page is built, not asserted. We publish the gaps in the same place as the totals, because a figure is only worth quoting if you can see what it excludes.
236,587 resulted runs · 2026-01-01 to 2026-09-12 · measured 2026-09-13
What we hold, per run
| Field | Runs | Coverage |
|---|---|---|
| Stewards' commentThe free-text note the racecourse steward writes about the run. Parsed into structured incidents — see below. | 231,155 | 97.7% |
| Starting priceThe on-course price returned at the off. | 225,162 | 95.2% |
| Betfair exchange priceBetfair Starting Price. Lower than SP coverage because the exchange only settles markets it ran. | 120,020 | 50.7% |
| Racing weightTaken at the kennel before racing. | 221,694 | 93.7% |
| Sectional timeSplit time at the first bend. Only some tracks record one. | 143,763 | 60.8% |
| Position at each bendRunning order at each bend, read from the stewards' note. | 171,763 | 72.6% |
The parsed stewards' comments
The steward's note is the only record of what actually happened in a race — whether a dog was baulked, missed the break, or ran a clear race. It is published as free text and nowhere as data. We parse it.
How much of it a second source confirms
Every result is checked against Sporting Life and Timeform as well as the governing body's own feed. A race counts as confirmed only when every runner both sources hold has the same finishing position and the same dog in the trap. Where they disagree, we say so on the race page rather than pick one.
What is wrong with it
These are defects we have measured and not yet fixed. They are listed here so that anyone using the data knows what to exclude.
- 54 races hold finishing places but no first place. A runner's row — usually the winner's — failed to load. These races should not be used for strike rates.
- 3,082 race numbers appear twice within one meeting. A track running two cards in a day is stored as a single meeting, so the numbering collides. This is now described rather than merely counted: every race also carries which card it is on and its place on that card, so 2,415 races are identifiable as belonging to a day's second card rather than being ambiguous.
- 2,731 races carry a source race number that disagrees with their actual place on the card. Sources number the same day differently — one runs 1..26 straight through, another restarts at 1 for the evening — so the number printed against a race is not a reliable way to refer to it. Use the race's own identifier or its off time.
- 13,573 races have no off time recorded, so they cannot be placed on a card at all. They are ordered by their source number, which is the best that can honestly be done with them.
- 372 races give more than one dog a first place, of which 213 are not dead heats and so contain an error. The sport does produce dead heats, and the regulator's own record does not separate the two cases. Where a second source covers the race, the verification above records which dog it confirms; where none does, we publish the count rather than pick a winner and present it as fact. Do not use these races for strike rates.
- 468 races carry no trip length. Any speed figure derived from them is meaningless.
- 2,175 of 180,544 beaten runners carry a time identical to the winner's, which is a copy rather than a measurement. Separately, and not counted here, many beaten times across the sport are calculated as the winner's time plus the recorded margin rather than clocked individually — a limitation of the sport's own recording, not of this database.
What was wrong, and when it was fixed
Defects we found by measuring the data against its sources, with the date each was corrected. They are kept here because anyone who exported the data before that date has the defect in their copy.
- Fixed 31 Aug 2026Irish and Northern Irish runs before 21 Aug 2026 carried the finishing order as the trap number, and every runner in a race carried the winner's time. Repaired from the regulator's race records; Irish trap statistics before that date should be treated as unreliable in any earlier export.
- Fixed 1 Sep 2026On days when a track ran two cards, one source's results could land on the other card's races because both cards numbered from 1. Races are now matched by their off time; the affected days were re-loaded. The two-winner count above is what remains of this.
- Fixed 2 Sep 2026Races after the twentieth on a long day (Harlow, Doncaster) were rejected before they reached the racecard; their results loaded, their form and selections did not.A run recorded as not finishing was dropped entirely; it is now kept without a finishing position.Sporting Life publishes off times in UTC; through British Summer Time none of its records matched ours, so no result had a second opinion. Converted; the affected days were re-checked.Irish sprint grades (S1–S9, 325 yards) were treated as schooling trials and left out, for both cards and results. Irish sprints are being re-collected from January 2026.Dundalk's racing had been filed under Derry and Drumbo Park's under Dundalk, from a wrong track-code table; Derry, Mullingar, Enniscorthy, Newbridge, Lifford and Longford had not been collected at all. Meetings were moved to the right tracks and the missing tracks are being collected from January 2026.Irish racecards are now loaded in advance (they were only ever created from results the night after).The trainer on the day is missing for most British runs from January to August 2026 and is being filled from the regulator's meeting records, about ten days' racing per night. Until it is, trainer figures for that period use the dog's current trainer, which is wrong for roughly one run in nine.
Where it comes from
Results and stewards' comments come from the Greyhound Board of Great Britain and from Rásaíocht Con Éireann (Greyhound Racing Ireland), the two bodies that regulate the sport; Irish racecards come from the same regulator, a day or more ahead. Results are checked independently against Sporting Life and Timeform, which is what the verification figures above count. Exchange prices come from Betfair. We link to the primary source on every page that quotes one, and we add no figures of our own to the record — the parsing turns the regulator's words into fields, it does not invent values.
If you cite this page, please quote the measurement date above: the numbers move as coverage improves.