What the sport writes down about incidents

After every race a steward writes a short note on each runner: whether it was crowded, baulked, missed the break, or had a clear run. Those notes are published as free text and nowhere as data. We have parsed them.

249,678 comments parsed · measured 2026-09-14

What is in the record

249,678
Comments parsed
44,687
Runs with recorded trouble
128,547
Runs recorded as clear
1,413
Falls and knock-overs recorded
Recorded incidentTimes recorded
Crowded58,245
Bumped29,290
Forced to check14,905
Baulked11,835
Badly crowded2,438
Badly bumped2,352
Badly baulked2,334
Checked2,130
Slightly crowded1,377
Impeded1,363
Stumbled1,252
Moved off951

These records cannot tell you which track is safer

This is the most important thing on the page, and it is why you will not find a safety league table here. Tracks do not record on the same system, and the difference is large enough to swamp anything you might want to measure.

TrackRunsDistinct terms usedFalls per 1,000
Thurles (IE)3,5071173.42
Kilkenny (IE)6,3281196.16
Cork (IE)4,4501196.97
Dundalk (IE)3,6151307.75
Youghal (IE)6,3411309.3
Towcester (UK)13,0961,8043.36
Hove (UK)10,7861,7224.36
Newcastle (UK)11,4201,6406.04
Doncaster (UK)9,6261,6342.8
Star Pelaw (UK)5,1201,4883.91

Irish stewards work from a controlled vocabulary of a little over a hundred terms. British stewards write free text and use more than ten times as many. One British track, Romford, records no falls at all across roughly ten thousand runs — and the words “fell”, “fall” and “knocked” do not appear in its comments even once. That is a property of its recording template, not of its racing surface.

Nor is the bias correctable, because it does not run in one direction: recorded fall rates vary from zero to over ten per thousand runs without following vocabulary size. Any ranking built on these numbers would be measuring how stewards write.

What the record can answer

  • Whether a particular dog met trouble. Within one dog's career the same stewards are writing, so a run described as badly baulked is comparable with its other runs. This is what the form pages use.
  • How one track's recording changes over time. A trend within a single track holds recording practice roughly constant.
  • What the sport does and does not write down. The gap between 44,687 runs recorded as troubled and 1,413 falls recorded across the entire corpus is itself a finding about the record, not about the racing.

Method, and how to check it

Comments come from the Greyhound Board of Great Britain and from Rásaíocht Con Éireann, the two bodies that regulate the sport. We parse the steward's own words into fields and add nothing: where the note says nothing, we record nothing rather than inferring a clear run. Coverage and known defects in the underlying data are published in full on our data quality page.

Journalists, veterinary and welfare organisations are welcome to quote these figures. Please include the measurement date above — the numbers move as coverage improves — and we will supply the underlying extract on request.