I enjoy my job making the Timeform Speed and Pace Figures every day, but I enjoy it best of all when the best and fastest horses race.

And so I was happy about the racing at Gulfstream Park last Saturday. Along with Grade 1 action in the Donn and Gulfstream Park Turf Handicap, two stakes races on the undercard included some of the top turf sprinters in North America. The Ladies’ Turf Sprint Stakes (race 4) and the Gulfstream Park Turf Sprint Stakes (race 10) were both intriguing, and race 4 was of particular interest, as it featured the brilliant filly Lady Shipman, runner-up in the 2015 Breeders’ Cup Turf Sprint.

Lady Shipman (1/5 odds) cruised in the 4th by more than 2 lengths, in 58.14 seconds, and Power Alert (5/2) took the 10th by a neck in 57.74. The times reported for the latter race seemed fine, but race 4 just didn’t make much sense given the stylish ease with which Lady Shipman won. It was time for me to go to work.

What did the data say?

Both races were run on the same turf course, there was no change in the turf course condition over the day, and the rail was set at 0 feet for both races.

Regarding the rail setting: Rails are portable, moveable. Tracks move them around to even out wear and tear on turf courses, but in this case they were in the inner-most position for both races.

The “run up” distance was originally reported at 69 feet for both races (for a refresher on run up, see this earlier piece on the topic).

Based on these data points, and the reported final times, my initial work with the data available led me to assign Lady Shipman a provisional 114 TimeformUS Speed Figure. However, several details gave me pause:

• Lady Shipman had run at least a 118 in all five of her most recent tries and as fast as a 123.
• She was now an older, more mature filly, one who looked thoroughly professional on Saturday: a top filly at the top of her game.
• The first quarter, reported as 23.22, was much slower than would be expected.
• The final time also looked slower than would be expected.

In short, that 114 Speed Fig just didn’t match my overall impressions of the race. This alone is never enough to justify making decisions about potential problems with data, but it will certainly cause me to look deeper into the issues.

Based on experience, my initial thought was that the actual run up was different from what was reported. The general practice is that racetracks relay the distance of the run up to the Equibase employee(s) on site, and that distance is then entered into the massive horse racing database that is maintained by Equibase. But there are times when the track will not follow the script. Instead, the track will unexpectedly change the position of the gate.

Gulfstream itself does a great job of providing replays of every race on Youtube, in HD no less, and I can then time the races myself with video-editing software. I don’t claim that timing races in this fashion is perfect. Due to varying camera angles, there is some margin of error, but this approach is accurate enough for this purpose: finding errors that affect clockings. From extensive testing of this method, I believe the margin of error on my end is no more than +/-  1/10th of a second for a race.

I started investigating Saturday’s video by looking at Race 10. The run up is pretty easy to see in this race because the 5/8ths pole is clearly visible. There are six lengths of rail between the position of the starting gate and the spot where timing of the race should begin: the pole.

I confirmed this by backing the video up from the finish line to the pole. The results matched the reported official final time, as they should have.  The next image shows the next landmark I found, a reddish, bush-like piece of shrubbery, and it can be seen on the replay that there are exactly 10 more lengths of rail after it to the 5/8ths pole.

Next, I went to the replay of Race 4, the Ladies’ Turf Sprint.  Something was definitely amiss.  The reddish bush was much closer to the gate, as can be seen in the following image:

I confirmed it was the same bush, and that it was only four lengths of rail (at most) from the starting gate to this landmark.

Clearly, the run up was vastly different in the two races under review.

In Race 10, there were a total of 16 lengths of rail between the starting gate and the bush. In Race 4, there were only four. There is some dispute about the length of a section of rail. Equibase uses eight feet, while others I’ve consulted have told me it is 10 feet. I don’t know for sure which is correct, but for our purposes here, it isn’t that important. The race was 12 lengths of rail shorter for Race 4, or anywhere from 96 feet to as much as 120 feet.

How long was Race 4?

This went some distance toward explaining the odd looking times, but something else didn’t add up. If Race 10 was run at five furlongs (a furlong is 660 feet) and had a run up of between 48 and 60 feet (six lengths of rail), just how long was Race 4 if it was at least 96 feet shorter?

The answer is that a race described as being run at five furlongs with a run up of 69 feet was actually run at less than five furlongs and with no run up at all. It was time for me to check in with Trakus.

Trakus is the official timing company for Gulfstream Park and also provides data on distance traveled for each horse in each race, at the finish and several points along the way. There was no race chart available for race 4 when I initially contacted the company. The Trakus employee I spoke with was very helpful and was able to fix the issue and display the chart. The Trakus chart confirmed what I suspected: The distance of the race was indeed less than five furlongs by their calculations as well.

Taking all this into account, I still had to calculate TimeformUS Speed Figures for these races. It would seem obvious that Race 4, run at a shorter overall distance in a slower official time, was not as fast. But this is where the role of run up asserts itself over the distance of a race.

Using the video replays and timing from the gate and making adjustments for the actual distance run, Race 4 was approximately 4/5ths of a second faster than Race 10 after Race 4 is normalized to the same distance—five furlongs with 60 rail lengths of run up.

How we’ll handle it in TimeformUS PPs

The TimeformUS Past Performances will display pace and speed figures for Race 4, but they will also show a ? due to the problems noted (Related article the TimeformUS question mark system). They will be based on times from the video replays, not the official chart times. This is the best that can be done with the information available. Lady Shipman will now be assigned a 123 TimeformUS Speed Figure for the race. I’m confident that the 123 reflects how well she ran on Saturday, and I am thrilled to be rid of the 114 that she would have received if her TimeformUS Speed Figure had been based on the information that was originally provided to us.