Final Time Quandary

[highnote]

Here's an interesting final time quandry. The speed figure that gets awarded depends on which final time chosen -- Keeneland Timer or Trakus.

http://www.drf.com/drfNewsArticle.do?NID=85172

Should you trust Trakus or the timer?
By BYRON KING

Without assigning blame, it is noteworthy that this spring at Keeneland - where American Teletimer, not Trakus, is used for its official times - there were at least two cases of official times differing greatly from times generated by Trakus. Trakus times from Keeneland are listed on the Keeneland website at ww2.keeneland.com/racing/Lists/Trakus/allitemsalt.aspx.

The time of one of those two races gets a test in a rather unscientific manner in Saturday's Louisville Handicap at Churchill Downs. Among the 12-horse field are five entrants exiting the April 27 Elkhorn Stakes at Keeneland, a race that Trakus timed as being run in 2:27.12, 3.28 seconds faster than the official timer at Keeneland did.

One of those two times - 2:27.12 for Trakus, 2:30.40 officially - is seemingly incorrect. It is up to horseplayers to determine which one they trust.

In the overwhelming majority of cases, the Trakus and official times were similar this meet, usually within a fifth of a second or so. That was obviously not the case with the Elkhorn.

{skipping ahead to the end of the article}

Hanshin: Purim exits fast key race

Following a similar Trakus data angle, Purim is my play in the Grade 3, $100,000 Hanshin Handicap at Arlington on Saturday.

Although entered on Polytrack in the Hanshin, his last race came on turf at Keeneland, a close third in the Grade 2 Maker's Mark Mile. That race, like the Elkhorn, received a much faster Trakus time than official time.

Trakus timed Kip Deville narrowly winning that race over Showing Up in 1:33.38, compared with an official time of 1:35.51.

I had awarded Kip Deville an 87 rating on my scale using the official time -- a very good figure.

Next, I changed the final time of the Kip Deville race to the Trakus time of 1:33.38 and recalculated the speed figure. This time Kip Deville earned an 86.

It's interesting that a faster time caused him to get a lower figure. This is explained by the fact that the faster final time made the daily variant lower, (i.e., a faster track). When the track is fast, the horses earn lower figures -- all else being equal.

How do I know the track was fast? Actually, I don't because the final times are in conflict. But if I assume the Trakus time is correct then I know the track was fast because I make projected daily variants. I look at previous races of all the horses that raced on the turf that day.

(see Mordin, Beyer, Brohammer or Quinn for how to make projected variants or figures)

So the main thing I want to point out is that the author of this article might be in error thinking that he should bet Purim in the Grade 3 at Arlington on Saturday. Purim may actually be a little slower than he thinks because he was aided by a fast track.

On the other hand, if the official time is correct then maybe Purim is a good play depending on what figures the other horses in the race might be expected to run tomorrow.

Interesting.

Thoughts?

[Edward DeVere]

So if KD had broken the track record, you would have given him an even slower figure?

And if he had broken the track record by, say, TWO FULL SECONDS, you would have given him, like, a negative twelve or so?

The faster a horse runs, the worse figure he should get - now I see where I've been going wrong all these years.

Hmmm. I wonder if I get fatter and uglier - will I attract more supermodels?

This has possibilities. . . .

[highnote]

So if KD had broken the track record, you would have given him an even slower figure?

And if he had broken the track record by, say, TWO FULL SECONDS, you would have given him, like, a negative twelve or so?

The faster a horse runs, the worse figure he should get - now I see where I've been going wrong all these years.

Hmmm. I wonder if I get fatter and uglier - will I attract more supermodels?

This has possibilities. . . .

It does seem counterintuitive. That's why I was hoping for some feedback. I wanted to know if other speed figure makers would get the same or similar results. Sometimes we make mistakes. When horses come back and run differently than expected we try to find out why -- is it the horse, the track, etc.?

A fast track can only speed a horse up so much and a slow track can only slow a horse down so much. Our software accounts for that. The system we use is self-limiting and statistically based. It bases the variant on the past figures horses have run and also on the times of the other races on the card. So even if the final time of one race is super fast, that race time is tempered by the other times of other races and also the past figures all the horses have run.

I won't go into all the details. You can find them in "Mordin On Time".

I'd say a slow track could probably slow a horse down more than a fast track could speed it up. A horse could run on some kind of a rubber track -- like humans do when they race indoors or it could run through quick sand. I'd say the quick sand would slow it down more than a synthetic rubber track would speed it up.

In Kip Deville's case when I lowered the final time to the Trakus time the difference was only one speed figure point -- about a length.

I just redid his speed figure and artificially lowered his final time to 90 seconds -- down from 93.38. Now the track is even faster, but his figure this time stayed the same -- 86.

[highnote]

The faster a horse runs, the worse figure he should get - .

Yes, if he was aided by the fast track.

Let's say Horse A runs 6 furlongs in 1:10.0 when the track is average -- 0 variant and earns an 80.

Next race Horse A runs 6 furlongs in 1:09.4 on a track that is four fifths fast.

Which is the faster race?

Time is relevant. 1:09.4 is the faster time. But the horse was aided by the fast track. If the track would have been normal then the time would be 1:10.3.

Big difference.

Should the faster race get a speed figure greater than 80?

[BillW]

The official track timer times the race from the first nose triggering the start timer to the first nose triggering the final timer. Trakus measures the individual horse from (i assume) start timer location to finish line with less precision (+/- a foot or two). A dead closer that stumbles out of the gate can show up as quite a difference in these two scenarios. Add in the chance of either system suffering from GP turf timer syndrome and all bets are off. It does bring up a good question (I will make sure and ask it when I'm up there this fall) - What steps does the track take to certify the official time? Backup timer?

As far as speed figs go - the Trakus system is probably better except for the fact that they would be incompatible with those from other tracks.

[highnote]

The official track timer times the race from the first nose triggering the start timer to the first nose triggering the final timer. Trakus measures the individual horse from (i assume) start timer location to finish line with less precision (+/- a foot or two).

According to the article the official time and the Trakus times are usually within 1/5 of a second of each other. So maybe Trakus correlates their times with the start timer.

As far as speed figs go - the Trakus system is probably better except for the fact that they would be incompatible with those from other tracks.

I don't think it would matter as long as you always use the same source of times for a given track. You just adjust your par times to reflect the differences. An 80 on a Trakus timed race at Keeneland should equal an 80 at Belmont.

[BillW]

I don't think it would matter as long as you always use the same source of times for a given track. You just adjust your par times to reflect the differences. An 80 on a Trakus timed race at Keeneland should equal an 80 at Belmont.

In the stumble scenario the horse won't get to the trigger point of the start until possibly a second or so after the "first nose" (let's assume the stumble costs the horse 5 lengths). If that horse closes to be the "first nose" at the finish line trigger point beating the horse that originally triggered the start timer by say 2 lengths the PP times would show the difference in race times between the 2 as 2/5ths of a second rather than 1 2/5ths second. In the classical timing system used by Equibase, each horse is assumed to reach the start trigger at exactly the same time which is not necessarily true. For a wire to wire situation both systems should measure the same +/- system variances.

I should add as a postscript that some tracks provide a start call beaten length figure. That solves the problem.

[highnote]

In the stumble scenario the horse won't get to the trigger point of the start until possibly a second or so after the "first nose" (let's assume the stumble costs the horse 5 lengths). If that horse closes to be the "first nose" at the finish line trigger point beating the horse that originally triggered the start timer by say 2 lengths the PP times would show the difference in race times between the 2 as 2/5ths of a second rather than 1 2/5ths second.


Good point. The horse that stumbled and lost a second, but won the race by a nose, actually ran the race 1 second faster than the second place horse -- assuming the 2nd place horse broke cleanly.

[highnote]

I should add as a postscript that some tracks provide a start call beaten length figure. That solves the problem.

I just checked the Belmont charts and they only give the position at the start. Beaten lengths at the start would be very helpful and would add a lot to the accuracy of the figures.

I suppose you could just make figures based on the time and beaten lengths of the first call to the finish. But that gets a bit messy.