What could you do with a model that accurately predicts final race time?

[coachv30]

Maybe I'm not as schooled and diverse in my ability, but I think knowing the final time within a second or so would still leave a lot to be determined.

Let's say it allows us to narrow a field from 11 to 7 horses. The handicapping is not done.

Who will run to these expectations today and who will bounce?

[davew]

The things I could see using the information for is making better variants for the day and figures in the race. The quality of the race could also be documented.


So not helping me on picking winner of current race, but improving forecasts for future races.

[JerryBoyle]

Originally Posted by davew

The things I could see using the information for is making better variants for the day and figures in the race. The quality of the race could also be documented.


So not helping me on picking winner of current race, but improving forecasts for future races.

Exactly, dave. That was the original purpose. Still going down that route, but just was giving thought to other ideas.

Quality of the race is interesting. Back when I did it, I sampled some races where the prediction was far off from the actual (> 3-4s). In those cases, something "odd" happened, whether it was an event that slowed all runners down, or a single runner that ran away from the rest. This had me thinking of also predicting the finishing time of the 3rd place horse. I could then use expected finish of 1st and 3rd as a sort of measure of the pack vs potentially an outlier.

[JerryBoyle]

Originally Posted by coachv30

Maybe I'm not as schooled and diverse in my ability, but I think knowing the final time within a second or so would still leave a lot to be determined.

Let's say it allows us to narrow a field from 11 to 7 horses. The handicapping is not done.

Who will run to these expectations today and who will bounce?

Correct. Not meant to be a perfect solution. Just to improve accuracy ever so slightly.

[steveb]

Originally Posted by JerryBoyle

Really two questions:

1. What could you do with this model?

2. What average error would you find good/ok/bad? That is, if the model, on average, is off by 1 second, would unacceptable would you consider this. Perhaps another way to think of it is, when looking at a race, how accurately could you guess the final race time?

I've toyed around with building one a few times, and they're reasonably accurate, but not great. The average error is about 1s across all races/distances. Obviously, 1s on a 6f race is much worse than 1s on 1 1/4 mile race. Originally, I thought I might use it as a way to determine if a track was running slower/faster on a given day by comparing the difference in estimate vs actual for all races.

would you not be better off trying to predict the speed, rather than the time?
pretty sure it would be easier, and with the added bonus that the difference
between your expected and the reality, would be an indication of going.

[JerryBoyle]

Originally Posted by steveb

would you not be better off trying to predict the speed, rather than the time?
pretty sure it would be easier, and with the added bonus that the difference
between your expected and the reality, would be an indication of going.

Hey steve, not sure what you mean exactly. If currently what I'm fitting the model to is time, do you mean first normalizing the time for distance prior to fitting?

The honest answer for why I chose time was simply because I have the final time provided in the historical data I purchase. But I'm definitely open to ideas.

[JerryBoyle]

Just as an example, below are the results for GP and SA for yesterday, 4/6/19. Note that some races are missing, as there are some minimum start and number of runner requirements to be able to predict. Endless possibilities around working w/ the numbers, but a quick glance suggest SA ran slower than usual yesterday.

SA:
   race_number  final_time   final_est  actual - estimate
0            1      106.97  107.265846          -0.295846
1            2       78.07   74.918915           3.151085
2            3      105.47   99.359329           6.110671
3            4       70.73   67.561844           3.168156
4            6       77.46   74.999336           2.460664
5            8      111.28  106.704811           4.575189
6           10      123.02  123.784393          -0.764393
7           11       93.71   96.135582          -2.425582

GP:
   race_number  final_time   final_est  actual - estimate
0            1      101.54  102.001259          -0.461259
1            2       97.68   97.258904           0.421096
2            3       65.94   66.565193          -0.625193
3            4       72.81   71.816132           0.993868
4            5       96.04   96.320381          -0.280381
5            7       87.75   90.156960          -2.406960
6            8       71.97   72.435570          -0.465570
7           10       88.30   89.940712          -1.640712
8           11       72.00   70.108253           1.891747
9           12       89.40   90.944725          -1.544725

[steveb]

well i don't know how you may do things over there.


for me.....i know how fast any given race should be run, and would express that in numbers similar to beyer(but they're in reality just ratios)


basically i expect any one race to be run in a proportion of standard(par).
so if the standard =100 and for 6f the standard time may be 70.
thus if it runs 70 seconds, you give it a 100.


for 'x' class, u know from past data that they average 98% of par(which would give a beyer of.....(98-90)*10=80


so if they ran 71 seconds..... (standard/time-.9)*1000 = beyer

or (70/71-.9)*1000 = 86



so you expected 80 and got 86


thus it's 6 points faster.


that's basically all it is in simple terms

[steveb]

and your estimates are out of whack, because you would expect 1 second slow at 6f to be something more than 2 seconds slow at 12f or 1.7+ at 10f.......
your actual - estimated needs to be a % figure and it would make more sense i think.

[JerryBoyle]

Originally Posted by steveb

well i don't know how you may do things over there.


for me.....i know how fast any given race should be run, and would express that in numbers similar to beyer(but they're in reality just ratios)


basically i expect any one race to be run in a proportion of standard(par).
so if the standard =100 and for 6f the standard time may be 70.
thus if it runs 70 seconds, you give it a 100.


for 'x' class, u know from past data that they average 98% of par(which would give a beyer of.....(98-90)*10=80


so if they ran 71 seconds..... (standard/time-.9)*1000 = beyer

or (70/71-.9)*1000 = 86



so you expected 80 and got 86


thus it's 6 points faster.


that's basically all it is in simple terms

Ahh, got it. Thanks, Steve. I do also get beyers in the data I purchase, so I could predict them, or my own variation of them if I try to make my own speed figures. I will take a look at this. It's very straightforward to replace the final time with the final beyer and regenerate predictions.

[bobphilo]

Originally Posted by JerryBoyle

Still processing the responses, but one thing I want to clear up is that the models I've played with/tested take the entire race as an input (all runners and some info about each) and outputs 1 number, which is expected time for the race. So I unfortunately don't have a time for each runner. Just the expected time for the race. Obviously, this is the model's expected time for the winner, but there's nothing connecting that time to which runner the model thinks will run that time. So operationally that restricts what can be done with this info.

I take the same view. Knowing what the final time of a race will be is useless unless I also know which of the horses will be be the one that runs that time. This is an entirely different proposition.

[bobphilo]

Originally Posted by storyline

If you're able to accurately project final figures for each horse this would be your starting point...work back from that.


Everything else is noise

if what your trying to do is find the winner (horse with he fastest time) for today's race, projecting final figures for each horse would be the final step. Estimating the winners final time would mainly be useful for getting a projected figure in creating a variant for the race.

[davew]

Originally Posted by JerryBoyle

Quality of the race is interesting. Back when I did it, I sampled some races where the prediction was far off from the actual (> 3-4s). In those cases, something "odd" happened, whether it was an event that slowed all runners down, or a single runner that ran away from the rest. This had me thinking of also predicting the finishing time of the 3rd place horse. I could then use expected finish of 1st and 3rd as a sort of measure of the pack vs potentially an outlier.

I have to wonder how often the 'official time' is off. For the race estimates that are 3 seconds off, can you hand time replay or does it depend on track and distance?

[ultracapper]

accurately predicting splits is exponentially more helpful than final time. An acceptable +- for final time should be pretty tight to be of value IMO. Maybe 2/5 give or take at 6 furlongs. Predicting 1:12.2 and accepting anything between 1:11.2 and 1:13.2 is much too liberal IMHO.

[JerryBoyle]

Originally Posted by davew

I have to wonder how often the 'official time' is off. For the race estimates that are 3 seconds off, can you hand time replay or does it depend on track and distance?

That's a great question, and I honestly do not know. I'd be hesitant to try hand timing after reading cj's posts, just because it sounded as if it's not as simple as starting a stop watch when the gate opens and stopping it when the first runner crosses the line. The predictions I posted for SA above were off by 3+ seconds so I'll give one of them a shot and see what I get.

I'd be interested to hear from anyone who does make their own figures on whether or not SA was running slower than expected that day.