Help creating sectional time based pars for individual runners

[JustRalph]

You are working too hard. Tons of great software out there that will do this kind of stuff for you with the click of a mouse

[gamblegamble]

Originally Posted by Dave Schwartz

I'm a bit confused.
If you don't have the distance, how will you measure the speed?


Also, you might want to study up on a pace approach called Simulated Pace. It was first mentioned in a book called Scientific Handicapping back in the '60s.

The concept is based upon the PACE OF RACE is whatever the winner ran. Thus, if the winner ran (say) a "90" speed rating, then the pace of race was a 90.

Thus, a horse that was (say) 5 lengths back at any call would get an "85."

This makes it very simple to base the ratings on just beaten lengths.

I recently did a large international project. Outside of N. America, the horses do not typically run fast paces. This creates ridiculously slow pace times and makes the velocity-based numbers pretty worthless.

Simulated pace did a much improved job, despite its obvious lack of "time reality."


Dave

You divided the pace par time by the pace of the race to get the pace rating.

To get the pace of the horse you calculated how manly lengths behind the horse was at the call. You then divided by the length of the race multiplied by the call. Essentially this race was 6f and the call was 4f so the margin was weighted at 66.6%.

I don't have lengths behind at the call, I have the individual time the horse ran to the call. So instead of the leader only I have the sectional time for all horses.

[Dave Schwartz]

Originally Posted by gamblegamble

You divided the pace par time by the pace of the race to get the pace rating.

To get the pace of the horse you calculated how manly lengths behind the horse was at the call. You then divided by the length of the race multiplied by the call. Essentially this race was 6f and the call was 4f so the margin was weighted at 66.6%.

I don't have lengths behind at the call, I have the individual time the horse ran to the call. So instead of the leader only I have the sectional time for all horses.

So, you are missing key pieces of information.

If I said that a horse went some distance in 1:05.37 how can you possibly know if it is fast or slow?

Question:
If you have (say) 50 races at this same track and distance, will the FRACTIONAL DISTANCES be consistent for all the races?

[steveb]

Originally Posted by gamblegamble

All very helpful and informative replies but my actual question was not answered, probably from poor phrasing from my part.

I have two data sets.

One is based on race time, leader to 400m, leader to 800m, final 400m, ect.

The other is the same but for individual horses. How fast the individual horse ran to the 400m, 800m, final 400m, ect.

When dealing with the second dataset of individual horses should I base the pars off the first dataset which is based on race time, leader to X, finishing time. Or is there merit in calculating some sort of median, mean or mode based on all the individual horse sectionals?

why do you even need 'pars' for the sections?
if you have the individual sections then all you need to do is find what % of total time the winners ran each section.


same for leader sections so that you have a baseline for leaders and one for winners.


then regression analysis is your friend.


every race will have a different best way to pace, depending on what the leaders have done.
you don't need pars as such.

i guess then that means a 'par'% for leaders and a 'par'% for winners

if the leaders go that %, then the winner should have went this % type of thing.
you will find very strong correlation.

[gamblegamble]

Originally Posted by Dave Schwartz

So, you are missing key pieces of information.

If I said that a horse went some distance in 1:05.37 how can you possibly know if it is fast or slow?

Question:
If you have (say) 50 races at this same track and distance, will the FRACTIONAL DISTANCES be consistent for all the races?

I think there's a misunderstanding, probably my fault for not being clear.

To answer that question, if par time was 1:05.00 then is was 0.37 above par. Yes the sectionals will be taken from the same points at the same track and distance.

I believe I have the required information to create a pace rating against a par. If for some reason I had to know the margin because time wasn't good, I can't compered why this would be though, I know how many lengths are in a furlong so I can calculate meters per second and produce a margin.

Lets use your example of a 6f race, the call is the first 4f.

You have the time of the leader to the first 4f and the margin of horse behind the leader at this call.

I have the time ALL horses individually ran to the 4f. So the horse that was 13 lengths behind at the 4f, I have what time that horse ran the 4f in.

This would be the format as an example

Horse A to the 2f 25.45 / to the 4f 47.96 / last2f 22.76 / time 70.72 / margin +1.3
Horse B to the 2f 25.47 / to the 4f 48.03 / last2f 22.92 / time 70.95 / margin -1.3

So my question is for the pace rating of horse b can we just divide the par against the time to the 4f.

Maybe this will make this clear for me, if you were working out the pace of the horse for the leader, how would you do it?

[gamblegamble]

Originally Posted by steveb

why do you even need 'pars' for the sections?
if you have the individual sections then all you need to do is find what % of total time the winners ran each section.


same for leader sections so that you have a baseline for leaders and one for winners.


then regression analysis is your friend.


every race will have a different best way to pace, depending on what the leaders have done.
you don't need pars as such.

i guess then that means a 'par'% for leaders and a 'par'% for winners

if the leaders go that %, then the winner should have went this % type of thing.
you will find very strong correlation.

Thankyou for the suggestion, very helpful.

[Dave Schwartz]

Originally Posted by gamblegamble

So my question is for the pace rating of horse b can we just divide the par against the time to the 4f.

Maybe this will make this clear for me, if you were working out the pace of the horse for the leader, how would you do it?

Yes, this is different and you can make this work.

But don't worry about the links. If you want them you can simply create some "phony-but-consistent" length values. The rest of us would much rather have the actual time of the horse.

Now, about what you want to do...

Theoretically, you could do that but the absolutely best way to make pace numbers I have ever seen comes from Jim Cramer at HDW.

What he does is to take the fractional times and extend them to the finish. It creates an amazing rating.

__________
Let's say that you have a horse that ran 6f in 22.00, 45.00, 1:10.0

Further, let's say that the par for the track at 6f is 1:11.00. (i.e. 71.00)
___________
70.00 is 1.00 seconds faster than par.
Convert that to create the SPEED RATING based upon whatever lengths-per-second you are using.
___________
6f/2f=3.0
22.00 x 3=66.00
Comparing the 66.00 to 71.00 tells us that this RATE OF SPEED is 5 seconds FAST.
Use your lengths-per-second method to turn that into the 2f rating.
___________
6f/4f=1.5
45.00 x 1.5 = 67.50
3.50 seconds fast.
Use your lengths-per-second method to turn that into the 2f rating.

So, the only piece you are missing is the lengths-per-second calculations.


Hope this helps.

[steveb]

Originally Posted by Dave Schwartz

Yes, this is different and you can make this work.

But don't worry about the links. If you want them you can simply create some "phony-but-consistent" length values. The rest of us would much rather have the actual time of the horse.

Now, about what you want to do...

Theoretically, you could do that but the absolutely best way to make pace numbers I have ever seen comes from Jim Cramer at HDW.

What he does is to take the fractional times and extend them to the finish. It creates an amazing rating.

__________
Let's say that you have a horse that ran 6f in 22.00, 45.00, 1:10.0

Further, let's say that the par for the track at 6f is 1:11.00. (i.e. 71.00)
___________
70.00 is 1.00 seconds faster than par.
Convert that to create the SPEED RATING based upon whatever lengths-per-second you are using.
___________
6f/2f=3.0
22.00 x 3=66.00
Comparing the 66.00 to 71.00 tells us that this RATE OF SPEED is 5 seconds FAST.
Use your lengths-per-second method to turn that into the 2f rating.
___________
6f/4f=1.5
45.00 x 1.5 = 67.50
3.50 seconds fast.
Use your lengths-per-second method to turn that into the 2f rating.

So, the only piece you are missing is the lengths-per-second calculations.


Hope this helps.

the problem with that, is that not all sections at all tracks are the same.


2f here might be downhill or on a turn, over there, it's uphill and straight.


it's assuming that everything is proportionally the same.
for an isolated track it would be fine, not for comparing them though.

[MJC922]

Originally Posted by steveb

the problem with that, is that not all sections at all tracks are the same.


2f here might be downhill or on a turn, over there, it's uphill and straight.


it's assuming that everything is proportionally the same.
for an isolated track it would be fine, not for comparing them though.

The research will show that it's a non-linear relationship. True to historical form (most) handicappers tend to do things that make sense but aren't supported by the hard data.

[Dave Schwartz]

Originally Posted by steveb

the problem with that, is that not all sections at all tracks are the same.


2f here might be downhill or on a turn, over there, it's uphill and straight.


it's assuming that everything is proportionally the same.
for an isolated track it would be fine, not for comparing them though.

Yep.
There is ALWAYS missing information.

For example, what is the run up?
And does it change?
(Of course it does.)

Numbers still need to be transformed into "pars" any way.

Thus, PAR FIGURE at 4f at one track might be (say) 87 and at another 91.

[classhandicapper]

Originally Posted by steveb

the problem with that, is that not all sections at all tracks are the same.


2f here might be downhill or on a turn, over there, it's uphill and straight.


it's assuming that everything is proportionally the same.
for an isolated track it would be fine, not for comparing them though.

I have a problem with turns.

It isn't difficult to figure out average times from looking at all the result data at different distances, but figuring out how much more "energy" is used running 23 on a turn vs. 23 on a straight can be trickier, especially when the turns are different from track to track also.

[classhandicapper]

Originally Posted by MJC922

The research will show that it's a non-linear relationship. True to historical form (most) handicappers tend to do things that make sense but aren't supported by the hard data.

Agreed.

My experience suggests it's not the same from horse to horse even when they have similar ability. Horses have a different balance of natural speed and stamina so they seem to react differently to same change in pace and then in non linear ways as it gets more extreme.

[MJC922]

Originally Posted by classhandicapper

Agreed.

My experience suggests it's not the same from horse to horse even when they have similar ability. Horses have a different balance of natural speed and stamina so they seem to react differently to same change in pace and then in non linear ways as it gets more extreme.

Agree, to be clear, I'm fine with adjusting for improper energy distribution and to a lesser extent even ground loss, I believe both are part of a true comprehensive rating for the overall past performance.

However if we want to be even more predictive we would then have to back out some of those adjustments for horses that are headstrong and refuse to rate. Similarly, horses that sweep wide every race due to running style would need the adjustments for ground loss to be rolled back at the tail end of the process.

There's only so much to be gained with all of this granularity and there are an awful lot of inputs missing to do it right in the first place. I'm ok these days with knowing what I have now is less than being fully comprehensive as a performance measurement, it's fairly solid but nowhere near perfect. The top trip handicappers will always be more formidable for re-ordering finishes within a race vs the computer IMO.

[ReplayRandall]

Originally Posted by MJC922

The top trip handicappers will always be more formidable for re-ordering finishes within a race vs the computer IMO.

One of the select few edges that are left...