Predicting Speed Rating from Past Ratings

[HUSKER55]

Mike,

beyer himself said his figures could float 20 points at the same track. How often do you see a horse go from 80 to 100 opposed to , 80 to 60?

I believe that pace determines final time, and pace is determined by a host of variables.

However, I have been wrong before.

GL

[Robert Fischer]

Interesting.

[classhandicapper]

Originally Posted by TrifectaMike

Ok. I (the new guy in post 1) take your advice (linear regression) make a visit to a local university and ask for help. A grad student, for pay, is willing to help. He directs me to get a years data for all tracks and performs a linear regression.

A short time later he delivers his results. The grad student knows nothing about horse racing. His results show that one weight is insignificant and it is not w4, which if I believe that more recent performances are more important...seems odd to me. But he can't help any further.

Mike

I don't know what kind of data you sent, but I think you might get different results if you sent data for high quality horses vs. horses from 3rd string tracks whose figures are much more volatile for a wide variety of reasons.

That's why when I was attempting the same study, I tried to control for every factor that causes volatility in figures. I wanted to know the underlying reality of how important each recent race was and try to deal with other factors (distance changes, surface changes, class changes, off tracks etc...) as a separate study. I didn't want those things clouding my weights for each race.

The results I got back were extremely logical.

[TrifectaMike]

I receive a call from this stat guy. He wants more time. He tells me after discussing the problem with two other stats guys, he will normalize the data differently and also transform it. Okay, by me. And I wait and wait.

Finally, he indicates that he is stumped and the pattern is REAL. However, he wants a friend, a Bayesian (he could have just as well said a Martian), to take a look at it.

Several days later I receive a stack of printouts. Once again all Greek to me.

[TrifectaMike]

The Bayesian guy agreed with the other stats guys.

When taken as a group, four speed ratings, the third rating back is INSIGNIFICANT!

What does it mean and what are the implications?

Mike

[DeltaLover]

Originally Posted by TrifectaMike

The Bayesian guy agreed with the other stats guys.

When taken as a group, four speed ratings, the third rating back is INSIGNIFICANT!

What does it mean and what are the implications?

Mike

That its impact is neutral and can be eliminated from the model without to affect its precision?

[classhandicapper]

Originally Posted by TrifectaMike

The Bayesian guy agreed with the other stats guys.

When taken as a group, four speed ratings, the third rating back is INSIGNIFICANT!

What does it mean and what are the implications?

Mike

That would be consistent with my results if he also told you that the 4th race back was insignificant.

I was given back the optimum weights for using the last 3 races and for using the last 2 races. He told me the gain from using the 3rd race back was so small, it probably wasn't worth the extra calculating effort. He said I would be fine if I just looked at last 2.

I did not have the 4th race back in my data. If I was told the 3rd race back was not significant but the 4th race back was significant, I'd disregard the entire study as having some kind of flaw in the sample.

It's logical that there's a point where if you go back any further the races are close to meaningless, but I refuse to believe 3rd race back is insignificant and older ones are significant.

What did he say about 4th back?

If you have the ability to continue down this line, the next step would be:

What if the last 3 races were on turf, but today's race is on dirt and he has a dirt race way back? (stuff like that)

[Cratos]

Originally Posted by TrifectaMike

Let me ask a question, which appears to have an obvious answer.

I am fairly new to horse racing and have heard about the importance of speed ratings to predicting winners.

So, I think I've come up with a good idea.

"I'll take a horse's last four races and assign each speed rating a weight"

Last speed rating weighted by w1
Second back speed rating by w2
Third back speed rating by w3
Fourth back speed rating by w4

The question; how to assign values to the weights?

Some here have probably done something similar. Anything goes.

Mike

Mike, what you attempting to do is weigh timefigures as though they were speedfigures; it’s not going to happen now or ever because time is constant and continuous.

When a horse runs a certain time for a race that time will not be changed by some arbitrary adjustments.

However what can be done is to create a timefigure (not speedfigure) based on “conditional time impacts” that will be a “what if” timefigure if the arbitrary adjustments are met and that is where the “weights” come in probabilistically.

[cj]

All this generalization will get people into trouble. Every horse is different, and every group of PPs is different. There are times the third or fourth figure is more significant than both of the most recent. There are times they are meaningless. Sure, as a group you can reach these conclusions. But when you ignore the outliers, you miss a lot of the best prices. This is especially true when we are talking about speed figures where the most recent ones drive the betting.

[ReplayRandall]

Originally Posted by cj

All this generalization will get people into trouble. Every horse is different, and every group of PPs is different. There are times the third or fourth figure is more significant than both of the most recent. There are times they are meaningless. Sure, as a group you can reach these conclusions. But when you ignore the outliers, you miss a lot of the best prices. This is especially true when we are talking about speed figures where the most recent ones drive the betting.

Couldn't agree more. On Sept. 26th, I posted the following:

I have been involved in a "group" methodology for over 2 years with long time friends, using DRF formulator, the Sheets, Thorograph, BRIS, Timeform, a private "X" program and replays. We only invest our synergies on Fri. Sat. and Sun. for most A tracks, some B and C tracks and rare "honey hole" tracks. We look for only one thing..... comparison outliers. This involves a bit of meticulous chart, odds and record keeping to find those entities that have an edge at certain tracks, at specific distances, over differing surfaces and class level, et al. It is a continuing work in progress with some successes but never attaining full confidence, as we still are limited by sample size......BTW, we as a group have concluded the risk/reward and profitability of playing Mon. thru Thurs. is simply not worth our time or energy......

[PhantomOnTour]

Originally Posted by TrifectaMike

Let me ask a question, which appears to have an obvious answer.

I am fairly new to horse racing and have heard about the importance of speed ratings to predicting winners.

So, I think I've come up with a good idea.

"I'll take a horse's last four races and assign each speed rating a weight"

Last speed rating weighted by w1
Second back speed rating by w2
Third back speed rating by w3
Fourth back speed rating by w4

The question; how to assign values to the weights?

Some here have probably done something similar. Anything goes.

Mike

You are?
You have started some threads that, on the surface, seem pretty deep and not what I'd expect from someone new to this game.

[Robert Goren]

Originally Posted by PhantomOnTour

You are?
You have started some threads that, on the surface, seem pretty deep and not what I'd expect from someone new to this game.

TM is hardly new to this game.

[arno]

Using regression for handicapping horses just doesn't work
as a horse with a back figure 8 races back might be more meaningful than the last race on a different surface, class, pace scenario etc.

By the way using regression analysis with a weighted average has a statistical term........Exponential smoothing.

[Magister Ludi]

Perhaps there is a cycle of equine performance consistency. W3 could be the minimum performance consistency point.

[PhantomOnTour]

Shouldn't you be projecting PACE first and SPEED (final time or figure) second?