Pacelines...which ones, how many?

[classhandicapper]

Originally Posted by Tom

Classhandicapper and Ranchwest talked about a composite speed figure, where the last three or four speed figs are weighted. Curious what, if any considerations are given the lines. Last 3-4 period? Last 3-4 same distance, surface, class?
Any thoughts?

I've done multiple studies.

In one I specifically looked for horses who last 5 races were all at the same track, same distance, same surface, similar class, and no layoffs, off tracks, trainer changes, equipment changes to complicate things. They were all older horses.

Then I had someone run a regression for me that asked, how should I weight the horse's last 4 Beyer figures to best predict how he will run today? (in this case it was the 5th race).

I can't remember the exact weights, but it was around 50% for last, 25% for next to last; 15% 2nd to last and 10% for 3rd to last.

(Tom, I'm pretty sure I gave you the exact numbers of that study one time in a private message. It was years ago).

The next thing I did was run tests using my own Class Ratings on the following categories using a large sample of Graded Stakes races (for consistency purposes)

Last Race
Most recent similar race
Best of Last 2
Average of Last 2
The 50, 25, 15, 10 formula above
The formula above but excluding the very worst race
Tweaking around with the formula above

Looking at the last 4 races and excluding the worst race produced the best result.

If I recall, all of them were within 2-3 percent.

The major problem with this approach is every horse's PPs are a different adventure. So rules that apply to a study like this of many horses may not apply to individual specific horses.

So I don't use a formula, but I do look at multiple races to appraise the horse giving more weight to recent races.

[cj]

Originally Posted by classhandicapper

I think in those terms also, but when I studied it a bit it didn't seem to gain me much of an advantage in terms of predicting winners.

Maybe that was because:

1. Those 2yos and 3yos I looked at were running against other 2yos and 3yos that were also improving.

2. If the race was long enough ago so that significant development was possible (5-6 months?) some of them did jump up but others were coming off layoffs or weren't the same horse anymore.

I really should do more work on the subject because we know for certain that on average those young horses get better.

Sure, you have to figure out if the horse was an early developer, and if coming off a long layoff if the horse is well meant and if it was injured and maybe won't develop. But if running regularly, it is pretty much a sure think IMO.

[classhandicapper]

Originally Posted by cj

Sure, you have to figure out if the horse was an early developer, and if coming off a long layoff if the horse is well meant and if it was injured and maybe won't develop. But if running regularly, it is pretty much a sure think IMO.

Another thing you can consider is the number of starts the horse has had.

A lot of the young horses are improving in part because they are getting physically bigger and stronger, but also in part because they are getting more racing experience.

Take a horse like Flightline. He wasn't developing because he was young. He was getting better with racing experience.

[cj]

Originally Posted by classhandicapper

Another thing you can consider is the number of starts the horse has had.

A lot of the young horses are improving in part because they are getting physically bigger and stronger, but also in part because they are getting more racing experience.

Take a horse like Flightline. He wasn't developing because he was young. He was getting better with racing experience.

Yes, I forgot to include that one, it is already in there. But the improvement tends to be less if they are older.

[ranchwest]

Originally Posted by cj

I give every race a "score" with how relevant it is to today's race when making figures and think it can be applied to handicapping. The more information, the better, so I'd rarely go with one race. But the relevance score might be so low as to exclude races. I've detailed how I do it here before so not going to type it out again.

Score factors:

Distance
Surface
Days since race
Recency of Race in PPs (last race, 2 back, etc.)
Class (how close is it to today)

One thing I rarely ever see mentioned is adjusting for a horse maturing. If you are using a line from 3 months ago on a 3yo in January, are you building in that it should be faster now? I know that I do.

I also use a weighted selection method a little bit similar to yours. I do have it baked in to try not to reach back into the 2 yo year. And, I don't like races where the horse was not at least a little competitive.

[Lt1]

The CSR factor in RDSS weighs the last 4 lines regardless of surface or dist. I don't put much stock in it. Sartins' guideline for picking pacelines was to use the best of the last 3 comparable at surface ,dist, and class recommending that one stay within the last 5 lines. Generally one will find the useable pl within the top 5 but there are times when I have back further when called for.
Tim

[Lt1]

Here is the definition of how csr is weighted.
CSR – Composite Speed Rating. A weighting of the RDSS Adjusted Speed Rating from the last 4 usable
lines, regardless of distance, surface, race conditions or finish. The weighting is as follows: 100% of the
last line SR + 62% of the average of the last 2 lines SR + 38% of the average of the last 3 lines SR + 24%
of the average of the last 4 lines SR. Thus the last line SR has a major weighting in the CSR, with the
weighting declining for older lines.
Tim

[classhandicapper]

Originally Posted by Tom

The Sartin Methodology thread started a few good discussiond on paceline theory that lost in the People v Sartin defamation trial so I thought this deserved a new thread.

Classhandicapper and Ranchwest talked about a composite speed figure, where the last three or four speed figs are weighted. Curious what, if any considerations are given the lines. Last 3-4 period? Last 3-4 same distance, surface, class?

Take the 4th at Aqueduct today. 6F TURF SPRINT NW1 STATE BRED

Three Unions -her two turf sprints are decidedly better than her two dirt races. Do you use just the turf lines or all four? If yiu use all four, are you concerned that the highest weighted line is a dirt stakes, neither of which are what she face today?

Any thoughts?

I'll get back to the original question.

In a case where a horse is clearly better on one surface than another, I tend to just use the races on the surface it's running on today and give more weight to the more recent races.

One exception would be if the horse hasn't run on today's surface in a long time and I'm looking way back in his record. Then I may look at relative form. In other words, if he's been getting better on dirt, he may also be getting better on turf now than he was way back in his record.

[traveler]

The best paceline to use is to "project" what the horse will run today. Picking a paceline is making an assumption that the horse will replicate a point in time.

Short of that, averaging multiple lines may be the best way.

[classhandicapper]

Originally Posted by cj

I give every race a "score" with how relevant it is to today's race when making figures and think it can be applied to handicapping. The more information, the better, so I'd rarely go with one race. But the relevance score might be so low as to exclude races. I've detailed how I do it here before so not going to type it out again.

Score factors:

Distance
Surface
Days since race
Recency of Race in PPs (last race, 2 back, etc.)
Class (how close is it to today)

One thing I rarely ever see mentioned is adjusting for a horse maturing. If you are using a line from 3 months ago on a 3yo in January, are you building in that it should be faster now? I know that I do.

I have a similar Query I use for testing 1 race.

Do you have any good ideas for how to handle multiple races?

I have Queries now that weight each of a horses last few races, but it doesn't look at surface, distance, layoffs etc... It systematically gives more weight to more recent races.

I want to improve that.

If for example today's race is a dirt sprint, I'd want to give more weight to recent dirt sprint races than dirt routes, but I'm perplexed about whether to eliminate some other races, given them less weight, some of each depending how different it was etc...

For example, maybe I still want to use a dirt mile with a lower weight than a dirt sprint but eliminate turf routes.

Maybe I want to eliminate any bad race if the horses had a layoff right after that?

Theses are the kinds of things I do subjectively in my handicapping, but coming up with rules is a little tricky.

If anyone has any ideas for rules I'm all ears.

[thaskalos]

Originally Posted by classhandicapper

I have a similar Query I use for testing 1 race.

Do you have any good ideas for how to handle multiple races?

I have Queries now that weight each of a horses last few races, but it doesn't look at surface, distance, layoffs etc... It systematically gives more weight to more recent races.

I want to improve that.

If for example today's race is a dirt sprint, I'd want to give more weight to recent dirt sprint races than dirt routes, but I'm perplexed about whether to eliminate some other races, given them less weight, some of each depending how different it was etc...

For example, maybe I still want to use a dirt mile with a lower weight than a dirt sprint but eliminate turf routes.

Maybe I want to eliminate any bad race if the horses had a layoff right after that?

Theses are the kinds of things I do subjectively in my handicapping, but coming up with rules is a little tricky.

If anyone has any ideas for rules I'm all ears.

Let's say that a sprinter has run its last two races in routes, and has tired noticeably in the stretch both times...but today it cuts back to another sprint, where the horse ran impressively three races back. I used to routinely look past the two route races and concentrate on the sharp sprint race 3 back...but then I started thinking: what sort of influence did the last two route races have on the horse's sprint form? Will it still retain its previous sharp form...or will the two recent route races act as a neutralizing force against the horse's sprint-running ability? The same question presents itself when the horse runs two "conditioning" races on the turf and then returns to the much-preferred dirt surface.

I don't like it when a horse runs two races in a row under unfavorable conditions...and then returns to where it belongs. One race is fine...but two races mess up the horse "psychologically", IMO...to the point where the horse fails to return to its prior "glory".

[classhandicapper]

Originally Posted by thaskalos

Let's say that a sprinter has run its last two races in routes, and has tired noticeably in the stretch both times...but today it cuts back to another sprint, where the horse ran impressively three races back. I used to routinely look past the two route races and concentrate on the sharp sprint race 3 back...but then I started thinking: what sort of influence did the last two route races have on the horse's sprint form? Will it still retain its previous sharp form...or will the two recent route races act as a neutralizing force against the horse's sprint-running ability? The same question presents itself when the horse runs two "conditioning" races on the turf and then returns to the much-preferred dirt surface.

I don't like it when a horse runs two races in a row under unfavorable conditions...and then returns to where it belongs. One race is fine...but two races mess up the horse "psychologically", IMO...to the point where the horse fails to return to its prior "glory".

I agree with what you are saying.

When I'm handicapping I can subjectively weigh all the available information. I usually treat something like your example as a small negative because I'm less certain how the horse is going to run than if his sprint race was his last race.

When it comes to creating some kind of computer rules for testing ideas like this, it's much trickier because it could depend on what else I see in the horses record, the trainer, and other factors. You can't code for everything.

Maybe I overestimated the ability of the CAWs also.

Computers are great for studying data, but they are less great for some of the complexities of racing PPs.

[Tom]

My gut instinct is to not rate a horse off lines that don't apply to todays race.
A trouble line, wrong surface, odd track......I don't want to penalize a horse for a racw where he was not expecyed to run well.
A horse wins a MC40 at Aqu. The 4 starts are in Str50000 and he fails to hit the baord.Today, he is in for C40000n2l. I don;t care about the 4 races in the higher class, I would use the MC40 line either alone or with a previous MC line if available. If the maiden line is good enough, that is another thing, but I want to see what the horse rated when it performs well.

[ranchwest]

Originally Posted by Tom

My gut instinct is to not rate a horse off lines that don't apply to todays race.
A trouble line, wrong surface, odd track......I don't want to penalize a horse for a racw where he was not expecyed to run well.
.

I agree.

[classhandicapper]

Originally Posted by Tom

My gut instinct is to not rate a horse off lines that don't apply to todays race.
A trouble line, wrong surface, odd track......I don't want to penalize a horse for a racw where he was not expecyed to run well.
A horse wins a MC40 at Aqu. The 4 starts are in Str50000 and he fails to hit the baord.Today, he is in for C40000n2l. I don;t care about the 4 races in the higher class, I would use the MC40 line either alone or with a previous MC line if available. If the maiden line is good enough, that is another thing, but I want to see what the horse rated when it performs well.

I know what you are saying.

Let me give you a different example.

We are looking at a horse that typically runs figures in the mid 80s. He runs on the slop twice and runs figures in the 60s.

The question is, did he dislike the slop or did he go off form and just happened to run on the slop those last two times?

If you assume he didn't like slop, some percentage of the time you'll find out he was just off form.

All else being perfectly equal, I think the horse that ran mid 80s last out would win more often.

But it's really tricky to code for something like that because there may be other information that can help you decide which it was, like pedigree, other races in the slop, how sloppy the track was etc..