Since Speed is the Long Pole in the Tent

[Capper Al]

Originally Posted by pandy

One of the main reasons why I developed my various handicapping methods is because you have to be able look at races from a different perspective than the average person, otherwise you'll land on the same horses, or favorites. So creating your own set of numbers certainly makes sense, if they're good.

I started testing a new method on my handicappingwinners website yesterday, hit a cold $255 exacta at Indiana last race. I had to sweat out an objection but it turned out it was on the third place finisher. I bet my tests.

This method looks for horses that finish well but are fast enough to stay within striking distance. For the tests, I'm don't look at the pps at all, just speed and pace ratings.

I bet my test too. Why? It doesn't make any sense except it is exciting when trying something new.

[pandy]

Originally Posted by Capper Al

I bet my test too. Why? It doesn't make any sense except it is exciting when trying something new.

Yes, it is exciting trying something new if you think you're onto something.

[Capper Al]

Raybo,

Once a speed figure is computed it becomes a noun so to speak and measurable. The hypothesis is in measuring speed as an attribute of a horse. A comprehensive figure like BPP is a summation of many hypothetical attributes into one final figure in hopes the whole is greater than any part like speed or class alone would be.

[thaskalos]

Does it only happen to me, or do you guys also encounter more than your fair share of "good luck" when you first try out a new and promising idea...only to be disappointed later?

Almost every original handicapping idea that I've ever come up with starts off in the most promising way when I first start testing it...causing me to keep dreaming of horses when I sleep.

[raybo]

Originally Posted by classhandicapper

That's the way I feel about class. No one has the numbers I have.

I have some ideas for late speed and early speed ratings that would be totally unique.

I have some unique ideas about how to incorporate pace into speed figures the way thaskalos is suggesting.

But I think final time speed figures have been researched to death.

I don't know how you arrived at where you are, other than your mention of "trial and error", and that is just longhand for more automated data analysis. Either way will help tremendously in the quest to find out what works best on a set of data. Personally, I did the same thing for over 20+ years using mostly the DRF and published results from the newspaper. The fact that it took me 20+ years should attest that there are better ways to arrive at the same points of understanding about the game. But, however you arrive at it doesn't matter, other than the time it takes (and the money you could have been making by arriving there quicker, of course).

The problem I see with trial and error or more automated data analysis is in the data that is used as inputs. My belief, and that of many others, is that the data input should be as close as possible to the measured data (yes, we know that the measured data contains flaws in many cases, but that is the lot in life for all players that use published data from commercial sources, not a whole lot we can do about that, we must use what we have access to). Otherwise, we never know if the problem is within ourselves or within the data we're using as inputs. That is the basis for my contention, and deep belief that published, often opinionated/biased, and externally manipulated/adjusted figures/assignments (by other human beings and/or their applications/software "of choice") can not be depended upon to represent the measured data adequately enough for serious data analysis.

But, that's just me, take it as you will.

[raybo]

Originally Posted by Capper Al

Raybo,

Once a speed figure is computed it becomes a noun so to speak and measurable. The hypothesis is in measuring speed as an attribute of a horse. A comprehensive figure like BPP is a summation of many hypothetical attributes into one final figure in hopes the whole is greater than any part like speed or class alone would be.

You may believe that, but I don't, and by the way, I don't use, nor do I recommend the BPP, because it puts you among too many others out there. But, the idea of a "performance" rating has a whole lot of merit, as long as it is your creation (or someone you really trust, to be unbiased, and as accurate as possible, and not used by too many other players). If TFUS data files were available, you can bet I would have already made a major change in the data that I use, because I seriously trust the processes that CJ uses to create his figures. And, yes, I would still create my own performance figures from that data.

[traynor]

Originally Posted by pandy

That's not the point of this thread. Remember, the thread is about taking the most predictive handicapping factor (for picking winners) and improving on it. The reason why I brought up these public handicappers is because some people said that speed is not the most predictive factor. As I've said, and I've studied handicapping for over 40 years, the handicappers I've known who picked the most winners said that speed figures were the most predictive single handicapper factor. Now, they knew how to interpret those numbers, but they did rely heavily on speed figures.

And believe me, these guys didn't have an occasional good day of picking winners. They had more good days than anyone, that's why, when it came to picking winners, they were the best. Now, whether or not that makes any difference, some would say that picking a higher percentage of winners than everyone else is meaningless, that's a moot point. If the guy who picks the most winners relies heavily on speed figures, then speed figures are the most predictive single handicapping factor for picking the highest percentage of winners. And someone is going to have to show me some pretty powerful evidence to convince me otherwise.

I don't think anyone beyond the stone cold rookie stage is going to bet every race at every track, every day. I don't think anyone beyond the stone cold rookie stage is going to use raw final times as the only selection criteria.

Once final times are manipulated--with variants, with distance adjustments, with whatever else is used--they can no longer be meaningfully compared to "pure final times," any more than a Ferrari can be meaningfully compared to a Ford pickup. Once the number massaging starts, the "speed figures" that result can no longer be (or should no longer be) lumped together with "raw final times."

Once beyond the "every race at every track every day" stage, the application of any type of filtering process essentially tweaks the results.

It might be more useful to consider what the "long pole in the tent" is for the specific types of races one wagers on. Personally, I have little interest in two-year old races, or maiden races, or races with first-time starters, among a number of other categories. I don't really care (and I definitely do not find it useful to know) what may or may not be predictive in the races I have no interest in betting on.

However, in the races I DO have an interest in betting on, it is a relatively trivial task to code a parsing algorithm to isolate single factors that are associated with winners. Not rocket science. Number of winners with the factor / number of entries with the factor. Ignore ROI completely--focus ONLY on attributes that can be used to predict winners. Absolute basic Handicapping 101 for anything beyond the stone cold rookie stage, either of research or of betting strategies. It makes no difference whatsoever to me what happened back in the day at Aqueduct in December when the wind was blah blah. Or who did what back in the day. I am highly skeptical (to say the least) of anyone's memories of "what happened back in the day."

In the races I DO have an interest in betting on (and in virtually every "advanced handicapping application" I have seen used by serious bettors), final times--including speed figures that are solely time-based--are not even in the top half dozen or so "most predictive factors." In many cases, not even in the top 10. Anyone who believes they are might do well to do a bit more research, rather than relying on what someone said way back when. All it takes to establish otherwise is a relatively modest level of programming skill and a few hours work.

Understand I am not commenting on the every race at every track every day scenario, but on improving the predictability of speed figures. I think that can most usefully be done by considering specific types of races as the first filter to apply. (NOT going backwards and trying to isolate in what types of races final times are most predictive--or seem to be. Finding the types of races first, then searching for ways to improve the performance of speed figures/final times in those specific race types).

[raybo]

Originally Posted by thaskalos

Does it only happen to me, or do you guys also encounter more than your fair share of "good luck" when you first try out a new and promising idea...only to be disappointed later?

Almost every original handicapping idea that I've ever come up with starts off in the most promising way when I first start testing it...causing me to keep dreaming of horses when I sleep.

Yup! They either fail miserably in the beginning, or look good early on and then fail miserably later. That is probably because we are using recent observations for those ideas, rather than hard data over a lengthy period of time. Heaven forbid that I ever bet one of those half-cocked ideas!

[whodoyoulike]

Originally Posted by Cratos

The poster "Thaskalos" is suggesting the right approach and I will add that Trakus data would be better than Equibase/DRF data for this analysis; and Bayesian statistics would be the "engine" to drive this analysis.

I mentioned this to Capper Al a couple years ago. He has good ideas but, he's been focusing on a piecemeal basis instead of combining speed, class and pace etc., into a comprehensive approach. But, before combining them he has to be certain each is valid.

And, I think another mistake he and others are making is ... they are attempting to apply their approach to all types of races (if not every) which is probably impossible to do and will be expensive IMO.

I agree with you about Trakus info. It is very good. I haven't used it as much as I should only because it wasn't available when I was developing my methodology and I may screw up my method(s) more if I change.

[traynor]

Originally Posted by classhandicapper

I've experimented with that kind of stuff endlessly.

My experience is that when you start introducing more factors into a number it can sometimes cause just as much damage as good in terms of accuracy.
That's why I decided to finally test all this stuff with live data and answer some of these questions definitively instead of by trial and error betting experience.

The evidence so far suggests to me that I am better off using my methods of classing horses as the primary driver and using speed and pace figures as a minor component instead of the primary components.

A point at which they may become more useful.

[Capper Al]

Originally Posted by thaskalos

Does it only happen to me, or do you guys also encounter more than your fair share of "good luck" when you first try out a new and promising idea...only to be disappointed later?

Almost every original handicapping idea that I've ever come up with starts off in the most promising way when I first start testing it...causing me to keep dreaming of horses when I sleep.

Yes, it happens to me too. Maybe that's why I bet when I'm testing.

[Capper Al]

Originally Posted by raybo

You may believe that, but I don't, and by the way, I don't use, nor do I recommend the BPP, because it puts you among too many others out there. But, the idea of a "performance" rating has a whole lot of merit, as long as it is your creation (or someone you really trust, to be unbiased, and as accurate as possible, and not used by too many other players). If TFUS data files were available, you can bet I would have already made a major change in the data that I use, because I seriously trust the processes that CJ uses to create his figures. And, yes, I would still create my own performance figures from that data.

You can believe what you believe too. I think this is silly because we both seem to end up in the same place but have different words to describe it. I make my own comprehensive figure also from my attributes that I make.

[whodoyoulike]

Originally Posted by classhandicapper

That's the way I feel about class. No one has the numbers I have.

I have some ideas for late speed and early speed ratings that would be totally unique.

I have some unique ideas about how to incorporate pace into speed figures the way thaskalos is suggesting.

But I think final time speed figures have been researched to death.

I understand you work for DRF but, I don't know how they're structured.

Do you have the opportunity to discuss handicapping with Randy Moss?

I think he's one of the best regarding handicapping in general. He gets it IMO.

[Capper Al]

Originally Posted by whodoyoulike

I mentioned this to Capper Al a couple years ago. He has good ideas but, he's been focusing on a piecemeal basis instead of combining speed, class and pace etc., into a comprehensive approach. But, before combining them he has to be certain each is valid.

And, I think another mistake he and others are making is ... they are attempting to apply their approach to all types of races (if not every) which is probably impossible to do and will be expensive IMO.

I agree with you about Trakus info. It is very good. I haven't used it as much as I should only because it wasn't available when I was developing my methodology and I may screw up my method(s) more if I change.

In the end, I handicap by race type. I've said that in this thread and elsewhere. Every attribute has to stand on it's own and have a positive impact value to make it into my comprehensive figure. Trakus sounds interesting, but I haven't finish my R&D with BRIS data and believe I will be finished when I finished with BRIS.

[Capper Al]

Class,

If you do work for DRF, I'm jealous.