New Pace - Reg Speed Sort - Opinions?
 

I don't see how the sort is valid using the horse's last 10 lines, IF the lines, as in many cases, are completely different that today's conditions. Let's use 2 contenders from the last race Richie put up in the Matcher's Forum from FG - a 5.5 dirt sprint.

High Toned has 7 turf lines - 5 short sprints, 1 long sprint, 1 mile route. 70% of his races have nothing to do with today’s condition.

Myleka has 6 route lines, 1 on the turf. 60% of his races have nothing to do with a dirt sprint and even one of the dirt sprints is on a good track.

I consider dirt, turf, and poly to be completely separate worlds differentiated even more so by distance structure. There are tracks where you will get fields with fairly homogeneous lines but a lot of the time, contenders have lines like these.

I don't see what kind of confidence I can have using these 10 races for each horse to determine a high low when today's race is a 5.5f dirt sprint.

Thoughts welcome.

Although I don't mean to speak for Dave Schwartz, I venture the opinion that he would say that he trusts his particular Speed Rating numbers (e.g. HDW's, or perhaps the DRF's as used to illustrate NewPace in his videos) to be properly equalized and manipulated so that they reflect (as best it is possible) a normalized view of these disparate distances, surfaces and tracks. So don't worry about it - if you trust the comparability of your Speed Ratings.

That said, no matter HOW GOOD any vendors SRs are, there are degrees and then degrees of confidence you should have in any attempt to compare the non-comparable. So, if I had to make fine-grained judgements, I might be more confident in the 90% confidence SR range produced for a horse whose distances, surfaces and track were more homogenous, and rather less confident for horses like the above examples.

Everything in our racing analysis deals in degrees of certainty or risk, and what we get paid for prudently undertaking that risk.

That said, maybe there is a way to attach a confidence factor to the ratings produced by the standard NP formula (or any formula for that matter), depending on a measure of non-comparability or noise in a horse's PPs compared to the conditions of the race its entered in today ...?

Ted

Good thoughts Ted. I have no doubt that logistically the SR's in RDSS, Dave's numbers, Beyer’s, whatever, are most likely translated logically but as you know, a turf route translated to a dirt sprint no matter how accurately doesn't represent reality if the horse can't run a lick on the dirt. Also agree that the risk/reward decision is the name of the game.

I want to know 2 things first: What is the par for the race and what did the favorite just do?

If we take the FG sprint, the RC for the race was 85 (I normally use BRIS pars and speed figs) and the favorite had just run that 85 at this track, surface and distance.

Then I want to know: who's got a shot at matching that in their race history and what have those candidates been doing lately (last 4 races). When I say "matching" that, I would look at dirt sprints to make some judgments.

Once I get it down to those horses that can match it and I think have a shot at matching it today, then I look at what class each horse has demonstrated his ability, using to day's class condition as the gauge - CL 5 NW 2 since July 29th, 2011.

I would eliminate Myleka for win contention and review High Toned a bit further, although I'm not thrilled with him.

Such is one of the big problems of handicapping: How do you address the unknown quantities?

Thinking of it in terms of Aristotlean truth (i.e. false/true, 0/1) is probably not a good idea. A better idea is to think of it in degrees of truth: 0.00 to 1.00 and everything in between.

Think of it as a confidence number, representing how confident you are in the ratings.

Some horses' ratings are more reliable because all of their races have been on similar surface, distance, and maybe even same circuit, without significant layoffs. These horses would be as close to a "1.00" as a horse can get.

On the other hand, some horses - more common in turf races - have never raced on today's surface and sometimes not even at the approximate distance. These horses are closer to the low end; closer to 0.00.

In Percentages and Probabilities 2012 I published stats for "Good Race at the Approximate Distance," to address distance shifts. It does not address very well this discussion but is a step in the right direction. Depending upon the conditions of the race (and especially the age of horses) the distance factor itself goes from very important to totally negligible.

And even if a horse was a "1.00," how reliable are the ratings at a particular track? In our par times we have a metric called the "Speed Reliability Index" (SRI) which rates each track on how well the horses run back to their previous ratings.

IMHO, even the 1.00 horse - same surface, distance, etc. - isn't really 1.00 because of the SRI. If you are trying to use speed at a track like Keeneland, forget about it (as they say in NY). On the other hand, at tracks like CBY the horses run very reliably.

Okay - enough diatribe.

For what it's worth, my two cents. All of this to me spells "Variable" and in the words of Jim The Hat it's a dang horse race. Some get out of the gate and then some get out a little late, a horse may set the pace, they may press the pace or they might overcome the pace. All these pars from anyone IMO is a waste of trees and ink. And in the words of Taulbot don't look for something not there. Remember every race is a one off, a horse maybe a repeat winner but he will never repeat the way he won though you will convince yourself through numbers he did.

Interesting comments Dave. My thought process is typically digital - 0's and 1's.

In SoCal, for ex., TUP horses very rarely win, even though their speed sort could be very competitive in the race. Another would be if Mike Mitchell had the horse for 8 races in the PP's and now it was handled by Declan Jackson - the speed sort would look very good but the chances of winning today minimal. I could probably think of more situations that would increase the risk.

I have to think about the "confidence index" and "sri" concepts. Maybe they can be worked in to come up with some sort of a risk adjusted speed sort or something like that.

Trotman: While I agree with most of what you said, I have to disagree with you on the pars. They can add tremendous value to the handicapping process. Class levels at a racetrack are there for logical and economic reasons and horses are campaigned by the trainer up and down those levels logically, at least if they want to succeed. Holds true for both claiming and non-claiming.

Pars can help you determine where a horse might be successful, whether they can successfully stretch out or cut back, etc., can they move up a coulple of levels off a win, etc. One can't be extremely rigid with them but they add a framework.

Class levels, trainers, and how a track is running by distance and surface are realities that can't be ignored. Whenever I have, it has typically cost me money.

In a race where enough races have been run at the distance and surface and I know that the preferred running style is E and 40% of the winners have gone wire to wire, it doesn't make sense to me to be looking for a closer. The closer might win a race or two, but I'll make more money sticking with the profile. How may predicted speed duels have you seen that don't materialize? Too many to count for me.

On the other hand, if the profile is neutral, then running style becomes secondary and just a normal part of the analysis.

Not to be a pendant, but the distinction that you need to make with Aristotlean logic is not that it functions in a binary way, that one can say that 'x is either true or false' but rather that Aristotlean logic functions from the universal to the particular, that is takes a universal premise and tries to see if a particular can be logically subsumed underneath it.

Rather than flog to death the old 'Socrates is a man...' example, here would be a more pertinent example:

All winning racehorses are fast.
Union Rags is a fast horse.
Therefore, Union Rags has won a race.

The general structure is:

All As (winning racehorses) are B (fast).
C (Union Rags, a particular horse) is a B (fast)
Therefore C (Union Rags) is an A (winning racehorse).

The problem with applying Aristotlean logic to horse racing is that it doesn't really apply. Even slow horses can win races - ask any trip handicapper. Theories on horse racing, or ideas which help to qualify which horse can win, are inductive, in that they take a pool of particular experiences, ie a number of races, and then move towards trying to craft a general or universal idea. The problem is that such an idea can't be applied universally due to the fact that it will not be correct every time it is applied. Ergo, there really can be no universal truth in handicapping, except to say that the outcome of a race is that at least one horse has won.

Sure, you need to speak of degrees of truth with regards to which inductively deduced ideas are applicable, but then handicapping becomes and effectively is a meta-problem, in that the question then becomes 'how do I determine which ideas are most applicable given the race I am trying to handicap?'

The differences in the types of handicappers out there, trip, pace, speed, etc, is that their differences in perspective lies at the meta-problem level, in that each thinks a certain foundational idea should be used to resolve the various questions each person faces when they handicap a race.

It's a digression, yes, and not meant to criticize you, but rather I just wanted to write that saying something as Aristotlean in logic as being binary with regards to truth/falsity is insufficient. Inductive reasoning claims to also arrive at true or false conclusions, it's just that one can't be 100% certain about them, a bit like the weather.

I completely agree. In fact, you actually made my point, perhaps better than I did:

Namely, that there are no true 0s and 1s. Everything falls in between. Examples of this can be found in the following statements (which I believe to be true):

1. There is no such thing as a horse that cannot win a given race.

2. There is no such thing as a horse that cannot lose a given race.

3. There is no such thing as an independent factor. All factors are related to all other factors, in some way.

4. All factors are bet by the public to one degree or another. This includes what we perceive as "unique" factors of our own invention. Those unique factors are inevitably comprised of more common individual factors. Those individual factors may well be bet by the public. The exception to this would be information that is totally unavailable to the public such as hidden workouts, barn information, etc.

In other words, what matters in putting together a system is the degree to which a horse is a contender (from 0.00 to 1.00) and the degree which a horse has a chance of winning (probability).

This is why spot play systems, which are, by their very nature, 0s and 1s (i.e. "Horse is a play" or "Horse is not a play") will cause wild fluctuations in results; some are actually good bets and some are not so good. Perhaps some are actually big losers.

One could (should) actually look at all assessments in this light. In my Basics of Winning class I had people grade each race on their confidence level, both in the value of the race but also in the reliability of the handicapping itself. In other words, some races, such as (say) a graded stakes at 6f is a much more reliable race to handicap than (say) a low-end claiming race for N2L. I am not saying that reliability is the most important issue; it is just an issue to consider, and all issues should be considered.