My fundamental approach to handicapping codified in a simple rating method

[Bill Cullen]

In my experience, using intra-race speed rating pars works better than using speed rating pars derived from an average of races for a particular class at a particular distance and surface at a particular track .

I define an intra-race par as the second highest speed rating among all the last out speed ratings of the horses in today's race. Using that as a cutoff, and using only non-maiden races where all the horses show at least 10 races in their past performances, I circle the three most recent speed ratings of each horse where those speed ratings equal or exceed the intra-race par. In addition, I put a check mark next to each of those three circled speed ratings where the particular race was on the same surface as today's race AND was within one-half of a furlong of today's distance.

Then I covert those circles and check marks to a recency score. The process is simple: A circle or check mark is given a score depending on how recent the race is in the sequence of each horse's last ten races. A horse that has a circled speed rating for its last race gets 10 points, a circled speed rating for the second race back gets nine points, etc. The tenth race back gets one point for a circled speed rating.

An example should make this clearer:

If a horse equaled or exceeded the intra-race par in its second and fourth races back, but in no other races among its last ten races, it's speed ratings for those two races should be circled and thus the horse would get 9 points and 7 points respectively for meeting or exceeding the intra-race par in those two races. In addition, if both of those races were on the same surface as today's race AND within a half a furlong of today's distance, then the speed ratings for those two races would also have check marks next to them and would also receive 9 and 7 points respectively. The total score for the horse 9+7+9+7 or 32.

In my experience, if play is restricted to the top rated horse (no ties), you should nail about 30% winners and roughly break even. If you take 5/2 or higher odds, you should beat the game.

This rating method could easily be tightened up by insisting on at least one par race within the last 45 days as well as backing no horse that ran a lifetime best Beyer last out.

In my opinion, the biggest deficiency of this rating method is that it doesn't into account the early pace. No matter. You can easily subject the highest rated horse to a simple pace analysis. As a matter of fact, even though this rating method is significantly more comprehensive than a spot play, you will still get the best results using this rating method in conjunction with other valid handicapping factors. That's how I use it.

It's been my experience with this approach that you can expect every third race to yield final odds of 5/2 or higher on the highest rated horse. Since the method generates about three plays per day per track, you should get about one play per day per track if you go for value (5/2 or higher).

This rating method is basically the one being tested currently under the "Selections" thread entitled Todays' picks from a new method being tested. But I have been using it for years. I personally feel quite confident about it. It's been and continues to be my most fundamental approach to handicapping. It's as close as I can come to quantifying a lifetime of experience into a simple method that can be used quickly.

Bill Cullen

[andicap]

Bill,
I've tinkered with the same approach in recent years, experimenting with 3rd best in last 90 days with inconclusive results. Seen a number of horses not meet the "par" and still win, but rarely older claimer/allowance/stake types.

Two issues I have are

a) number of races back may not be as relevant as "days ago"
In other words if horse "A" bettered par in his last two races but they were 350 and 400 days ago, he gets a better score than horse "B" who bettered par in his 2nd and 3rd races back, 45 and 80 days ago but who might have a valid excuse for last race OR just threw in a random clinker and figures to bounce back today.
It happens.

Wouldn't you be better off putting a chronology on your points instead of "races back?"

e.g. purely hypothetically with no background on this at all.

21 days -- 10 points
35 days -- 9 points
49 days -- 8 points

etc. down to

1 to 2 years -- 2 points
more than two years -- 1 point


B) Wouldn't work nearly as well for lightly-raced horses who continue to improve and haven't met the pars but figure to be "explosive" or steadily improving based on recent races.

For lightly-raced younger horses I have pretty much foresworn any type of "par" in favor of

1) pace analysis
2) form cycle analysis looking for horses that have recently bettered their earlier ratings.
3) class analysis-- e.g. drop to 3 yr old claimers is huge here.

Obviously speed ratings are important because the more the horse has to improve the higher the odds I want! And horses that don't figure to improve can be figured to run what they have in the past or worse if they the erratic sort.

[Shacopate]

(dramatic pause)......and the winner of "Outstanding Title for a thread and coolest closing line in a post" goes to......Bill Cullen.

As for the method, like Andicap I use something similiar for contender selection. I like your "new top" pitchout and Andi's ideas for points based on days.

What if you did the same thing for pace ratings and combined the two? Cj's pace figs are free. If the track in question isn't available, maybe you could use something simple like simulated pace.

My best days have come when I took the time to make an odds line and bet the highest rated horse that exceeds fair odds. The Fiero templates and Meadow bonus reduce the time involved. Nothing is completely accurate in racing, but one thing I've noticed when comparing speed fig's is "fast is fast" and in most cases "close" is usually good enough when making odds line.

[andicap]

The new tops idea is tricky. With drugs horses are pairing up more than ever now -- its cost me a few shekal throwing these out.
I now only throw out horses who ran MAJOR tops. 9 or more Beyer points OR who are runing back within 17 days or so.

Still, knowing individual trainers is probably the best weapon for know when to trust tops or not.

[Bill Cullen]

Originally posted by andicap
Bill,
Two issues I have are

a) number of races back may not be as relevant as "days ago"

B) Wouldn't work nearly as well for lightly-raced horses who continue to improve and haven't met the pars but figure to be "explosive" or steadily improving based on recent races.



Andicap,

I am taking your first suggestion into account but with a different twist: instead of the first circled race being ten, the second back being 9, etc, I am assigning the most recent circled race by any horse 50, then the second most recent race by any horse is 49, etc. This takes more time in terms of work but gives more weight to more recent races by actual racing date and also makes recency a competitive inter-horse measure between horses rather than being an intra-horse standard measuring recency serially (ie, 10, 9, 8, 7) without reference to actual racing dates across horses.

Your second suggestion is not as relevant for me personally since, if you look at my post again, I only use this rating method against races where all horses have at least ten races in their pp's.

Thanks for your excellent comments on this rating method.

Bill C

[Bill Cullen]

Originally posted by Shacopate
[B

What if you did the same thing for pace ratings and combined the two? Cj's pace figs are free. If the track in question isn't available, maybe you could use something simple like simulated pace.
[/B]

Yes, I thought about using pace figures. I know from Nunamaker's and other folk's work that pace figures like the Brisnet ones have impact values approaching speed figures and that the ROI's are sometimes as good or better.

Because I use the Form primarily and because I tried to take into account the recreational handicapper (thus the desire for a relatively simple and fast rating method), I chose not to use pace figures.

But the same principles would seem to apply and you might well get a better ROI. I think that poeple tend to asociate speed figures more with recency than with pace figures, which folks tend to associate more with a horse's habitual running style at the earlier stages of the race than with recency per se.

Thanks for your insight. Should you reserach this using pace figures, please let me know your results and I will likewise keep you abreast of my adapting the rating method to some of the suggestions offered by you and Andicap so far.

Bill C

[andicap]

Bill,

Why do you use the 2nd best last racing rating as the "par.?"
Have you tested this?
Just wondering becuase I'm looking for the best rating as well to use as the par. been more trial and error for me.

[Bill Cullen]

Originally posted by andicap
Bill,

Why do you use the 2nd best last racing rating as the "par.?"
Have you tested this?
Just wondering because I'm looking for the best rating as well to use as the par. been more trial and error for me.

Andicap,

I used the second best SR last out as the intra-race par because I recall several studies including Quirin and other empirically-sound guys finding out that the second best SR last out's winning percentage is pretty close to that of the best SR last out's winning percentage.

And it works. I say this more from years of observational experience than just two sets of one hundred races each that I just tested.

I just finished the summary of the second hundred races for the method (I haven't posted the results yet). The results were

100 playable races
28 winners
181.10 return (9% loss)

The winners are definitely at least short two or three (there were a disproportionate number of horses coming in second).

Think of it: just using one intra-race par and then simply converting races at par or better into a 10, 9, 8, etc scores and adding the scores gives you 30% winners with a near break even rating method (the jury's still out on the ROI but, again, from years of observation, I believe the ROI will fluctuate between -3% to +3%). This is a strictly mechanical rating method (and, I might add, I very simplified one) that gets three plays per day per track. Just by confining your bets to 5/2 or better, I am firmly convinced that you will get at least one play per day per track and will beat this game!

Now imagine what you can bring to this rating method by subjecting the highest rated horse to a simple pace analysis, post position, current form, jock, trainer, and other valid considerations.

If I were to summarize the key insights that I learned over years of handicapping that are embodied in this rating method, I'd say:

1) Handicapping only races where every horse has at least 10 races gives you an edge of in terms of having sufficient data to work with. Obviously for this kind of rating method using serial intra-horse recency, you really need ten races per horse.
2) Par is best determined by the more recent (doesn't necessarily have to be the last race out) speed ratings recorded by the horses themselves in today's race.
3) Having a lot more speed than today's intra-race par is not really that much of a decisive factor because most of the time horses tend to run at the class/speed levels where they belong. Thus I am able treat races run at today's intra-race par or better as being functionally equal in value. This ties in with the idea of using the second highest SR last out.
4) Since the intra-race par filter is binary (either a race meets par or it doesn't), races that qualify on par can be measured on recency alone (and intra-horse serial recency at that.)

Althought the rating method isn't totally comprehensive, it's fairly comprehensive as far as a quick and simple to use rating method can be. It considers significant aspects of speed/class/pars/recency/distance/surface/form.

Thanks,

Bill C

[andicap]

Thanks, Bill.

Re your point about giving the most recent checkmark 50 points and 2nd most recent mark 49 points. I disagree.

I'm not so sure there should be a major inter-horse recency competition. Don't believe you should give a horse more credit for having his race 14 days ago versus another horse's race 22 days ago.
It's often a negative to have a good race in such a short time frame.
And statistics today often show no difference at all in IV between horses coming off 28 or 45 days rest. Often the freshening is just what they need.
Hence, my sugestion for a RANGE of days off giving more credence to recent races because horse's are generally more likely to run back to recent races, JUST NOT necessarily to THE MOST recent races.

Good discussion,

[Bill Cullen]

Originally posted by andicap
Thanks, Bill.

Re your point about giving the most recent checkmark 50 points and 2nd most recent mark 49 points. I disagree.

I'm not so sure there should be a major inter-horse recency competition. Don't believe you should give a horse more credit for having his race 14 days ago versus another horse's race 22 days ago.
It's often a negative to have a good race in such a short time frame.
And statistics today often show no difference at all in IV between horses coming off 28 or 45 days rest. Often the freshening is just what they need.
Hence, my sugestion for a RANGE of days off giving more credence to recent races because horse's are generally more likely to run back to recent races, JUST NOT necessarily to THE MOST recent races.



Good discussion,

Andicap,

You're probably right. I have no doubt that if someome uses ranges of days since the last race and applies this to the rating method, they will see significantly better results.

Bill C

[Bill Cullen]

Originally posted by Shacopate
(

What if you did the same thing for pace ratings and combined the two? Cj's pace figs are free. If the track in question isn't available, maybe you could use something simple like simulated pace.



I believe this method could be well suited for use with the Brisnet Class Ratings and might well generate a positive ROI considering that intra-race pars based on speed ratings seem to generate a break-even ROI. One would simply plug in the Class Ratings in the method where speed ratings are used.

Any thoughts?

Bill C