Horses that ran their lowest beyer last out

[Bill Cullen]

A few years ago I checked five samples of a hundred horsres each in open non-maiden claiming company to see what the results were when a horse ran his lowest Beyer within the last 10 races shown. The following filters were used for inclusion in the sample:

1) Today's race (only the race where the horse's sample was pulled; today's race was never used for compiling the results) was an open non-maiden claimer. The definition I'm using here for an open claimer is a race where there were no restrictions on past performance. Restrictions on gender, age, state of breeding, etc, were all permitted. Optional claimers were not allowed.

2) The horse being considered for sample inclusion had to show at least 10 races in his past performance lines.

3) For the horse's last ten races, there had to be a number ( zero included) for each Beyer listed for the last 10 races. No dashes or blank spots were acceptable.

4) There could be no maiden races for the horse's last 10 races.

5) The horse's lowest Beyer for the last 10 races had to be unique (ie, no ties for the horse's lowest Beyer was acceptable).

6) The horse's lowest Beyer could not have occured in its last race.

That's it as far as the filters were concerned. What MUST be kept in mind was that I if a horse met the above criteria, I checked the very next race after it's lowest Beyer race. The 'result race' could have occurred anywhere from the 9th race back to the last race (the most recent race)

Here are the results for the 5 samples averaged together:

15% winning percentage
14 % ROI on the dollar

The range of the winning percentage across the samples was from 12% to 18%. Every sample was profitable.


The median winning payoff across the five samples was S10.20.
Each of the five samples median winning payoff was also ten dollars and change.

From scanning the Racing Form over the last few years, it seems that a playable race comes up about once a day per track. Keep in mind that I have not played this system consistently and methodically enough to determine if it is indeed profitable in the long term. I strongly suspect, though, given both the statistics I cite above, plus an informal workout of the sytem over the last two years or so, that the system is indeed profitable over the long term.

Does anyone have any thoughts about this system and where I could go to get it independently verified?

I'm a new member to Pace Advantage as of today.

Looking forward to talking with you all going forward.

Take care,

Bill Cullen

[BillW]

Bill,

Welcome to the site.

I like the lowest beyer last out angle and i'm actually surprised that the median mutuel is as low as it is, but I guess you find a lot of injured horses with big drops in class that the crowd likes to make the favorite. I play CT and PHA and see a lot of NYRA horses getting bet hard (happened today at PHA ... he finished out of the money).

There are a few here that have beyers databased that may be able to help you. It may take a day or so for them to come across your post.

Again, welcome,

Bill

[andicap]

I'm confused.

Your last rule is:

6) The horse's lowest Beyer could not have occured in its last race.

So when do you bet the horse if not off his last race? What race are you evaluating? Or do you mean his last race in the PPs (on the bottom of the page)?

In Speed to Spare, Cardello talks about cycling patterns where horses are recovering from bad races with a good race still below his top where the horse may have bounced off to run his worst race.

In Ragozin's material, he stresses this "recovery" race can't be closer than 2 points (I believe around 6 Beyer points, right?) off his recent top (unless the horse had proper race) or it could set back the horse again.

[Bill Cullen]

andicap,


The result race was the one immediately following the race in the horse's 10 past running lines with the lowest Beyer. I didn't go about looking for races with lowest-beyer-last-out contenders; I only looked at the past performances of horses that were entered in non-maiden open claimers. Then I would examine that horse's past ten races, look to insure that a unique lowest beyer existed in any race from the second race back to the 10th race back, and then, assuming the horse qualified on the other filtering rules, I looked at the NEXT race immediately following the lowest beyer race to see if the horse won that race or not.

That's it.

Bill Cullen

[andicap]

OK,
so if I'm looking at Wed's card, I'm looking for horses who ran their lowest Beyers in their last race, correct? As long as they fit all of your rules.

I would also check horses who had their lowest beyers in their NEXT TO LAST RACE and improved by at least X points in their last, while remaining somewhat below their tops.

Also have you checked your data to see how they did at various odds points?

Also, was there a single $80 winner that skewed the results so that if you missed that day you would have broken even.

With a system like that I would think you also want to see if you can eliminate losers to get the win% closer to %20 and cut down the losing streaks.

[ranchwest]

If I understand Bill C correctly:

1) Use a qualifying horse's PP's as a data set
2) Find the low Beyer among the races in the PP's
3) Utilize the race following the low Beyer

If that is the case, my problem is that today you don't know whether the last race Beyer will be the low Beyer five races hence.

Now, is EVERYONE confused?

[BillW]

Unless I'm misunderstanding, the confusion lies in the differentiation between Bill's test method and his implementation. He described his test method above (lowest Beyer not the last race out) in which he arrived at the statistics he posted. i.e. he did not test his method in real time but rather with a stack of pp's.

When playing the angle, the low Beyer IS the last race out and meeting the posted filter criteria.

Bill

[CryingForTheHorses]

When I look at the form at my race. I reall dont pay a lot of attention to the beyers, A horse can have a 50 beyer and the next race the beyer shoot up to a 72.
I look at the speed rating and the track varient..I add them together...I add all the same distance numbers on each horse. You will always find the the top 3 numbers will be in the hunt at the finish.

[Bill Cullen]

Originally posted by BillW
Unless I'm misunderstanding, the confusion lies in the differentiation between Bill's test method and his implementation. He described his test method above (lowest Beyer not the last race out) in which he arrived at the statistics he posted. i.e. he did not test his method in real time but rather with a stack of pp's.

When playing the angle, the low Beyer IS the last race out and meeting the posted filter criteria.

Bill

Bill W,

Thank you for clarifying the ambiguity in my original message. You got my meaning exactly right!

To those whom I was less than clear, my apologies.

Just read Bill W's remarks I quote above.

thanks,

Bill Cullen

[ranchwest]

Originally posted by Bill Cullen
Bill W,

Thank you for clarifying the ambiguity in my original message. You got my meaning exactly right!

To those whom I was less than clear, my apologies.

Just read Bill W's remarks I quote above.

thanks,

Bill Cullen

What if the Beyer continues to plunge? Wouldn't the test case suggest that the horse not be played until the Beyer hits bottom? How do you currently know when the horse has reached bottom?

[Brian Flewwelling]

Bill, welcome to the site, and thanx for the interesting (and reasonably clear) post.

Your idea suggests a very interesting area of study.

Brian

[Secretariat]

Mark Cramer would love that method Bill. Great contrarian thinking. How large was your sample size? What tracks? Small or large...

You've got me wondering if highest odds ever might work as well.

Ranchwest makes a good point. If one could find an additional filter or two that could up the win percent a bit with no diminishment in the ROI you might have something. Fundamentally, no one likes a horse who appears to have tanked in the last race. Does your method also include horses who broke down or did not finish, OR threw their jockey or are coming off extended layoffs?

I gather it did.

[Bill Cullen]

In the first post in this thread, I laid out how I tested my low-beyer-last-out system. The emphasis was on how I filtered out horses instead of selecting horses that qualified. Down below in this post, I re-phrase the rules so they appear as a system normally would:

1) Today's race is an open non-maiden claimer. The definition I'm using here for an open claimer is a race where there were no restrictions on past performance. Restrictions on gender, age, state of breeding, etc, were all permitted. Optional claimers were not allowed.

2) For a horse to be a system play, it has to show at least 10 races in his past performance lines today.

3) For the horse's last ten races shown today, there had to be a number ( zero included) for each Beyer listed for the last 10 races. No dashes or blank spots are acceptable.

4) There can be no maiden races for the horse's last 10 races.

5) The horse's most recent race must have a Beyer that is the lowest for the horse's last ten races shown today and there can be no tie among two lowest beyers (ie, the lowest beyer must be unique)

That's the system. If more than one horse qualifies, it's OK to play more than one horse in the same race.

Since the test of the system across 500 horses yielded a win percentage of 15%, that roughy 1 out of every six horses or so, I would look for odds at 5/1 or better. But note, I did not use an odds filter in my test of the system.

I strongly suspect that if you follow these rules exactly, you will win around 15% of your bets and have a positive ROI of 14 cents on the dollar. Expect about one play perday per track.

[RXB]

Bill, it's an interesting idea, but my concern is that your study did not replicate the conditions that a handicapper faces in his daily challenge (going through today's races).

If I go back into a horse's last 10 PP lines and find its lowest Beyer, I don't know that this low number would've been the lowest number in the PP's on that very day that the horse actually ran his 'post-low' race, because there might have been an even lower number in the races that have since disappeared from the horse's PP lines. This would especially apply to younger horses.

Not to say that there isn't some validity in the theory; horses, like people, typically return toward the mean and thus will frequently follow a bad performance with an improved one. But any study would have to be carried out in such a manner that it would be applicable to someone handicapping today's cards, not someone going back through old PP's. Otherwise, the results will not be reflective of reality.

[cj]

rxb,

I was wondering the same thing. You would need lifetime PPs, ala Formulator, to accurately test this theory.

[Bill Cullen]

Originally posted by RXB
Bill, it's an interesting idea, but my concern is that your study did not replicate the conditions that a handicapper faces in his daily challenge (going through today's races).

If I go back into a horse's last 10 PP lines and find its lowest Beyer, I don't know that this low number would've been the lowest number in the PP's on that very day that the horse actually ran his 'post-low' race, because there might have been an even lower number in the races that have since disappeared from the horse's PP lines. This would especially apply to younger horses.

Not to say that there isn't some validity in the theory; horses, like people, typically return toward the mean and thus will frequently follow a bad performance with an improved one. But any study would have to be carried out in such a manner that it would be applicable to someone handicapping today's cards, not someone going back through old PP's. Otherwise, the results will not be reflective of reality.



I generally agree with your above statements. I have tested this system in real life with real money and it seems to work. I say "seems" because my real life test has been ad hoc and spread over time and I havn't keep records but my sense is that after playing a hundred horses or so, I am still profitable.

Forget about why the system should work; I still submit based on the extraordinary consistency of the 5 original samples I tested plus my own informal experience to date playing the system in real life, that there's a good chance that this system is the real McCoy! I offer it as a hypothesis that has had some excellent verification to date but which could use some additional testing against new samples with an improvement in the experimental methodology/research design. The original research design was a compromise: I didn't want to buy a hundred racing forms so I used 500 horses' intra-10 race past performances from open claiming races for both isolating the lowest beyer plus using the the very next race in a horse's 10 races for the 'results race.' Some of the filter's i used, including picking horses only appearing in open claiming company, no maiden races in their last ten races, no dashes or gaps where a beyer number should be, no ties, etc, were intended to mitigate against the less-than-perfect research design.

I personally think that the system works not so much because of a bounce off a lowest-beyer-last out (the beyer being the lowest of the horse's last ten visible races in the racing form on the page), but because people disregard the horse's chances greater than even dismal appearing situation warrants.

Hope that helps to clarify my position.

Bill Cullen

[RXB]

Exactly, cj. Not only is there the lowest fig problem, but some of his other filters (minimum 10 races, no maiden races, not in a conditioned claimer) almost certainly would have thrown many of those 'qualified' horses out if they had been applied on that very race day.

[Bill Cullen]

Originally posted by RXB
Exactly, cj. Not only is there the lowest fig problem, but some of his other filters (minimum 10 races, no maiden races, not in a conditioned claimer) almost certainly would have thrown many of those 'qualified' horses out if they had been applied on that very race day.

Yes, it would have thrown out many of those qualified horses but it also wouldn't have thrown out a lot of horses!

THE MOST COMPELLING EVIDENCE FOR THE SYSTEM IS THE EXTRAORDIANRY CONSISTENCY OF THE STATISTICS ACROSS THE FIVE INDIVIDUAL SAMPLES OF ONE HUNDRED HORSES EACH.

EACH SAMPLE WAS PROFITABLE INDIVIDUALLY (THE RANGE WAS SOMETHING LIKE 8 TO 19 % ROI).

EACH SAMPLE'S MEDIAN PAYOFF WAS TEN DOLLARS AND CHANGE.

NO ONE SAMPLE DEPENDED ON ONE HORSE FOR PROFITABILITY.

When I was in graduate school doing psychology research, preliminary findings such as the above would generated trememdous excitement among researchers and students and would have unleashed a torrent of additional follow-up research.

Enough said.

[cj]

Bill,

Don't take my post as criticism, I was having trouble understanding exactly what you were saying. It certainly does generate thinking and further analysis, keep it up.

Believe me, if people here thought you were full of it, they would have told you by now! I think the posters were trying to get a clearer picture. I know I was.

Craig

[Bill Cullen]

Originally posted by cjmilkowski
Bill,

Don't take my post as criticism, I was having trouble understanding exactly what you were saying. It certainly does generate thinking and further analysis, keep it up.

Believe me, if people here thought you were full of it, they would have told you by now! I think the posters were trying to get a clearer picture. I know I was.

Craig

If my reply sounded defensive, i did not intend that. No, I took yours and other's statements as an indication that I had been less than articulate and clear in my explanations. I was just trying to clarify what I thought were the strong points for the system being worthy of continued testing.

I'm extremely grateful for yours and the other's feedback.

Many thanks,

Bill Cullen

[andicap]

I want to know what its like going through life with the same name as a famous game show host.

:D

[Bill Cullen]

Originally posted by andicap
I want to know what its like going through life with the same name as a famous game show host.

:D

Years ago, folks knew who Bill Cullen was, but with the passage of time, I don't get much name recognition anymore.

[RXB]

As I said previously, there is likely some validity in your theory, Bill, and it is worthy of study. All we're saying is that in order to be accurately tested, you would have to set up the study from a database that allowed you to construct 10-race PP blocks where the most recent of those ten races contained the low Beyer and the horse passed your other filters based on his racing record and PP lines following that race.

In a game with a 15-20% takeout, it's best to be skeptical regarding systematic profitability until verified under rigorous conditions.

[Bill Cullen]

Originally posted by RXB
As I said previously, there is likely some validity in your theory, Bill, and it is worthy of study. All we're saying is that in order to be accurately tested, you would have to set up the study from a database that allowed you to construct 10-race PP blocks where the most recent of those ten races contained the low Beyer and the horse passed your other filters based on his racing record and PP lines following that race.

In a game with a 15-20% takeout, it's best to be skeptical regarding systematic profitability until verified under rigorous conditions.

I agree. I've done it for a hundred plays or so for real and my sense was that the winning percentage was about 15% and the profit was 5 to 10 %.

I'm all for rigorous emprical testing. I posted the sytem should anyone want to do that. I would do it myself if I thougt there was software availble and past performance data available at a reasonable price that didn't involve an inordinate amount of programming or doing SQL queries.

Bill Cullen

[RXB]

I'm sure someone has the capabilities and could run the study.

My intuitive thought, for whatever that's worth, is that if it falters it will likely be because the 15% win figure doesn't hold up. I would expect the median win price to maintain at the level you've indicated.

The other thing is, I'm just so leery of anything that doesn't produce at least 20% wins. Once you head below that level, two things happen:

1. You need a huge sample to have legitimate results; and

2. The long losing streaks that will inevitably occur can wreak psychological havoc.

But I guess if The Price Is Right...

[Secretariat]

Originally posted by Bill Cullen
I re-phrase the rules so they appear as a system normally would:

1) Today's race is an open non-maiden claimer. The definition I'm using here for an open claimer is a race where there were no restrictions on past performance. Restrictions on gender, age, state of breeding, etc, were all permitted. Optional claimers were not allowed.

2) For a horse to be a system play, it has to show at least 10 races in his past performance lines today.

3) For the horse's last ten races shown today, there had to be a number ( zero included) for each Beyer listed for the last 10 races. No dashes or blank spots are acceptable.

4) There can be no maiden races for the horse's last 10 races.

5) The horse's most recent race must have a Beyer that is the lowest for the horse's last ten races shown today and there can be no tie among two lowest beyers (ie, the lowest beyer must be unique)

That's the system. If more than one horse qualifies, it's OK to play more than one horse in the same race.

Since the test of the system across 500 horses yielded a win percentage of 15%, that roughy 1 out of every six horses or so, I would look for odds at 5/1 or better. But note, I did not use an odds filter in my test of the system.



A couple of questions. So a n2l is not allowed by Rule 1.

Rule 3 - If a horse did not finish or a jockey was thrown in ANY race it is not included. If a horse had a zero Beyer in any previous race he could NEVER qualify because of Rule 5 based on low ties right? So a zero Beyer would effectively eliminate a horse unless it occurred ONLY in the last race right?

Rule 4 - No maiden races at all. Why no Maiden Special Weight Races?


Race 5 - So if a horse had a low Beyer 5 races ago of 50 and matches that last race of 50 ,he is disqualified right?

Again thanks for the thoughts.

[Bill Cullen]

Originally posted by Secretariat
A couple of questions. So a n2l is not allowed by Rule 1.

Rule 3 - If a horse did not finish or a jockey was thrown in ANY race it is not included. If a horse had a zero Beyer in any previous race he could NEVER qualify because of Rule 5 based on low ties right? So a zero Beyer would effectively eliminate a horse unless it occurred ONLY in the last race right?



Rule 4 - No maiden races at all. Why no Maiden Special Weight Races?


Race 5 - So if a horse had a low Beyer 5 races ago of 50 and matches that last race of 50 ,he is disqualified right?

Again thanks for the thoughts. Number 3 above: Yes,That's Correct.

Number 4 above: I wanted only horses from open non-maiden claimers in my sample. A lot of claimers don't have the necessary full ten previous races in their record.

Number 5 above: Yes,That's Correct

[Bill Cullen]

Originally posted by Bill Cullen
Number 3 above: Yes,That's Correct.

Number 4 above: I wanted only horses from open non-maiden claimers in my sample. A lot of claimers don't have the necessary full ten previous races in their record.

Number 5 above: Yes,That's Correct



Number 4 above should read: I wanted only horses from open non-maiden claimers in my sample. A lot of maidens don't have the necessary full ten previous races in their record.