Using e-ponies factors as handicapping tool

[mickey_arnold]

Been using a manipulation of the free e-ponies data using MS Excel to see what I can see. Very, very preliminary results looks interesting, especially in the 2nd through 4th finishers. Looks like an improvement over using the e-ponies data right out of the can, as far as selection of actual race finish, when comparing my results to the findings of a very old study published by Liam, the creator of e-ponies. Haven't yet explored spot plays

Profitability, in the straight bet spot is another matter...But exotics results are starting to catch my eye...Always fun to see long odds contenders finishing in the 2nd to 4th spots...Large fields (12-14 horses) and many scratches, tend to make predictability a difficult proposition, but there is always the "Pass" option.

Much too early and too limited a sample to talk about here. Still tweaking it anyway, because, since retirement, much of my intermediate level analytical, statistical and Excel skills have eroded.

Briefly, it takes the e-ponies data, normalizes it, re-weights it, creates a Power rating, determines the contender threshold (that was the toughest part of the design) and creates an odds line, all while using existing MS Excel functions.

It's been fun and somewhat frustrating (but not unbearably so)... Would have loved to use database analysis and report creation, but don't have the patience anymore and my programming abilities were minimal anyway...So, there is a small degree of manual transfer of data...

When I am done tweaking and analyzing a respectable number of races, I would be happy to share the results, especially, compared to the original data with no manipulation. I would love to see myself just put into the range of profitability with rebates.

If I run into any further technical issues using MS Excel, I won't hesitate to post them here.

I'm satisfied with the conceptual methodology developed... Not ashamed to admit that I borrowed heavily from existing computerized handicapping approaches, in deciding what e-ponies race factors to use plus how and when to employ them.

Thank you to all for the diverse knowledge, skills and perspective shared on this website that have both inspired and assisted me in this lengthy project.

[mickey_arnold]

Forgot to note that I used published statistical studies that examine race factors, to analyze their degree of predictability and profitability, so I can choose when and how to use those employed in e-ponies ratings.

[pandy]

So I assume you've changed the point structure? I've always thought that things like the points for the morning line dragged the ratings down, in a sense that if you improve a horse's ranking just because it is the morning line favorite, you are going to pick the favorite too often and you can't win betting favorites. Any sort of computer system that tries to be profitable should not be attempting to pick the most obvious horses. I would think that the odds line points are counterproductive.

[mickey_arnold]

Originally Posted by pandy

So I assume you've changed the point structure? I've always thought that things like the points for the morning line dragged the ratings down, in a sense that if you improve a horse's ranking just because it is the morning line favorite, you are going to pick the favorite too often and you can't win betting favorites. Any sort of computer system that tries to be profitable should not be attempting to pick the most obvious horses. I would think that the odds line points are counterproductive.

Pandy,

Thanks for your comments. You are right on target about the counter-productivity of using Ml odds...I had additional reasons to ditch them from my re-configuring of the e-ponies selection process.

I ditched the point structure immediately for five reasons:

1) the variability of competence in ML makers from track( or circuit) to track (or circuit),

2) in checking out e-p's own line vs. the official ML line, there can be wide discrepancies and I didn't think that Liam uses his own line to formulate the points for this factor, (if I am correct about the latter, it seems to denote that he doesn't trust the ML and thinks his line is superior).
If I am wrong about the reason for Liam's odds line and its absence from the derivation of total e-p points, the next comment about the use a projected public odds line (or a calculation of a pre-scratch odds line) is still not invalidated,

3) if it is used as a tie-breaker in some type of races (as Liam has asserted in his explanation of the meaning of each race factor that is on his site) and it appears to be a significant one in others (and it is), I think point #1 above has even more validity,

4) in my analysis of the e-p figures, I found that in most race situations, eliminating it from Liam's formula, seemed to be a relatively small factor ( as Liam"s explanation of its use would infer) in the long-term accuracy of predicting overall group in-the money(ITM) or lower results.

As for selecting the winner of a race, if my re-working of e-p's figures has improved results in ITM or lower accuracy (and profitability in exotic wagering), I am not overly concerned about improved precise winner selection.

Concerning your expectation that i changed the point structure,I think you were referring to how the points were derived,i.e. for example what does 1 point in my methodology represent for speed, class etc. Well, if you don't assume that the entire e-p point derivation is completely garbage and not much better than selection of ITM or lower horses based on actual odds
or even random selections, then, re-calibration of the point system is less crucial.

I think Liam's old study of e-ponies accuracy shows this quite conclusively, as far as comparing the relative ranking of his selections versus the overall order of finish....Sure, his ID of the precise order of finish is not as high as some perfectionists might prefer, but the grouping of his selections by, say, which horses might comprise the top 3 or 4 or 5 horses in a field is fairly accurate..... And that is why much of my future focus will be on how I may improve the prediction and the potential wagering profitability of the GROUP of ITM or top 4 or 5 horses in a race, in race type or "spot" scenarios.

Oh, I know there are many, many similar comprehensive out-of-the box handicapping methods developed that can make the same claim, but I am only concerned here with e-ponies.

If you want to praise the results of your own methodology or someone elses that you utilize, then certainly do it.

(If you are reading all this, Liam, hats off to you !!!)

BTW, I think what has helped in my re-working of e-ponies figures is a process of double normalization and double re-weighting of race factors that does not result in redundancy in the individual ranking of contenders,
compared to not using this multi-level statistical approach.

P.S. I am not a shill for e-ponies...I'm a big fan of imagineering in methodology analyses and development.

[pandy]

What about the points for "Connections?" I would think that it would be best to eliminate them.

[mickey_arnold]

Originally Posted by pandy

What about the points for "Connections?" I would think that it would be best to eliminate them.

No surprise here either, Pandy...The evidence is right before anyone's eyes...Liam says straight out that "Connections data" doesn't become useful until the first few weeks of a race meet have passed... I think it may take a bit longer. and some shorter race meets could be nearly finished by then.

Haven't tried to unravel if the use of racing circuit-based "Connections" data, instead of track meet data might have selective significance.

Again, if you simply look at the data as presented for each race , it becomes obvious that there is usually a low number in and proportion of the final rating attributable to "Connections"...

And I would imagine that the data used to develop the rating for 'Connections" is what appears in the past performances available to just anybody who cares to peruse that data, and not based upon more sophisticated profiles resulting from some serious "drilling down" to finer and finer granules.

The use of more meaningful data ultimately leads to the use of that data as angles or spot plays. And, which will not appear as such in just about any of the free stuff out there....

As you said about the ML, this is the kind of factor where a high number in e-p Connections points and a high number compared to the other factors in the e-p rating again leads to the obvious lower odds choices and a consequent obstacle to profitability.

Could there be profitability with a Connections number after a few weeks in a race meet have passed? Perhaps. But if a person is using e-ponies data exclusively, they had better be prepared for some serious correlational analysis covering a long period of time, in a variety of race type, race distance and race class scenarios, with careful attention paid to those situations where the other e-p race factors usually have a high degree of dependency influence. And the careful selection of scenario constants, if that is even possible or useful...

I feel only compassion for those novices who attempt to use e-ponies or other methodologies in an indiscriminate manner...I refer to that alternatively as the "innovative","sophisticated", "methodological" or "technological" imperative.

If it exists and it seems and claims to be better than the way you do things ( Mind you, i am not suggesting that Liam and e-ponies do that) and, especially, if you lack the ken to honestly critique its basis, its strengths and its limitations, you are very, very inclined to just kick its wheels and go for a ride, paying severe tolls along the way.

Haven't given your computer based stuff a try, but I love your insight into profitable longshot spot plays...Beats anything I might come up with.

Mickey

[Capper Al]

The connection points are better reworked.

As far as e-Ponies ratings, they were called digital chalk for a reason. Yet, they will out pick most handicappers and most morning lines. There's potential there. I ran a piece in PA on the Jimmy spreadsheet that used e-Ponies data.

Good luck

[Dave Schwartz]

Mickey,

What you have described is exactly what I have in mind for the new (and free) software I am developing. The idea would be that you get the angles/picks, etc into the program - perhaps with copy/paste, an csv import or by typing - and the system tracks it (and does a whole bunch of other stuff).


Good luck with your endeavor.


Dave

[Flysofree]

What about the E-ponies Class and Speed figures? Have you made any determinations on their use?

[mickey_arnold]

Originally Posted by Dave Schwartz

Mickey,

What you have described is exactly what I have in mind for the new (and free) software I am developing. The idea would be that you get the angles/picks, etc into the program - perhaps with copy/paste, an csv import or by typing - and the system tracks it (and does a whole bunch of other stuff).


Good luck with your endeavor.


Dave

Originally Posted by Dave Schwartz

Mickey,

What you have described is exactly what I have in mind for the new (and free) software I am developing. The idea would be that you get the angles/picks, etc into the program - perhaps with copy/paste, an csv import or by typing - and the system tracks it (and does a whole bunch of other stuff).


Good luck with your endeavor.


Dave

Dave,

Tried to use CSV format for import into a data base using comma delimiter but got stuck on the race descriptor field, because it was a pain to standardize and appeared on the same line as the race factor headings.. so I simply "fixed" the race factor fields,separately loaded the race descriptor and pasted the data into the race lines, which, of course, contained many more headings for the algorithm output.

Should have used a DB program, where I was more comfortable attempting command statements..But once I got rolling on the spreadsheet, why stop?

Will confess that it took a lot of trial and error to first paste the E-P selections data into one spreadsheet with no format changes involved and, then, pasting them correctly into the computation worksheet. As a manual process, it worked more effectively for me, than pasting directly into the worksheet. Had more confidence that I wouldn't foul up the the integrity of the format and algorithm of the "fixed fields by mis-loading the data. Once I am satisfied with the final tinkering, I will password protect the appropriate fields.

If anyone wants to know how I did the manual cut and paste process for e-ponies selection data using a spreadsheet, I'll be glad to share...Isn't too hard to do and I saved everyone else the trouble of trial and error.

Since I'm the only one using the spreadsheet design at this point, I can hold off on automatically importing the computed data for each race and the race results into a database to develop analyses and, for the time being, do it manually.

With the limited number of fields in e-ponies, I think the matrix of variables to compute race results will be somewhat finite, especially if I standardize and/or limit the detail of the race descriptors shown in e-ponies selections.

I'd love to know how many handicappers use e-ponies data to confirm or supplement their own approach and how they utilize it. Don't know if a survey has been done on the e-ponies forum.

Dave, your data tracking feature seems very interesting in the context of the software model you are developing... great way to help those with neither the inclination, time or ability to personally track what they do.

For those in love with the selections of a single public handicapper or a tailored consensus of public handicappers, this will make life easier in finding the sweet spots where they can ID the plays and structure the wagers.

If you need test data you know where you can find some. Probably no shortage of folks here willing to cooperate in acceptance testing, either.

I spent loads of time at work doing data warehouse and database acceptance testing for the data file content and reports intended to enhance my job as a forensic data analyst. I think the first word out of my mouth after birth was GIGO.

Ciao,
Mickey

[mickey_arnold]

Originally Posted by Flysofree

What about the E-ponies Class and Speed figures? Have you made any determinations on their use?

Originally Posted by Flysofree

What about the E-ponies Class and Speed figures? Have you made any determinations on their use?

I hope I understood your question.

Speed and Class are integral...Cut them out and you literally won't have anything left to use. Remember we are using only six fields of data for computing a rating for each horse ( seven if you wish to include the e-p line as opposed to the M.L.). Besides, you'd be hard pressed to find any composite factor handicapping method that excludes either or both speed and class or pace and class.

If you meant how to use them... The detail is in 1) the actual devilish decision of when and how to modify a weighting scheme and, perhaps, even more importantly, 2) the conceptual approach used to construct a
weighting factor.

And that doesn't even touch the question of when the actual e-p data looks "funny", what do you do about it and how do you do it.

It required a lot of thinking and a lot of research into a variety of statistical areas and a variety of possible statistical functions to utilize.

BTW, I am a virtual novice at this compared to a couple of of the contributors on P.A. forums. They sure know how to make me humble.

Darn, I was looking for an example to illustrate difference in expertise between me and the far more knowledgeable and adept statistical afficianados that offer commentary on the PA site.

Oh well, I hope that answers your question. If not let me know.

[Capper Al]

Originally Posted by Flysofree

What about the E-ponies Class and Speed figures? Have you made any determinations on their use?

His speed and class figs aren't bad. I prefer EP's speed with Euro's. I prefer BRIS class to EP's.

[Flysofree]

Originally Posted by Capper Al

His speed and class figs aren't bad. I prefer EP's speed with Euro's. I prefer BRIS class to EP's.

Euro's? This type of race is somewhat rare from the EP races or am I missing something.

[Capper Al]

Originally Posted by Flysofree

Euro's? This type of race is somewhat rare from the EP races or am I missing something.

Breeders' Cup Day and maybe some races like the Arlington Million.

[rog9170]

where can i get this jimmys spread sheet