Analyzing your ALL-WAYS data base.

[formula_2002]

Analyzing your ALL-WAYS data base.

Find the date at the mid point of the specified track data base by going into the HANDICAPPING module and scrolling down to about the middle.

Note the date.

Now go into the ANALYSIS module and click on MODIFY to make a profile based on all races prior to the noted date.

Run the IMPACT VALUES ANALYSIS.

Create a profile.

Now, using the DB RUN in the ANALYSIS module, test the profile against the races prior to the noted date. You should generally show a very nice profit.

Now, test the same profile against all the races beyond the noted date. You should generally get a loss.

Just do that once for any one track. Do no try to “tweak” your data for better results.
Perform the analysis for each of the race categories, DS,DR,TS,TR etc. but only once for each.

If you wish, post your observations here.

Joe M

[CBedo]

First the disclaimer. I use All-Ways. It is an invaluable TOOL in my handicapping. It helps me in race selection, contender selection, and gives me the tools I need to generate my odds lines. I have never used it "black box," betting its top selections.

BUT, I thought this would be fun, so I bit. To make it even more fun, I did in on a track where my database is fairly small 650ish races, and not only that, 346 of the races came from 2003 (the last time I used All-Ways at Del Mar) until this month. So we have limited data and old data to generate profiles. To create the first profile, I segmented down to non-maiden sprints 3 & up that paid less than or equal to 15.00, using all races on or before July 31, 2006. and generated impact values to create my profile. This was a database of something like 77 races. Here's what the profile generated "fit" to the data it used to create it:

45% winners at avg mutual of 6.25 for an edge of just 40.6%

Now I reran it in the same time frame, but added back in all payoffs. New results:

38% winners at avg mutual of 7.39, edge: 40.4%

Finally, I then tested it on data it hadn't seen, all the races in August of this year. Results......

33% winners at avg mutual of 7.94, edge: 31.0%.

SO, what does this prove......the only thing I can think of right now is that it proves I should be handicapping tomorrow's races instead of playing another edition of "Statistics don't lie, Statasticians do!"

Good fun.

[CBedo]

Using the same data as used to create the first profile, but instead of eliminating the big payoffs, I only looked at payoffs of 15.00 or more to generate a "longshot" profile.

First run over the data used to create it (really small data sets now--good retro fits!), but added back all races, not just the big payoffs, and then arbitrarily set the criteria that we would be playing "any horse that is one of ALL-Ways top three choices going off at 7-1 or better," determined before looking at the results. Here's what we got:

12.1% winners at a 28.81 avg mutual, Edge: 74.3%

Now running it over the new August races:

10.3% (3/29 chances) winners at 29.33, edge: 54.1%


I completely understand the point I believe you are trying to make (from reading this and most of the other posts I could find that seemed related), and I completely agree with you that there are many pitfalls that await someone who blindly best fits past data. But that doesn't necessarily mean it's not useful in some way.

[traynor]

Analyzing your ALL-WAYS data base.

Find the date at the mid point of the specified track data base by going into the HANDICAPPING module and scrolling down to about the middle.

Note the date.

Now go into the ANALYSIS module and click on MODIFY to make a profile based on all races prior to the noted date.

Run the IMPACT VALUES ANALYSIS.

Create a profile.

Now, using the DB RUN in the ANALYSIS module, test the profile against the races prior to the noted date. You should generally show a very nice profit.

Now, test the same profile against all the races beyond the noted date. You should generally get a loss.

Just do that once for any one track. Do no try to “tweak” your data for better results.
Perform the analysis for each of the race categories, DS,DR,TS,TR etc. but only once for each.

If you wish, post your observations here.

Joe M


That is the best "quick and dirty" test method I have ever heard. You get roughly the same insights as you would from slicing the races into random groups of the same size, and comparing the results. Your method is a lot faster, easier, and just about as revealing. Thanks!
Good Luck

[traynor]

First the disclaimer. I use All-Ways. It is an invaluable TOOL in my handicapping. It helps me in race selection, contender selection, and gives me the tools I need to generate my odds lines. I have never used it "black box," betting its top selections.

BUT, I thought this would be fun, so I bit. To make it even more fun, I did in on a track where my database is fairly small 650ish races, and not only that, 346 of the races came from 2003 (the last time I used All-Ways at Del Mar) until this month. So we have limited data and old data to generate profiles. To create the first profile, I segmented down to non-maiden sprints 3 & up that paid less than or equal to 15.00, using all races on or before July 31, 2006. and generated impact values to create my profile. This was a database of something like 77 races. Here's what the profile generated "fit" to the data it used to create it:

45% winners at avg mutual of 6.25 for an edge of just 40.6%

Now I reran it in the same time frame, but added back in all payoffs. New results:

38% winners at avg mutual of 7.39, edge: 40.4%

Finally, I then tested it on data it hadn't seen, all the races in August of this year. Results......

33% winners at avg mutual of 7.94, edge: 31.0%.

SO, what does this prove......the only thing I can think of right now is that it proves I should be handicapping tomorrow's races instead of playing another edition of "Statistics don't lie, Statasticians do!"

Good fun.

I'm not sure how many people would be willing to bet on a model that started out with a clump of 77 races, or how indicative the application of that model would be. The problem with small samples (a topic hammered to death on this forum) is that they can indicate almost anything. The fact that you got a positive result in using the impact values applied to another (small) sample is not especially surprising; one or two mutuels in a month of races can seriously distort the results.

It might be interesting to re-run your latter sample with the original constraints on odds. That would diminish the effect of rare mutuels distorting the application of the model to the subsequent sample. It is curious that the very issue at the heart of such models--aberrant mutuels distorting the results--was "corrected" in the original sample, but not in later application of that model. That implies that the distortion of large mutuels in the latter sample was necessary to get the positive results (whether that was the case or not). Else why not constrain to mutuels $15 or less in the latter samples?
Good Luck

[Maxspa]

All,
I checked out the DDSS Grading Ratings website and found it to be professionally constructed with plenty of information regarding their product.
They have a free sign up for 3 days of their choices. Bettors have different plans to choose from each based on monthly service. The service is not inexpensive but if their program is successful one of their plans would fit most handicapper's needs.
I signed up for three days of their free selections and will report back on the quality of the DDSS Rating System.
As always, I do not represent this company and do not suggest anyone purchase their product!
Maxspa

[traynor]

All,
I checked out the DDSS Grading Ratings website and found it to be professionally constructed with plenty of information regarding their product.
They have a free sign up for 3 days of their choices. Bettors have different plans to choose from each based on monthly service. The service is not inexpensive but if their program is successful one of their plans would fit most handicapper's needs.
I signed up for three days of their free selections and will report back on the quality of the DDSS Rating System.
As always, I do not represent this company and do not suggest anyone purchase their product!
Maxspa

This sounds like an ad. Because it is against the agreement with this forum to advertise, I just decided to cut off your free sample. I appreciate your kind words, but we are not really actively soliciting subscribers. Thanks anyway.
Good Luck
traynor

[Maxspa]

Traynor,
I'm very disappointed in your response! First of all I'm not a shill, but as a member of this board, I have given my honest opinion on many handicapping software offerings. I will continue to do so!
First of all, your email address on the bottom of your response IS a form of ADVERTISEMENT, I would not have contacted your website without that information.
Maxspa

[ryesteve]

This sounds like an ad. Because it is against the agreement with this forum to advertise, I just decided to cut off your free sample.
You've got to be kidding me! If you want to weasel out of something, at least come up with something that makes even the slightest bit of sense. At this point, you might as well be honest and just admit that you're afraid of someone actually reporting the results obtained via your service, since it's impossible not to draw that conclusion at this point.

[Tom]

Gotta agree herre. Max is a fixture here for his helpful suggestions, reviews, comments. I repsect him a lot, and always read what he has to say about things.
Sounds pretty shallow to me, to cut off a freebie becasue you don't want results posted, yet still advetise the site. :ThmbDown::ThmbDown::ThmbDown:

Curious, which one of the user agreement points do you say Max violated?


BTW, I played a three race model this weekend at Philly - went 4 for 6 over two days. I thought, having three races, GEE, more than enough data to start betting. :lol:

[CBedo]

It might be interesting to re-run your latter sample with the original constraints on odds. That would diminish the effect of rare mutuels distorting the application of the model to the subsequent sample. It is curious that the very issue at the heart of such models--aberrant mutuels distorting the results--was "corrected" in the original sample, but not in later application of that model. That implies that the distortion of large mutuels in the latter sample was necessary to get the positive results (whether that was the case or not). Else why not constrain to mutuels $15 or less in the latter samples?
Good Luck

I don't believe that implies that at all. I often take out some of the bigger mutuals when I create profiles, especially in smaller samples, so that one payoff doesn't skew the returns generated by the profile. But as far as testing the profile on the original data, I then add back in the larger payoffs, which usually has the effect (as seen above) of lowering the win percentage, but possibly catching a higher mutual--if you catch one of the higher priced horses.

As far as removing the higher payoffs from the new data, well, that doesn't seem to make much sense to me. If you were blindly following the profile as a system, and it picked a 10-1 horse, would you pass the bet because that odds range was excluded from the original formulation?

But just to be fair, I did rerun it, removing any payoff over $15.00 which resulted in removing 11 races from the sample. The profile did pick one of these 11 winners (Heirloom on Aug 7th) which paid 17.40. So the new numbers actually look better, not worse. Higher win%: 39%, Lower mutual: 7.21, Higher edge: 40.6%.

[shanta]

BTW, I played a three race model this weekend at Philly - went 4 for 6 over two days. I thought, having three races, GEE, more than enough data to start betting. :lol:

That's how it's done!

Shoot 3 races might be 3 too many :eek:

:lol: :lol:

[headhawg]

This sounds like an ad. Because it is against the agreement with this forum to advertise, I just decided to cut off your free sample. I appreciate your kind words, but we are not really actively soliciting subscribers. Thanks anyway.
Good Luck
traynorHmmm...let me compare what Maxspa has done for us here vs. you. There's really no comparison. I just decided to put you on my Ignore list.

Thanks anyway.

[xtb]

This sounds like an ad. Because it is against the agreement with this forum to advertise, I just decided to cut off your free sample. I appreciate your kind words, but we are not really actively soliciting subscribers. Thanks anyway.
Good Luck
traynor

wtf

[traynor]

You've got to be kidding me! If you want to weasel out of something, at least come up with something that makes even the slightest bit of sense. At this point, you might as well be honest and just admit that you're afraid of someone actually reporting the results obtained via your service, since it's impossible not to draw that conclusion at this point.

You are absolutely correct.

[traynor]

Traynor,
I'm very disappointed in your response! First of all I'm not a shill, but as a member of this board, I have given my honest opinion on many handicapping software offerings. I will continue to do so!
First of all, your email address on the bottom of your response IS a form of ADVERTISEMENT, I would not have contacted your website without that information.
Maxspa

I agree. To avoid any implication that offering free samples is an attempt to solicit "customers," I have removed the link from my signature.

[traynor]

Traynor,
I'm very disappointed in your response! First of all I'm not a shill, but as a member of this board, I have given my honest opinion on many handicapping software offerings. I will continue to do so!
First of all, your email address on the bottom of your response IS a form of ADVERTISEMENT, I would not have contacted your website without that information.
Maxspa

I apologize to you, and to anyone else who misinterpreted my signature as an attempt to advertise, or to "lure" readers away from this site. The link was posted for the convenience of readers, nothing more. Other than that, it had no value to me whatsoever.

[PaceAdvantage]

I'm getting so confused! What does this DDSS Grading Ratings have to do with the original intent of this thread? Seems like things have gone totally off topic.

And as for Maxspa and Traynor.....

Traynor, thanks for the help, but NO THANKS. Maxspa hasn't done anything to violate any terms of this board. He has proven to be a valuable resource to this board, and after meeting him in person on numerous occasions throughout the years, I can attest to the fact that he is a stand up guy, and he means what he says (not that my word carries much weight, nor should it, but you get the drift).

He's NOT affiliated with any commercial endeavor that I am aware of, thus he can NOT be breaking any of the terms of this board. Only folks like you, Traynor, who operate commercial endeavors can be guilty of breaking the terms of this board of which you accuse Maxspa of breaching.

[Hank]

The tiny bit of credibilty that you had left after the Lil Tommy debacle is now gone.:ThmbDown:

[traynor]

I'm getting so confused! What does this DDSS Grading Ratings have to do with the original intent of this thread? Seems like things have gone totally off topic.

And as for Maxspa and Traynor.....

Traynor, thanks for the help, but NO THANKS. Maxspa hasn't done anything to violate any terms of this board. He has proven to be a valuable resource to this board, and after meeting him in person on numerous occasions throughout the years, I can attest to the fact that he is a stand up guy, and he means what he says (not that my word carries much weight, nor should it, but you get the drift).

He's NOT affiliated with any commercial endeavor that I am aware of, thus he can NOT be breaking any of the terms of this board. Only folks like you, Traynor, who operate commercial endeavors can be guilty of breaking the terms of this board of which you accuse Maxspa of breaching.


I did not in any way imply that Maxspa is anything less than impeccably credible, or had done anything to violate the terms of board use. My comments were intended to preempt the erroneous conclusion that Maxspa and I were somehow collaborating to market ratings. It had nothing to do with Maxspa; it was related to the erroneous idea that my only motivation in providing free samples or posting selections is for "marketing." Which, as you point out, WOULD be a case of breaking the terms of board use.

I have no idea how you could interpret what I said as a condemnation of Maxspa. My intent was to avoid a situation in which Maxspa produced a glowing testimonial--which would then be interpreted as "marketing puffery" designed to sell ratings.

[traynor]

Traynor,
I'm very disappointed in your response! First of all I'm not a shill, but as a member of this board, I have given my honest opinion on many handicapping software offerings. I will continue to do so!
First of all, your email address on the bottom of your response IS a form of ADVERTISEMENT, I would not have contacted your website without that information.
Maxspa

Maxspa,
Let's get this cleared up before it goes any further. I did not in any way intend my comments to imply you were a shill, or anything less than perfectly honest. Since I began posting on this forum I have consistently argued against the popular--but erroneous--notion that a small sample is representative of a larger population. That issue is generally referred to by the term Amos Tverky and Daniel Kahnemann use; the "Law of Small
Numbers." It is a common decision-making fallacy that has been the topic of numerous academic research projects and studies.

That issue, along with the related issue of adjusting datasets to correct for outliers, was the main topic of a 4000+ page hit thread recently. That thread clearly indicated that the issue of small sample distortion is literally ignored by "software developers" who write handicapping applications. The reason is simple; the inclusion of anomalous outliers invariably improves the historical "ROI" figures, and excluding those anomalies invariably diminishes the perception of "profitability."

The notion of small samples replicating the distribution of larger samples--when applied to random variable events--is silly.

We offered (past tense) free samples to provide illustrations of the format used, not as indicators of potential earnings or losses. In most cases, the samples are from "rag tracks" like Charlestown, Finger Lakes, or Evangeline Downs and comprise less than 10% of the ratings for a given day. The samples are not intended to "market" ratings, any more than the occasional selections I posted on PA were intended as "marketing." The notion is absurd, given the fact that I have consistently argued against the use of small samples as indicator of long-term performance.

If you have had experience evaluating software applications, you are probably well aware of the fact that small samples can be seriously misleading, in either direction. Buying a software application because it picked the winner in two of the last three races, or "did really well in yesterday's races at Mountaineer" is almost always a prescription for serious losses. You may not have been aware that the free samples were intended primarily to illustrate format, not potential; the samples are for one track only (randomly extracted from the eight to ten tracks provided each day), and that track changes daily. There would be very little basis on which to make a meaningful evaluation of anything except the printing format, or something of that nature.

Finally, personally, it is obvious from your syntax that you are a very precise, orderly person, and fairly serious. I appreciate that, and respect it. I again apologize for my abruptness, and for any implication that you were in any way anything except completely honest and straight-forward in your comments.
Good Luck
traynor

[traynor]

The tiny bit of credibilty that you had left after the Lil Tommy debacle is now gone.:ThmbDown:

The "Lil Tommy" debacle? Oh, you mean Anderon and his hissy fits? Not to worry. He has discovered how to express his wagers in a fashion that he can simultaneously claim Big Wins while posting losses. I had a fine time watching that "$50,000 to win" post. I was kinda hoping he would lose, just so I could watch it go up to $100,000, then $200,000--whatever it took to claim a profit on paper. Oh, yeah, sorry--we were discussing "credibility." What was your point, anyway?

[PaceAdvantage]

it was related to the erroneous idea that my only motivation in providing free samples or posting selections is for "marketing." Which, as you point out, WOULD be a case of breaking the terms of board use.

If you were to post free samples or free selections on this board, I would NOT view that as a case of breaking the terms. It's been done many, many times on this board with zero problem for the vendors in question....

[traynor]

If you were to post free samples or free selections on this board, I would NOT view that as a case of breaking the terms. It's been done many, many times on this board with zero problem for the vendors in question....

Perceptions may be distorted when events are taken out of context, or when only a part of something is viewed.

When I saw your post on another thread, 8/28, you stated (in regard to Slick's comments):

[quote PA]
Freud once said "sometimes a cigar is just a cigar"

And my question to you, "Wanna buy an ad?" was just that. A solicitation to purchase advertising for your product. Your user name is your product, and your signature gives people a way to contact you about your product.

If you aren't interested in drumming up more business, why does your signature contain an email address for folks to contact you about your services?

It was a fairly innocent question which I intend to ask others with something to sell....sorry if it seems I have offended you somehow.... [end quote]

That seemed to me a fairly clear statement of intent on your part, and that you regarded the inclusion of a link signature as "marketing." It is your forum, and it is your prerogative to do so. However, when that statement was followed shortly by Maxspa's kind assessment of our website, along with the announcement that he intended to do an "evaluation" of our sample ratings, it seemed to me that it would be easy for a casual reader to conclude that it was all a marketing ploy.

As a number of forum members have pointed out, this forum is not an appropriate venue for marketing our services. The information we provide is intended for wagering professionals, which is a completely different demographic than that served by this forum. Aside from downloading the free samples we offer, there has been no significant activity from this forum to any other portion of our site that would indicate the usefulness of this venue for marketing our services.

My use of the link in my signature was as a convenience to your members, a number of whom take advantage of the fact that we post free selections for early races early, and late races late in the day. By visiting twice a day, they can get selections for two (or more tracks), every day.

Why do I feel the need to include a link, rather than just posting the selections on PA two or three times a day? Time constraints. The selections are generated in a format for uploading, and editing those selections into a text format for posting on PA is time consuming. Editing another document with the results takes still more time. On days that I am pressed for time, I might only post one track, or no tracks.

I appreciate your position, and I appreciate your statement that posting selections would not be considered marketing. Missing was the statement that a link could be used in the signature line of those postings, as long as the postings avoided marketing-type commentary. I don't mind providing selections and free samples to forum members. I DO mind that the inclusion of a link in my signature seems to have been re-defined as "marketing."

[PaceAdvantage]

That seemed to me a fairly clear statement of intent on your part, and that you regarded the inclusion of a link signature as "marketing." It is your forum, and it is your prerogative to do so.

You've been here a long time. You know that I allow folks to include a link to their site in their signature. Whether I consider it marketing or not isn't the point. The point is, it's allowed, and has always been allowed.

As a number of forum members have pointed out, this forum is not an appropriate venue for marketing our services.

Actually, it's an excellent venue for marketing your service in the RIGHT MANNER (that is, with a very affordable, paid advertisement). Case in point, the recently closed thread in the handicapping software section, where a vendor writes about the major jump in hits his website received after a thread about their software program was posted by a user. In fact, the operator of the website stated, and I quote:

wow, suddenly the whole world is going to our webpage

So you see, I can't think of a more affordable or proper venue to market your product then this website.

The information we provide is intended for wagering professionals, which is a completely different demographic than that served by this forum. Aside from downloading the free samples we offer, there has been no significant activity from this forum to any other portion of our site that would indicate the usefulness of this venue for marketing our services.

Ouch! Is that supposed to be a put down? For I all I know, your stuff sucks, and that's why nobody has gone further than the free samples....as long as we're being honest here, that's all I have to say. Any further comments on this matter should be taken to a new thread, as I will delete any further attempts to take this thread off topic.

[formula_2002]

First the disclaimer. I use All-Ways. It is an invaluable TOOL in my handicapping. It helps me in race selection, contender selection, and gives me the tools I need to generate my odds lines. I have never used it "black box," betting its top selections.

BUT, I thought this would be fun, so I bit. To make it even more fun, I did in on a track where my database is fairly small 650ish races, and not only that, 346 of the races came from 2003 (the last time I used All-Ways at Del Mar) until this month. So we have limited data and old data to generate profiles. To create the first profile, I segmented down to non-maiden sprints 3 & up that paid less than or equal to 15.00, using all races on or before July 31, 2006. and generated impact values to create my profile. This was a database of something like 77 races. Here's what the profile generated "fit" to the data it used to create it:

45% winners at avg mutual of 6.25 for an edge of just 40.6%

Now I reran it in the same time frame, but added back in all payoffs. New results:

38% winners at avg mutual of 7.39, edge: 40.4%

Finally, I then tested it on data it hadn't seen, all the races in August of this year. Results......

33% winners at avg mutual of 7.94, edge: 31.0%.

SO, what does this prove......the only thing I can think of right now is that it proves I should be handicapping tomorrow's races instead of playing another edition of "Statistics don't lie, Statasticians do!"

Good fun.
Two years,, I waited two years for a reply!!! :jump:
Come on Wilson, we are out of here

CBDO, the next 650 races will kill you..

[CBedo]

Have to wait a year to end this story. The next sample will be hard to collect until next summer as Del Mar ends Wednesday.

But I completely understand what you're saying (I think). I believe anyone who doesn't understand where the impact values & ROIs are coming from and how they were calculated is headed down a dangerous road. But I believe that like many things in life, just because they're not perfect, it doesn't mean you can use them if you comprehend their underlying strengths and weaknesses.

As far as getting me "off" ALL-Ways, I'd love to hear what you think I ought to be using and why. Email me.

I'll take all the help I can get!

CBedo

[formula_2002]

Cbedo
A question if you would.
Does always include the dollar odds and pay-offs in their export file?
It’s an improvement I had highly encouraged, but never happened.
If you are dbase familiar, I would suggest the $1.00 DRF single file format.
I just never got any value from those $7 All-Ways files, and I have, I would say about 1500 or more files.
I have been using JCapper since, about Jan’06. It has everything I need and it uses the drf files and xrd files.

[CBedo]

I'm running an old version (10), which does have the payoffs in the export file, but not the going off odds (M/L only). But I know the going off odds are in the Exotic Results data file, and ALL-Ways uses them. Otherwise it wouldn't be able to do the "what-if" profile analysis. I should have version 13 in a week or so (been out of handicapping for a while), and when I get the upgrade, I'll let you know.

[CBedo]

By the way, thanks for the recommendations. I'll check out JCapper. As I said, I'll take all the help I can get.

whoops, guess not. no more demos.

[formula_2002]

I'm running an old version (10), which does have the payoffs in the export file, but not the going off odds (M/L only). But I know the going off odds are in the Exotic Results data file, and ALL-Ways uses them. Otherwise it wouldn't be able to do the "what-if" profile analysis. I should have version 13 in a week or so (been out of handicapping for a while), and when I get the upgrade, I'll let you know.
I have 10 point something myself.
I ran many test of all 80 plus factors and found them, when taken individually, to be very similar in roi, except for the prime power rating, which is a bris figure also contained in the single file format.

Do let me know if the going off odds are in the release 13 version.
If you like to work with All-Ways data, outside of the program, inaddition to extracting data using the "export" feature, you can also extract data using a TIF program that bris makes avilable.
It's located at http://www.bris.com/cgi-bin/dynamic.cgi?page=utils. It's called INFOTRAN.
When using INFOTRAN, I think you have to change the .ALW file extension to .DRF