Another contrarian system

[Bill Cullen]

More than 10 years ago I tested this system with Jim Bayle of Sportstat. I don't recall all the stats but I remember that the winning percentage was 17% and the ROI was minus ten cents on the dollar. All races excepts jumps and hurdles are playable.
The sample that Jim Bayle tested was in the thousands of races.

Here are the rules:

1) The horse must have been in first, second or third in one of the first three call positions in its most recent race.
2) The horses odds in its last race must have been less than 4/1.
3) The horse must have finished 5th or worse in its last race.
4) The winning horse of the last race must have won by 5 or more lengths.

That's it.

[NoDayJob]

Originally posted by Bill Cullen
I don't recall all the stats but I remember that the winning percentage was 17% and the ROI was minus ten cents on the dollar.

I have a fairly complete library of losing systems. This one be added. Thank you.

NDJ

[andicap]

Hey,not that I'll be using it, but a purely mechanical system that gets back. 90 ROI ain't the worst thing I've ever seen.

These things have some value -- at least in giving newcomers to the sport something simple to have fun with and have a chance of winning some money. 90 cent return isn't good, but it sure as hell beats the lottery, keno, and some craps bets.

Bill, these things are always fun. Keep on posting!!

one of my favorite books when I was starting out was one that listed 300 angles and systems. Of course none of them were winners, but it was great fun to read through it. And they might give you an idea for an angle that you can tweak to help you.

:)

[NoDayJob]

Originally posted by andicap
Hey,not that I'll be using it, but a purely mechanical system that gets back. 90 ROI ain't the worst thing I've ever seen.
:)

Why not just bet the favorite in every race? Less work, approximately the same return. Eliminate the false favorites and viola' you're an instant winner.

NDJ

[Tuffmug]

Instead of playing these systems or trying to tweak them to profitability, I suggest you focus on analyzing what these data sets tell you about trainer intent and moves used by the trainer to get to that next race where he wants to score.

To me, your first system "works" because trainers are darkening form in last race to obtain a bad Beyer and a better price next out.

Your second system "works" because this showing of early speed in the last race is an indication of improving form that may portend an improved race next out and may have been nothing but a workout race used by the trainer to prep him for today's race.

[Bill Cullen]

Originally posted by Tuffmug
Instead of playing these systems or trying to tweak them to profitability, I suggest you focus on analyzing what these data sets tell you about trainer intent and moves used by the trainer to get to that next race where he wants to score.

To me, your first system "works" because trainers are darkening form in last race to obtain a bad Beyer and a better price next out.

Your second system "works" because this showing of early speed in the last race is an indication of improving form that may portend an improved race next out and may have been nothing but a workout race used by the trainer to prep him for today's race.

Some good observations and food for thought.

Thanks,

Bill Cullen

[Secretariat]

Bill,

The system may have merit with some additional tweaking, AND may be profitable at some specific tracks or race types or distances. Mechanical systems are tough, and whenever handicappers tell you to forget something, look again cause it means they're looking the other way.

A 10 cent loss is pretty good. An Odds filter on that might push up that ROI a bit or demanding a decent jock or trainer or a recency range. Some people scoff at that stuff. Trust your own records and don't get hung up on what others here think of angle investigation.

I'll check out your system on some of my data tomorrow and see how they hold up on a couple of sample tracks.

[Bill Cullen]

Originally posted by Secretariat
Bill,

The system may have merit with some additional tweaking, AND may be profitable at some specific tracks or race types or distances. Mechanical systems are tough, and whenever handicappers tell you to forget something, look again cause it means they're looking the other way.

A 10 cent loss is pretty good. An Odds filter on that might push up that ROI a bit or demanding a decent jock or trainer or a recency range. Some people scoff at that stuff. Trust your own records and don't get hung up on what others here think of angle investigation.

I'll check out your system on some of my data tomorrow and see how they hold up on a couple of sample tracks.

I'll be curious about what results you pull up.

Bill Cullen

[kenwoodallpromos]

Sounds like you are looking for a bad trip.

[Bill Cullen]

Originally posted by kenwoodallpromos
Sounds like you are looking for a bad trip.

The rationale behind the system was to have a legitimate contender (odds < 4/1) 'contending' at some point in the race (1st, 2nd or 3rd at the first, second or third call position) and then a run-away winner blows by the opposition in the stretch winning by 5 lengths or more and the qualifying horse finishes 5th or worse. Presumably the jock on the qualifying horse is less-than-motivated to push his horse hard with the winnner opening up daylight in the stretch or maybe the winning horse had a huge lead all the way.

Although the system loses money, it only loses 10 cents on the dollar and the winning percentage was about 17%. I don't recall the impact value but it was significant. Like I said Jim Bayle of Sportstat validated the system across thousands of races.

[John]

Bill . How ofen can a play come up. If you are looking for the winner of your rules to have won by 5 or more lengths.

[Bill Cullen]

Originally posted by rocajack
Bill . How ofen can a play come up. If you are looking for the winner of your rules to have won by 5 or more lengths.

If memory serves, a play showed up every couple of days per track.

[Secretariat]

Bill,

I ran both your systems against Belmont 2001-2002-and 2003 total was 267 cards, and Delaware 2002 and 2003.

Here are the results. First Belmont.

Old Bayle Method

28 plays out of 267 cards or 0.1 plays per day, or a play every 9th or 10th day.

28 attempts
2 wins
7.14% Win
$15.40 Money won
0.28 ROI or a 72 cent on the dollar loss.

Low Beyer Method

384 plays out of 267 cards or 1.44 plays per day.

384 attempts
41 wins
10.68% Win
$616.50 money won
0.80 ROI or a 20 cent on the dollar loss

Now Delaware

Old Bayle Method

49 plays out of 271 cards or 0.18 plays per day or a play every 5th day.

49 attempts
12 wins
24.49% win
$105.20
1.07 ROI or a 7 cent profit on the dollar

Low Beyer Method

594 plays out of 271 cards or 2.19 plays per day.

594 attempts
54 wins
9.09% win
$643.00
0.54 ROI or a 46 cent loss on the dollar.

Now it's possible I programmed something wrong, but the Low Beyer isn't showing much at these tracks except perhaps as an eliminator. The Old Bayle Method needs more research. A more extended workout is needed and a review of the Low Beyer rating, but I manually checked ten samples and they do seem to comply with your stated rules.

[hurrikane]

49 plays in 3 yrs. I'd fall asleep waiting for the next bet.

And then with my luck I'd probably be in the head taking a leak when the bet came up.

[Bill Cullen]

Originally posted by Secretariat
Bill,

I ran both your systems against Belmont 2001-2002-and 2003 total was 267 cards, and Delaware 2002 and 2003.

Here are the results. First Belmont.

Old Bayle Method

28 plays out of 267 cards or 0.1 plays per day, or a play every 9th or 10th day.

28 attempts
2 wins
7.14% Win
$15.40 Money won
0.28 ROI or a 72 cent on the dollar loss.

Low Beyer Method

384 plays out of 267 cards or 1.44 plays per day.

384 attempts
41 wins
10.68% Win
$616.50 money won
0.80 ROI or a 20 cent on the dollar loss

Now Delaware

Old Bayle Method

49 plays out of 271 cards or 0.18 plays per day or a play every 5th day.

49 attempts
12 wins
24.49% win
$105.20
1.07 ROI or a 7 cent profit on the dollar

Low Beyer Method

594 plays out of 271 cards or 2.19 plays per day.

594 attempts
54 wins
9.09% win
$643.00
0.54 ROI or a 46 cent loss on the dollar.

Now it's possible I programmed something wrong, but the Low Beyer isn't showing much at these tracks except perhaps as an eliminator. The Old Bayle Method needs more research. A more extended workout is needed and a review of the Low Beyer rating, but I manually checked ten samples and they do seem to comply with your stated rules.

Wow! Excellent research! Obviously my more informal testing was way off the mark as regards the low beyer method! It did indicate some initial promise (yes, I was aware of the methodological shortcomings in the research design) but I stand empirically corrected and rightly so. Sorry to have wasted anyone's time on this board.

Regarding the Bayle Method, yes I agree: more research and a larger sample of actual plays is needed.

Many thanks to you, Secretariat!

[Secretariat]

Bill,

No problem. Personally I thought it might provide some potential, but there were long losing streaks in the Low Beyer method. THe Old Bayle Method was profitable after two years at DEL. I don't care how many plays. Any thing in the arsenal is worth checking. I remember checking Mark Cramer's camoflauged win and there are very, very frew plays, but no one's complaining about his ideas.

Keep them coing Bill. It shows a willingness to participate and a thinking about what the other players might not be doing. Just curious what your initial tracks were for your initial test. BEL and DEL don't mean LBM doesn't work better elsewhere.

[Bill Cullen]

Originally posted by Secretariat
Bill,

No problem. Personally I thought it might provide some potential, but there were long losing streaks in the Low Beyer method. THe Old Bayle Method was profitable after two years at DEL. I don't care how many plays. Any thing in the arsenal is worth checking. I remember checking Mark Cramer's camoflauged win and there are very, very frew plays, but no one's complaining about his ideas.

Keep them coing Bill. It shows a willingness to participate and a thinking about what the other players might not be doing. Just curious what your initial tracks were for your initial test. BEL and DEL don't mean LBM doesn't work better elsewhere.

Thanks for your encouragement. Because I'm more of a poet than a statistician, I tend to come up with a fair number of right-brained hypotheses that tend to be very counter-intuitive but I had training in statistics and experimental design in grad school and in my work (software engineering) so I believe in the absolute need for empirical validation.

A question: how does one obtain the data and software to do their own system testing in an automated fashion (I have programming and SQL skills although a little rusty now since I'm more of a manager now)?

Many thanks, Secretariat.

Bill cullen

[Secretariat]

You can import data into MS Access, or learn a little Visual Basic to facilitate the process.

[hurrikane]

Ok, I was only joking around. Sorry if it sounded negative. You guys need to lighten up.

But seriously. 49 races is not enough of a sample to make a betting decision. You would have to wait anouther 3 years to get enough plays to MAYBE say you had a profitable play.

There just are not enough plays to make an itelligent decision about this play.

MHO

[Mike at A+]

I am looking for a new programming project to keep me out of trouble and I came across this folder. Looking at the above 4 step "system", I can easily see that programming something like that would be a piece of cake given the complex stuff I've done with A+ Thoroughbred (i.e. regression analysis to come up with impact values, extraction of hundreds of pieces of data from BRIS and Trackmaster PP files, utilizing the extracted data and impact values to come up with probabilities and ultimately an odds line, etc etc). Programming a 4 step system like the one mentioned above is nothing when compared to the thousands of hours I've put into A+ Thoroughbred since 1989 so I have this "proposition" for anyone who can provide some input. I'd like to write a program that produces selections for a large number of these simple "angle systems", say at least 15 or 20 of them.

Here's the "proposition". For those providing input for one or more angle systems that make it into the final product, I will email the executable code along with installation and usage instructions. I will also offer the finished product for sale to the general public at a modest price to be determined. Barring any unforseen circumstances, I should be able to produce something pretty quickly (but no guarantee on a target date).

To be acceptable, all systems must be fairly simple to describe and should use data contained in both the BRIS (single file DRF) and Trackmaster (Type "C" files). You can easily obtain free file layouts for both of these files but to get an idea of what I am currently extracting for A+ Thoroughbred, you can take a look at the View/Edit screen shot on my website. Go to http://users.rcn.com/aplus/indext.htm and find the link to "View 7.0 screen shots".

Please send any suggestions to me via email and not in this forum. Thanks. Address is "aplus@rcn.com".