How will my spot play run today?

[Carl]

(With apologies to William Scott).

I am a member of the mechanical school of though on spot plays, I don't wanna have to handicap at all. Think I too am like Fred4now, don't know how and too lazy to learn.

So how to make as sure as possible that I am finding plays in rummaging through past data that have future value becomes the question. I have in the past been
1) Using pristine data to confirm.
2) Checking for a "normal" ratio of 2nd's and 3rd's as well as winners with the potential spot play.

Wouldn't mind a third way.

Well, I now have a new way, thanks to Tomcat's and Mike's posts on my previous spot play thread. I'll put the idea out there to see if anybody has any comments, i.e. "been there, done that" or "that's a meaningless approach" or "whatta genius" or........

Anyway, here is what I have come up with.
1) Win percentage is as important as ROI. They are inseparable.
2) "Average off odds" for ALL horses the play picks is an important number to know about a spot play.
3) The ratio of the "average off odds of the winners your play picked"/"average off odds of all horses picked by play" tells you whether or not your play has been lucky with the winners it has been getting.
For example, in Mike's post about his 20% spot play that showed an average pay of $9 at RD and an average pay of $13 at MTR (the difference between being profitable or no), I would love to know the average off odds of all the horses selected. IF the RD average off odds were 50% less than the MTR off odds, my hypothesis is he has value. If the average off odds are the same, that MTR is "lucky".
4) Looks like my longshot plays (average pay around $20) have a 1.5 ratio of "winners off odds"/"horses picked by spot play off odds".
In other words, IF my average winning horse in a particular spot play went off at an average price of 10-1, I am finding that all horses picked by that spot play are going off at an average of 15-1
5) I think a ratio less than this is "bad". For instance, I have one spot play that has an average off odds of all horses of 11.55, and an average pay for my winners of 11-1. I am chucking it, I think it has been "lucky".

So my questions for fellow posters lurkers are
1) What do you think? Meaningless number this ratio or valid?
2) Anybody look at similiar?
3) What other confirmation factors do you use to check a spot play's potential validity?

later,
Carl

[Glen]

Carl,

You bring up a good point. Dunno the answer but along the same lines I kinda like Gordon Pine's A/E ratio. I'm not smart enough to figure out if it is a good measurement but on the surface it seems to make some senes. BTW, I am thinking about adding A/E ratio (if I could figure it out) to the ROI report if it is valid.

Following taken from G. Pine web site:
Track Tracts

Impact Value of Questionable Value:
A/E is the Gold Standard
by Gordon Pine

I’ve still got my copy of Winning at the Races by William L. Quirin, Ph.D. Picked it up as a
new book in 1979 at Smith News in Pershing Square in downtown Los Angeles. At the time,
Smith News was the only place in town that catered to horse-playing degenerates, so it was
a Mecca for me. I could stop there during my 2 * hour-each-way, two-transfer bus ride that
took me from West L.A. to Arcadia, where sat Santa Anita, the beloved home of all
horse-playing degenerates.

Winning at the Races was and is a classic in the field. Quirin looked at many aspects of
handicapping through a statistical lens, and he used (gasp!) a computer to do it. Well before
the time of ubiquitous personal computers and databases galore, Quirin, a professor of
mathematics and computer science, was on the job. And a fine job he did.

One of the metrics that Quirin developed to analyze handicapping factors he called "Impact
Value" or "IV." As he put it, "IV statistics are calculated by dividing the percentage of
winners with a given characteristic by the percentage of starters with that characteristic."
According to Quirin, "An IV of 1.00 means that horses with a specific characteristic have won
no more and no less than their fair share of the races." An IV of 2.00 meant these horses won
twice as much as expected; an IV of .50 meant they won half as much as expected.

So, Impact Value is a ratio: the win percentage of a group of horse divided by the percentage
of starters that belong to that group. For instance, let’s say you’re looking at the group of
horses that won their last start. Their win percentage is 17.0% and their percentage of
starters is 11.2%:

Impact Value = win% / % of starter
Impact Value = 17.0 / 11.2
Impact Value = 1.52

For last-out winners, you end up with an Impact Value of 1.52. In other words, the IV says
that these horses win 52% more than they should.

But do they really? Trouble is, there’s a flaw in this metric. What if you knew that horses
that won their last start went off at odds lower than average? In fact, they do. That would
mean that the horses in this group might be winning a lot because of their low odds. Their
winning the last race might not have anything to do with it. In fact, it’s possible that they
are winning less than they should given their odds.

And that’s the key: given their odds. The denominator in this ratio needs to factor in the odds
of the horses in the group being looked at. That’s why the better way to measure the impact
of a certain factor is to divide the actual win percentage of horses with a given characteristic
by the expected win percentage of that group, given their odds. I call this the A/E Ratio – the
Actual Win% divided by the Expected Win%. It’s a much better metric than the old IV
measurement. (I don’t mean to dis Quirin - he was apparently aware of the limitations of IV,
since he always combined its use with a couple of other metrics which he called $NET and
EW [Expected Number of Winners].)

Anyway, taking our example of the group of horses that won their last start, their win
percentage was 17.0%. Let’s say their average odds was 4.0 to 1. We first need to calculate
their expected win percentage. The calculation is:

Expected Win% = (1 / (Odds-to-1 + 1)) * (1 - Track Take)

If we imperfectly estimate the track take as 18 cents on the dollar or .18, then the calculation
goes like this:

Expected Win% = (1 / (Odds-to-1 + 1)) * (1 - Track Take)
Expected Win% = (1 / (4.0 + 1)) * (1 - .18)
Expected Win% = .2 * .82
Expected Win% = .164

We expect horses with average odds of 4.0-to-1 to win about 16.4% of the time. Plugging this
into our A/E calculation, we get:

A/E Ratio = Actual Win% / Expected Win%
A/E Ratio = 17.0 / 16.4
A/E Ratio = 1.03

So, given these numbers, the true impact of a horse winning the last race is 1.03, not 1.52.
Winning last out means a horse will win only 3% more than it should, not 52% more, as the
IV indicates.

In the time that I’ve used the A/E ratio, the only drawback I’ve seen is that it can be
misleading with small samples. Say you have 10 horses being studied: half of them went off
at 2/1 and half went off at 20/1. Their average odds would be 11/1. If two of the 2/1 horses
won, you’d have an A/E ratio of 20% / 6.8% or 2.94. This is obviously misleading. This
problem disappears with a larger sample because you will tend to have a mix of odds levels,
not just a bunch of low odds horses and another bunch of longshots.

"The A/E ratio is the single best measurement of the power of a particular
handicapping factor."

This may all seem pretty esoteric, but it’s my contention that the A/E ratio is the single best
measurement of the power of a particular handicapping factor. It truly tells you the impact
that a factor has, and removes the confounding influence of the odds of the group of horses
being studied. It avoids the drawbacks of Return On Investment calculations, which can be
easily thrown off-kilter by a couple of longshot winners. A/E is not something that you can
readily calculate in your head, but it is the gold standard of the art and science of
thoroughbred predictions. NC

[MikeDee]

well guys I'm going to have to re-read these 2 or 3 times to figure out what hell you said:)

Carl I'll take a look at you average off odds thing, I should be able to figure it and get back to you with some results.

Glen I'll see if I can run some numbers on your A/E thing as well

Carl I couldn't help but notice the Glen is a "active member" while you and I are just "members". (probably all those damm polls he puts up) :D

[MtKen]

Good stuff Carl & Glen.

What I like to do is somewhat simple. I routinely break my plays down into odds segments.
<2/1, 2/1 to 7/2, 4/1 to 6/1, 7/1 to 9/1, 10/1 to 19/1, 20/1 to 40/1, 40/1+.

If the play is legitimate I think there should be a logical progression down in win % as the odds climb while the ROI conversely should climb. @ some point in the upward odds progression the ROI may flatten out. If the 10/1 to 19/1 horses had a win % higher than a lower odds segment then I would expect that these had been 'lucky' & were due to flatten out. For my regular spot plays there is a break even point in the lower ranges & the profit
ranges become fairly evident. I depend on profit in the 10/1 to 19/1 range, but I think the play should be profitable a range or two below that or also profitable above 20/1 for it to be viable.

With the actual odds available this is easy enough to query. This is a simple approach but it does detect those $90 bombs that distort an ROI. The average 3/1 shot & the avg. 11/1 shot are different animals even if they share similar ratings. Maybe relying on an avg. $Win which includes such a wide odds range just distorts a valid 7/1 to 20/1 play.

I like Carl's idea of having a decent ratio of 2nd & 3rd place finishers. I guess I think there should also be a similar ratio in odds segments for winners or there is something fishy about a big ROI.

Ken

[MikeDee]

OK did some homework and here goes.

Carl - I have some trouble following your logic on the ratio. You said in item 3
ave odds of winners / ave odds of all picked

then you gave and example of average winner price of 10-1 (I think you meant to say ave off odds) and average odds of all picked of 15-1.

going back to the ratio this would be 10 /15 or 0.6666, but in your explanation you quoted it as a ratio of 1.5, which equate to 15 /10.

Is the ratio ave odds all / ave odds win -OR- ave odds win / ave odds all ?

Glen I computed the a/e ratio for the plays attached. I used the 18% take out that was in the article, suppose it would make a minor difference if I had the actual track takeouts but don't know if the difference is worth the effort.


Here are the definitions you need to decipher the attached spreadsheet
aao - average of all odds
awo - average of winner odds
w/a - ratio of average winner odds /average all odds
ewp - expected win % (from Glen's post)
a/e - ratio of ave win % /expected win % (from Glen's post)

Bottom line for me -- spend less time watching the Indians lose on TV at night and more time playing MNR on good ole UBET!!

[Glen]

Thanks Mike,

Whatcha think of the of the A/E thing? I have yet to form an opinion about it. But the idea is an interesting one. Anybody else have an opinion?

[George]

Glen.......not sure of the value of A/E in terms of future predictability
of a spot play, but do know that Gordon Pine used it heavily in
ALL-IN-ONE as a measurement of how individual factors were doing.

Think MtKen approach of reviewing odds-range is simple and probably
a very good indicator of balance of the spot play.

My measurement preference has always been to develop queries over
a limited date range and then test across date ranges that contain
data the query has never seen before. That will usually tell you if the
play will work going forward. On many queries there is a real problem
in getting enough data to measure it properly.

Thanks to Carl for starting this thread and to Glen for digging out Pines
A/E formula. Great topic and I will be following it with much interest.

[Carl]

Good stuff Mike, thanks.

1) The ratio I was looking at was "the ratio ave odds all / ave odds win". As in
A. Your ratio at RD would be 7.1/2.7 or 2.62.
B. Your ratio at MTR would be 8.7/5.8 or 1.5.
C. the ratio off all tracks is 59.8/28 or 2.13
D. Hence, Criswell predicts that over time your RD and MTR ratios will even out so that a "normal" curve would give you the following over your next 280 bets at RD and next 240 bets at MTR:
RD 7.1/2.13 or 3.33 to one or average pay of 8.66. If your win percentage stays constant, I predict you will show a 54 races X amut 8.66 =$468 amount won. Loss cut to $112.
MTR 8.7/2.13 or 4.08 or an average pay of $10.16. Win cut to $8.

(Amazing.
Change of topic.
I just found I could acually Re-format the excel sheet you attached. I did my predictions for RD' next 280 bets and MTR's next 240 bets there and attached).

This board's technology is amazing.

Mike,
1) If you don't understand what I am doing please post back. Helps me too, I understand it better my self when/if I can explain it.
2) If you have some "pristine data" for RD or MTR, run 280/240 plays and see how my predictions work out. I guess the short of it is I am saying a "play's average off odds/over average winners odds" is a mean, and I think all tracks will gravate to that.
3) I think large fields are a big edge in getting a decent price. Would suspect that is MTR's edge. Be interesting to relate "average off odds" to "average field size". I have a >=7 nfld on most of my plays and stick to it.

Glen,
I still have to learn what Gordon is talking about. Mikey figured it out, I wanna be just like him when I grow up, I'll spend some time on it today, see if I can figure it out.

I'll attach my predictions file (lines 12 & 13 are my projects forward).

Interesting stuff.

later,
Carl

[Carl]

Opps, forgot one thing.

If I were you Mike, I would watch the Indians (do they ever win anymore?) evenings and take a vacation down to Keenland while they are still going. Can't believe them Kentucky types are letting your horses go off an average of 11-1!. I would guess that's where your next big winner will come from.

(And I am still wondering if my ratio means anything. It's simple anyway, I do like it for that).

later,
Carl

[MikeDee]

Thanks Carl for the clarification on the ave odds ratio. Is Criswell the author of that book you mentioned some time ago about data bases? Could your repost the name and author of that book? didn't get it yet and I want to.

You may be right about the results coming together, time will tell.

When I look at the average odds ratio here is what makes sense to me.

I would look for plays where the average of all odds and the average odds of the winners are closer together. I guess a ratio of one or less and the lower the better. In my way of thinking if the average odds of my winners are a lot lower then my average odds of all my picks then it is telling me that what I am doing is just picking caulk and why bother, I don't need to do anything to pick caulk and my spot play is picking the obvious.

Glen - don't know that I agree with the Pine article about needing something over and above impact value.

Impact value makes sense to me. It is telling us that the factor matters and is able to make picks better then what one would normally expect. I'm not sure that looking at A/E odds ration is any better or worse then just looking at the ROI. Unless there are situations where ROI points in one direction and the A/E ratio points in the other.

[Carl]

Humm,

Another country heard from. So you think the lower the ratio of off odds all horses/off odds winners the better? I never considered that. Maybe. Do you have some old data you could run to see if the low ratio plays hold up over time?

Criswell was a local psychic (maybe Canadian only) who used to do Jean Dixon foresee the future type stuff. The book I mentioned is I think called "The Predictors" and it is by Thomas Bass. I picked it up at GBC, it should be in their catalogue.

In any case, thanks for the second opinion. IF my play that I benched for having too low a ratio gets cranked up, I'll second your vote for "low ratio good".

And sometimes opposites both work in this crazy game. I remember when one of my track buddies liked to play horses that the leading jockey was getting on (he or the agent had to see something in the horse). I agreed that was a play, and then said I like to play horses the leading jockey was getting off of (he was on it for a reason, and now I'll get more price with him coming off). He agreed with my logic also......

later,
Carl

[Cliff]

Ya'll,

I had some limited success with "All-Ways" software which utilized 71 factors and their related ROI and IV. Some of you may be familiar with it. I'm not sure if it used Gordon Pine's IV calculation or the more simple version. What you were looking for were the factors that had both an ROI of $2.00+ AND an IV of 1.00 +. Surprisingly, it was sometimes tough to find 5 to 10 of the factors that would meet both criteria. But, if you did, you could use those factors in a "profile" to project the outcome of future races. Of course the research needed to be broken down by distance, class, maiden/non-maiden, etc. etc. and it got real expensive real fast to build up any kind of a statistically valid data base. I suppose it was back-fitting anyway.

But that brings up a point or two I have wondered about anyway:

1) Would it be possible (in HTR) to add the IV calculations along with the ROI as another measure of which factors are valid statistically?

2) What are the odds (in general) of a horse winning, and, coming back to win the next consecutive race, and then three in a row, etc.? Surely, someone has done some studies on that?

Later,
Cliff

[tomcat]

However.....the pragmatic view:

When at the track, if the horse looks good, and you like the odds, and looks good on the tlc,

bet'em!

The question is: If it wins most of the time, or some of the time...will it win today?

[George]

Unfortunately I have never been able to accurately answer Tomcat's
question "will it win today?". That's why I play mechanical plays
most of the time. At least I know it probably will win "somedays".

Ran the A/E on 8500 races in my db on all HTR factors. Attempting to
past a table of top 10 plus one of my spot plays. If table is messed up
will attach the word document.

Really liked the A/E as it confirmed the factors I already knew were
important in about the order I expected.
AE IV ROI
F1 2.06 1.42 0.94
SPOT 1.92 1.60 1.27
F3 1.89 1.30 0.89
ACL 1.88 1.58 0.90
TRN 1.82 1.62 0.97
JKY 1.81 1.57 0.99
HTR 1.23 2.25 0.92
POW 1.20 2.14 0.85
K 1.14 2.22 0.86
PRB 1.12 2.33 0.89

[George]

Unable to fix prior table in edit. Trying to attach as rtf file.

[Rick]

George,
The trick is to use [ code ] at the beginning and [ /code ] at the end. You just have to leave the spaces out when using the code.

I went ahead and fixed it for you. You should be able to click on edit and see what I did.

[George]

Thanks Rick. Looked at it and you are right, very simple if
you know how. Maybe I will get the hang of it shortly.

BTW, the AE run was using PL1 and checking "ranked"
factors equal 1. Have not had time to run on lower rankings
but will report any gems if find any.

[MtKen]

George:

Have you ever done any form analysis with your spot plays besides layoff days?
I have started (barely) to use Prat to look at # of races since a layoff to see if that might improve anything? Wondered if you or anyone has looked at this.

Ken

[Glen]

Very nice George, Like those numbers. Good to see the flow of those figs. Are you using ricks access template and then making adjustments to the ROI calculator?

[George]

MtKen.......interesting you mention layoffs, as a recent spot I am working on concerns playing long layoff horses. Think recency is an old horseracing myth. I know from statistcal studys that recency is meaningless in modern racing. Many people play second after layoff, third after layoff, etc. Suspect there is some merit in that type of analysis.

I use prat mostly for horse's record at the track and distance. Also like to analyze form pattern to try and predict "bounce" but not very good at it.


Glen............not smart enough to modify the template. Wrote three sql statements and ran them one at a time for each ranked factor. Am trying to combine into one query but have not been able to get it to work yet.

Think the AE is an excellant tool to find factors to use as main basis in spot plays. The ratio is a nice balance of win% versus "betting pressure". Am sure one could use any one of F1,F3,ACL,TRN,JKY and combine with a couple ugly factors for a nice spot play. Could use the same base query to check AE by track and probably raise roi considerably.