Purse Value

[jerry-g]

Originally Posted by HUSKER55

I promise, I will not post till my second cup of coffee.

I think I miss read your chart. is the WPS net points?

Dayjob has 13 starts and 56 wins, 44 places and 44 shows.

I thought "form" was the WPS value

Right! Form is the value of the races ran for W P S in only the last six
races applied to the formula developed by Pitlak when he improved upon
Mitchel's formula. The math section is only if needed to total up the values
when you have several to compute.

[raybo]

Originally Posted by jerry-g

Thank-you for your input and you're previous post as well. I think it is very
good advice and I will incorporate it into my new template. I'm going to let
the Mitchell template stand on its own, to be or not to be.

My second template uses a form number derived out of a formula for W,P,S,
and then adds that to the EPS of the horse. Bob Pitlak improved upon the
original done by Larry Voegele in 1972 but I am sure he did not limit his
WPS values to surface and distance. Your idea makes more sense to me as
I have always realized the importance to comparing apples to apples.

I'm going to give my new template a whirl and see how it does. I have it
attached to this post.

The highlighted section, in my own philosophy, is extremely important. That being that, once one has found something that is both logical and consistent, keep it separated ("treasured" as you will) from all future work. That applied to, for me, early Sartin, Quirin, and Giles. I never went past the original works, trying to keep up with others' "improvements" to them. Instead, I used the originals as the core, and modified as I saw fit, without destroying the original concept.

By keeping the original work separate, one does not lose that initial "inspiration/genius" as easily as if one starts modifying it immediately, trying to make the results better (by backfitting it to results). Sure, the idea of making something better, is great, but often we don't ever really know if the mods made it better, long term, until we hit the wall of negative result (it's a slippery slope we travel). At that point, we have already invested too much time and effort, into trying to improve the original concept, to try to backtrack to a point where the mods actually negated the original premise. By saving the original, we can easily return to it, and begin again, often finding that the original concept is still viable, and that it was the mods that weren't.

[raybo]

Going back to the 2nd attachment you posted, you have gotten rid of all the "%" columns as well as the "min" and "max" data, from the "Mitchell" workbook. What was your thinking in doing that?

Also, how are you using the data in the 2nd section, below the CCR section at the top?

Am I to assume that in the CCR section, "Wins", "Place", and "Show" columns, you are using the same time eras as in the original "Mitchell" workbook (this year plus last year when less than 3 races this year), but the "Form" column only looks back 6 races? Or have you switched to only the last 6 races for all those columns?

I do think that "recency" (regarding the 6 races back "Form" number) is important, regardless of what you're working on in horse racing. And longer term data dilutes any current value that any method might have, except in things like "general" factor evaluation querying in database work (2 schools of thought regarding databases, large sizes encompassing years of data, and shorter, more recent data).

My thoughts are to continue testing the original Mitchell method, at more tracks, to see if the amazing consistency I saw at Aqu repeats at other tracks. If it doesn't, then - and only then - proceed to the next evolution. I feel that it is important to test, thoroughly, each evolution before modifying it, ie: fully evaluating the original concept, before discarding it as valueless.

[raybo]

Originally Posted by raybo

Am I to assume that in the CCR section, "Wins", "Place", and "Show" columns, you are using the same time eras as in the original "Mitchell" workbook (this year plus last year when less than 3 races this year), but the "Form" column only looks back 6 races? Or have you switched to only the last 6 races for all those columns?

I think I have answered my own quoted question after rereading prior posts. The number of wins, place, and show data should have been for only the last 6 races. I missed that in my testing of Aqu, instead I used the same time era as for "earnings". I am assuming that you were using the same time era for "Starts" as "earnings". If that is wrong please let me know, then I'll make those changes and retest Aqu.

[jerry-g]

Originally Posted by raybo

I think I have answered my own quoted question after rereading prior posts. The number of wins, place, and show data should have been for only the last 6 races. I missed that in my testing of Aqu, instead I used the same time era as for "earnings". I am assuming that you were using the same time era for "Starts" as "earnings". If that is wrong please let me know, then I'll make those changes and retest Aqu.

Yes I use the box for the starts and earnings and the WPS columns for the
value of the races in last six starts. I still have the original CCR intact. Might
be simpler to use and less chance of error. But I would like to add the odds
table to it as well since it is helpful knowing R.O.I. on a bet.
I'm afraid I've confused this issue by uploading that second template. I should
have kept it in my top secret drawer. The original Mitchell CCR uses numbers
not value for the WPS. I need to kiss the Blarney Stone and dance around
a tree several times.

[raybo]

Originally Posted by jerry-g

Yes I use the box for the starts and earnings and the WPS columns for the
value of the races in last six starts. I still have the original CCR intact. Might
be simpler to use and less chance of error. But I would like to add the odds
table to it as well since it is helpful knowing R.O.I. on a bet.
I'm afraid I've confused this issue by uploading that second template. I should
have kept it in my top secret drawer. The original Mitchell CCR uses numbers
not value for the WPS. I need to kiss the Blarney Stone and dance around
a tree several times.

Ok, so if I am understanding correctly this time:

Starts = this year's starts unless there are less than 3 races, then we use this year plus last year.

Earnings = this year's earnings unless there are less than 3 races, then we use this year plus last year.

Wins, Place, Show = last 6 races only

I'm pretty sure I have everything else correct in the original Mitchell CCR calculations.

[classhandicapper]

"value of the races in last six starts. "

Did you explain how you get the values for each race?

[jerry-g]

Originally Posted by classhandicapper

"value of the races in last six starts. "

Did you explain how you get the values for each race?

This is my fault for trying to do two things so close together. The first
template refers to Mitchell's original method. The value for each race is
the 1st, 2nd or 3rd finish in a horses last six races. If he has one win,
0 place and 2 show you would enter in the WPS column 1-0-2 and the
formula would do the rest.
That method was improved upon by another fellow and that is why I
created the second template. For value in that template, you use the
actual purse value if known or the dollar value of the race. So if you have
in the last six races 1 win at 50K and 1 win at 25K you would enter the
total of 75 in the win column. If no place finishes you would enter 0 in
the P Column and for no shows a 0 in the S column and the formula will
compute the result and display it in the Form column which then would
get added to the EPS and totaled making the CCR rating.
We are now trying to determine the value of either of these approaches.
My original intent is to isolate contenders by at least capturing the winner
in the top three picks. Neither of these two templates are suppose to be
used as a horse picking system to win the race. It is suppose to aid in
isolating the three horses to further handicap for the win position using
your favorite method. It still remains to be seen if these templates are of
any value at all. Mitchell his self said that the method is of little use in
predicting winners. I think the templates can be used by anyone to
customize them in different ways that makes sense to the individual. It
seems to me a good way to start trying to get a handle on Class and
then add your favorite factors for Speed, Pace, RS, ets. Hopefully, I will
be helping people in some way instead of continuing to confuse the issue.

[raybo]

Originally Posted by jerry-g

Yes I use the box for the starts and earnings and the WPS columns for the
value of the races in last six starts. I still have the original CCR intact. Might
be simpler to use and less chance of error. But I would like to add the odds
table to it as well since it is helpful knowing R.O.I. on a bet.
I'm afraid I've confused this issue by uploading that second template. I should
have kept it in my top secret drawer. The original Mitchell CCR uses numbers
not value for the WPS. I need to kiss the Blarney Stone and dance around
a tree several times.

After making the changes to #win, #place, and #show to only use the last 6 races, for the original Mitchell CCR calculations, I reran the same cards for Aqueduct again. The results were better than the first test.

Feb 2013 - Hit rate for the top 3 rankings ranged from 58.33% to 66.67% and ROI ranged from 0.78 to 1.05.

Through Mar 2013 - Hit rate for the top 3 rankings ranged from 56.35% to 64.44% and ROI ranged from 0.81 to 0.92.

Through Apr 2013 - Hit rate for the top 3 rankings ranged from 55.19% to 58.42% and ROI ranged from 0.81 to 0.88.

Through Nov 2013 - Hit rate for the top 3 rankings ranged from 57.36% to 59.39% and ROI ranged from 0.84 to 0.88.

Through Dec 2013 - Hit rate for the top 3 rankings ranged from 58.62% to 59.49% and ROI ranged from 0.83 to 0.85.

Through Jan 2014 - Hit rate for the top 3 rankings ranged from 58.32% to 60.16% and ROI ranged from 0.81 to 0.83.

Through Feb 2014 - Hit rate for the top 3 rankings ranged from 58.85% to 60.54% and ROI ranged from 0.81 to 0.84.

Through Mar 3rd, 2014 - Hit rate for the top 3 rankings ranged from 58.13% to 58.31% and ROI ranged from 0.80 to 0.81.

So, hit rates were generally around 58% - 59% and ROIs were generally around 0.83 - 0.85. That's more like a fair 3 horse contender selection method. Not far from what the public does.

However, the individual hit rates between the 3 ranked picks are more closely grouped than is the public's. Top ranked hit rate was 22.66%, 2nd ranked was 18.69%, and 3rd ranked was 16.77%. So, although lower than the public's, they are more consistent as a group, with only about 6% separating the 1st ranked from the 3rd ranked.

Can you, through this consistency between the 3 rankings (or 4 or 5 maybe), use the actual CCR ratings to create a good fair odds line? I don't know, but my initial thinking is that you could, if you factored in odds line power and confidence multipliers differing by individual track results.

More tracks need to be tested to see if these Aqu hit rate and ROI ranges are fairly consistent, or if they will vary widely by track.

Here's the screenshot from that Aqu test:

[jerry-g]

"Can you, through this consistency between the 3 rankings (or 4 or 5 maybe), use the actual CCR ratings to create a good fair odds line? I don't know, but my initial thinking is that you could, if you factored in odds line power and confidence multipliers differing by individual track results."

I need to see if I understand what you are asking. I could use an odds
table with Vlookup to find the odds based on Mitchell's CCR total number.
I don't think it would be a good fair odds line as it would be based on one
factor, Class Consistency Rating. It would be more like a factor fair odds.
Are confidence multipliers the same as Impact Values? If so, the IV could
be applied easily to the CCR number and be reflected in the odds line.
Let me know. I would not have any way to generate IV values across the
different tracks.

[classhandicapper]

Originally Posted by jerry-g

This is my fault for trying to do two things so close together. The first
template refers to Mitchell's original method. The value for each race is
the 1st, 2nd or 3rd finish in a horses last six races. If he has one win,
0 place and 2 show you would enter in the WPS column 1-0-2 and the
formula would do the rest.
That method was improved upon by another fellow and that is why I
created the second template. For value in that template, you use the
actual purse value if known or the dollar value of the race. So if you have
in the last six races 1 win at 50K and 1 win at 25K you would enter the
total of 75 in the win column. If no place finishes you would enter 0 in
the P Column and for no shows a 0 in the S column and the formula will
compute the result and display it in the Form column which then would
get added to the EPS and totaled making the CCR rating.

I like it.

What is the value of a 2nd or 3rd in a 50K race vs. a 1st in a 50K race?

[classhandicapper]

Originally Posted by raybo

After making the changes to #win, #place, and #show to only use the last 6 races, for the original Mitchell CCR calculations, I reran the same cards for Aqueduct again. The results were better than the first test.

Feb 2013 - Hit rate for the top 3 rankings ranged from 58.33% to 66.67% and ROI ranged from 0.78 to 1.05.

Through Mar 2013 - Hit rate for the top 3 rankings ranged from 56.35% to 64.44% and ROI ranged from 0.81 to 0.92.

Through Apr 2013 - Hit rate for the top 3 rankings ranged from 55.19% to 58.42% and ROI ranged from 0.81 to 0.88.

Through Nov 2013 - Hit rate for the top 3 rankings ranged from 57.36% to 59.39% and ROI ranged from 0.84 to 0.88.

Through Dec 2013 - Hit rate for the top 3 rankings ranged from 58.62% to 59.49% and ROI ranged from 0.83 to 0.85.

Through Jan 2014 - Hit rate for the top 3 rankings ranged from 58.32% to 60.16% and ROI ranged from 0.81 to 0.83.

Through Feb 2014 - Hit rate for the top 3 rankings ranged from 58.85% to 60.54% and ROI ranged from 0.81 to 0.84.

Through Mar 3rd, 2014 - Hit rate for the top 3 rankings ranged from 58.13% to 58.31% and ROI ranged from 0.80 to 0.81.

So, hit rates were generally around 58% - 59% and ROIs were generally around 0.83 - 0.85. That's more like a fair 3 horse contender selection method. Not far from what the public does.

However, the individual hit rates between the 3 ranked picks are more closely grouped than is the public's. Top ranked hit rate was 22.66%, 2nd ranked was 18.69%, and 3rd ranked was 16.77%. So, although lower than the public's, they are more consistent as a group, with only about 6% separating the 1st ranked from the 3rd ranked.

Can you, through this consistency between the 3 rankings (or 4 or 5 maybe), use the actual CCR ratings to create a good fair odds line? I don't know, but my initial thinking is that you could, if you factored in odds line power and confidence multipliers differing by individual track results.

More tracks need to be tested to see if these Aqu hit rate and ROI ranges are fairly consistent, or if they will vary widely by track.

Here's the screenshot from that Aqu test:

You might be able to improve the "hit rate" by insisting on good current form (like best last race of the top 3 or something like that).

[jerry-g]

Originally Posted by classhandicapper

I like it.

What is the value of a 2nd or 3rd in a 50K race vs. a 1st in a 50K race?

The values are reduced by the formula used by Pitlak. Click on any
cell in the FORM column and look at the formula.

[classhandicapper]

Originally Posted by jerry-g

The values are reduced by the formula used by Pitlak. Click on any
cell in the FORM column and look at the formula.

Got you.

Thanks to both of you for making the spreadsheet available and sharing the results of the studies. I wish I had that kind of database access.

At some point I could use something this as a core for what I do and put some personal tweaks into it to match my own approach.

[raybo]

2nd track test. I decided for the 2nd test to contrast it with a track that plays nothing like Aqueduct. Delta Downs fits the bill. I expected the hit rate and ROI to be lower than Aqu because of the large number long shots at Delta, and good class winners there usually result in very low prices. I was correct in my expectations, but, the consistency of this method was still very good, even though lower. The cards started on 2/2/2013 through 3/23/2013, then continued from 10/11/2013 through 3/1/2014. 842 races were played by the method (passed all races with 20% or more of the field not distance or surface qualified, as before).

Feb 2013 - Hit rate for the top 3 rankings ranged from 41.98% to 57.14% and ROI ranged from 0.64 to 0.76.

Through Mar 2013 - Hit rate for the top 3 rankings ranged from 40.32% to 42.42% and ROI ranged from 0.64 to 0.67.

Through Oct 2013 - Hit rate for the top 3 rankings ranged from 41.38% to 42.55% and ROI ranged from 0.64 to 0.68.

Through Nov 2013 - Hit rate for the top 3 rankings ranged from 41.93% to 42.88% and ROI ranged from 0.67 to 0.68.

Through Dec 2013 - Hit rate for the top 3 rankings ranged from 42.32% to 43.06% and ROI ranged from 0.68 to 0.70.

Through Jan 2014 - Hit rate for the top 3 rankings ranged from 42.90% to 43.50% and ROI ranged from 0.69 to 0.70.

Through Feb 2014 - Hit rate for the top 3 rankings ranged from 43.48% to 44.97% and ROI ranged from 0.69 to 0.72.

Through Mar 1, 2014 - Hit rate for the top 3 rankings 44.89% and ROI 0.72.

So, generally, 3 pick combined hit rate was around 42% and ROI was around .068 - 0.69.

Individual pick hit rates hovered around 18 - 20% for top pick, 13 - 16% for 2nd pick, and 11 - 14% for 3rd pick, with the overall hit rates finishing at 18.05%, 14.13%, and 12.71% respectively. That's a difference, between the 3 picks, of a little less than 6%, about the same as what Aqu produced, even though the hit rates there were higher.

Here's the screenshot from that test: