PPs - Equibase vs Brisnet vs DRF

[mikesal57]

I just went thru this whole thread looking for any comments on HDW.

Very little was said.

I doubt anyone can give an answer to this question because of the extensive cost to subscribe to all of them and test. Unless someone did

Anyway , here is Sat CD 6TH 2020.....

The first picture is Bris PP....The winner is #8 at 26-1....The #10 was the even money favorite...
The second picture is HDW in Alldata software....

Who would you think is the better horse?
This isn't a slight difference...it is HUGH

If CJ can put up his Timeform numbers for this race , it will make it interesting.


Comments?

[cj]

Attached.

[mikesal57]

Originally Posted by cj

Attached.

Thanks CJ....

So your opinion is?

[Jeff P]

I actually took a stand against the favorite and made a small bet on the #8 in that race. (Just some modest win bets in a race with a big field where I thought the favorite was overbet.)

Nothing like Jay68802 in the other thread where he posted that big DD.
http://www.paceadvantage.com/forum/s...8&postcount=28

Originally Posted by jay68802

Good hope someone is right

CHURCHILL DOWNS
RACE 5
DB
$2.00 /

Will Pays:

$369.90
$818.40
$1643.80
$993.60
$63.60
$290.20

Got to play that Santana at CD.

Gotta say it Jay: Nice!




Below are two screenshots of JCapper PPs using HDW data for #8 and the race favorite #10 from Sat 05-30-2020 Churchill R6:

• First, #8 at 26-1 --

The most recent turf line is highlighted in blue. Of note to me is the race distance (8.5F half a furlong shorter than today), the class level (MSPLWT likely lower than today) the late pace fig (113), and the final time speed fig (98.)

Also highlighted in yellow are selected data points from an older turf line Jan 18, 2020 FGX R7. Of note to me is the race distance (9.0F same as today), the class level (MSPLWT) the late pace fig (99), and the final time speed fig (103.)


• Next, #10 at 1-1 --

The most recent turf line is highlighted in blue. Of note to me is the race distance (8.5F half a furlong shorter than today), the class level (G3 higher than today) the late pace fig (114), and the final time speed fig (81.)

• Then, some comments about both horses and my thought process leading up to post time:

#8 at 26-1 has higher final time speed figs and higher late pace figs in both turf lines than #10 at 4-5.

Additionally, #8's best late pace turf fig (113 albeit at a shorter distance) is competitive (within a point) of #10's best late pace turf fig (114 also at a shorter distance.)

One of the little things that had me thinking the #8 might have a shot was the possibility he might appreciate 9F on the turf (the distance where he ran his lifetime best final time speed fig on the turf.)

#10 at 1-1 had earned his 114 late pace fig in a G3. This is one of the little things that had me thinking the #10 might be legit. (The #8 horse hadn't been asked to race in stakes company yet.)

One of the little things that had me thinking the #10 might be overbet was the thought (provided the figures were accurate) if both riders worked out good trips and squared off during the final 3/8ths, if the #10 fired a 114 and the #8 fired a 113:

Both horses would be right there and it would come down to whichever rider worked out the better trip.

With the #8 at 22-1 (or so) facing up to the gate and the #10 at odds on it was kind of a no brainer.

This might be a good race to take a stand against the favorite.

As it happened Brian Hernandez aboard #8 worked out a great trip and won without really being challenged in the stretch.

Truth be told I didn't use just the #8. A similar thought process found me including the #1 and the #6.

FWIW, the strategy of weighing relevant figs of longer odds horses against those of the favorite - especially in big fields - is something that pays off every now and then.



Because this is a thread about figs made by the various data vendors I'll throw in one final thought --

The above race represents a sample of 1.

Using a strategy similar to that described above, if I looked at a month's worth of races, and if I cherry picked my spots --

I'm pretty sure I could point out many individual races where (reading between the lines) the figs were suggesting an overbet favorite who (in hindsight) was defeated by a get-able longshot.

Imo, this would be true no matter whose figs I used.

Imo, the real question (likely never to be answered) isn't: Which figure maker does the best job?

But rather: Whose figures offer the most value?

What I'm about to say may be counter-intuitive to many of you reading this post.

The figure maker doing the best job may not in fact be the one garnering the most votes in a thread like this or the one most widely distributing his numbers.

Imo, the figure maker doing the best job --

Would be defined as the one whose figures consistently offer the most value at the windows.

Imo, that's not a question likely to be answered soon.

Imo, figures that consistently offer value at the windows have to be decent enough and be grounded with a solid basis in reality.

But at the same time, in order to consistently offer value at the windows they can only be seen by a limited number of eyeballs each day.

The day you find them, if you are smart, you will happily use them without ever saying a word.



-jp

.

[mikesal57]

Originally Posted by Jeff P

I actually took a stand against the favorite and made a small bet on the #8 in that race. (Just some modest win bets in a race with a big field where I thought the favorite was overbet.)

Nothing like Jay68802 in the other thread where he posted that big DD.
http://www.paceadvantage.com/forum/s...8&postcount=28



Gotta say it Jay: Nice!




Below are two screenshots of JCapper PPs using HDW data for #8 and the race favorite #10 from Sat 05-30-2020 Churchill R6:

• First, #8 at 26-1 --

The most recent turf line is highlighted in blue. Of note to me is the race distance (8.5F half a furlong shorter than today), the class level (MSPLWT likely lower than today) the late pace fig (113), and the final time speed fig (98.)

Also highlighted in yellow are selected data points from an older turf line Jan 18, 2020 FGX R7. Of note to me is the race distance (9.0F same as today), the class level (MSPLWT) the late pace fig (99), and the final time speed fig (103.)


• Next, #10 at 1-1 --

The most recent turf line is highlighted in blue. Of note to me is the race distance (8.5F half a furlong shorter than today), the class level (G3 higher than today) the late pace fig (114), and the final time speed fig (81.)

• Then, some comments about both horses and my thought process leading up to post time:

#8 at 26-1 has higher final time speed figs and higher late pace figs in both turf lines than #10 at 4-5.

Additionally, #8's best late pace turf fig (113 albeit at a shorter distance) is competitive (within a point) of #10's best late pace turf fig (114 also at a shorter distance.)

One of the little things that had me thinking the #8 might have a shot was the possibility he might appreciate 9F on the turf (the distance where he ran his lifetime best final time speed fig on the turf.)

#10 at 1-1 had earned his 114 late pace fig in a G3. This is one of the little things that had me thinking the #10 might be legit. (The #8 horse hadn't been asked to race in stakes company yet.)

One of the little things that had me thinking the #10 might be overbet was the thought (provided the figures were accurate) if both riders worked out good trips and squared off during the final 3/8ths, if the #10 fired a 114 and the #8 fired a 113:

Both horses would be right there and it would come down to whichever rider worked out the better trip.

With the #8 at 22-1 (or so) facing up to the gate and the #10 at odds on it was kind of a no brainer.

This might be a good race to take a stand against the favorite.

As it happened Brian Hernandez aboard #8 worked out a great trip and won without really being challenged in the stretch.

Truth be told I didn't use just the #8. A similar thought process found me including the #1 and the #6.

FWIW, the strategy of weighing relevant figs of longer odds horses against those of the favorite - especially in big fields - is something that pays off every now and then.



Because this is a thread about figs made by the various data vendors I'll throw in one final thought --

The above race represents a sample of 1.

Using a strategy similar to that described above, if I looked at a month's worth of races, and if I cherry picked my spots --

I'm pretty sure I could point out many individual races where (reading between the lines) the figs were suggesting an overbet favorite who (in hindsight) was defeated by a get-able longshot.

Imo, this would be true no matter whose figs I used.

Imo, the real question (likely never to be answered) isn't: Which figure maker does the best job?

But rather: Whose figures offer the most value?

What I'm about to say may be counter-intuitive to many of you reading this post.

The figure maker doing the best job may not in fact be the one garnering the most votes in a thread like this or the one most widely distributing his numbers.

Imo, the figure maker doing the best job --

Would be defined as the one whose figures consistently offer the most value at the windows.

Imo, that's not a question likely to be answered soon.

Imo, figures that consistently offer value at the windows have to be decent enough and be grounded with a solid basis in reality.

But at the same time, in order to consistently offer value at the windows they can only be seen by a limited number of eyeballs each day.

The day you find them, if you are smart, you will happily use them without ever saying a word.



-jp

.

Good to hear from you Jeff
and an excellent response...

In a nut shell you said it...
"Imo, the figure maker doing the best job --

Would be defined as the one whose figures consistently offer the most value at the windows.

Imo, that's not a question likely to be answered soon."

Wouldn't it be nice if some did or created a program that can identify these differences?...Wink Wink

Mike

[Jeff P]

Writing a program to do the comparison is the easy part.

I have HDW data dating back to 2009.

The only thing I'm missing is 3-5 years of back data for every thoroughbred race for Equibase, Brisnet, Beyer, TimeformUS, Ragozin, Thoro-Graph, etc. Some of it not available and some of it priced at $1k a pop or thereabouts for each year.





-jp

.

[mikesal57]

Originally Posted by Jeff P

Writing a program to do the comparison is the easy part.

I have HDW data dating back to 2009.

The only thing I'm missing is 3-5 years of back data for every thoroughbred race for Equibase, Brisnet, Beyer, TimeformUS, Ragozin, Thoro-Graph, etc. Some of it not available and some of it priced at $1k a pop or thereabouts for each year.





-jp

.

Jeff...your hitting the extreme here..LOL

what about starting with one as a test....

PLUS...What are you doing up so early...arent you on the west Coast?

Mike

[Speed Figure]

Numbers wise the isn’t to far off from the I think the biggest factor was that the finished 2nd by a length in a G3 last race and the class handicappers went overboard because of that and the colt had the big name trainer and jockey!

[mikesal57]

Originally Posted by Speed Figure

Numbers wise the isn’t to far off from the I think the biggest factor was that the finished 2nd by a length in a G3 last race and the class handicappers went overboard because of that and the colt had the big name trainer and jockey!

I agree....

Timeform and HDW had horses ,IMO, correctly and evenly matched..

Bris , on the other hand, under estimated the 8....


I'm excusing the 10 for running 3rd to a possible "bounce" off top figure...

The point from the start is we really dont know what kind of figures we have because no one has ever exploited them...

[cj]

Originally Posted by mikesal57

Good to hear from you Jeff
and an excellent response...

In a nut shell you said it...
"Imo, the figure maker doing the best job --

Would be defined as the one whose figures consistently offer the most value at the windows.

Imo, that's not a question likely to be answered soon."

Wouldn't it be nice if some did or created a program that can identify these differences?...Wink Wink

Mike

The hard part here isn't running the test, it is defining it. Last race? Best of two/three? Average of some sort? Only use same or similar distance? Surface? What do you do with trainer changes? The list goes on and on. I'm not saying there isn't an answer to be found, I'm sure there is. But first you have to decide on the above.

[mikesal57]

Originally Posted by cj

The hard part here isn't running the test, it is defining it. Last race? Best of two/three? Average of some sort? Only use same or similar distance? Surface? What do you do with trainer changes? The list goes on and on. I'm not saying there isn't an answer to be found, I'm sure there is. But first you have to decide on the above.

I believe the start is with figures....

You can use ......best...best 2 of PP's, avg whatever...etc

Once you gather all the info in a compatible format...with results

the rest is easy

[Jeff P]

If I were designing it there would be definitely be a number of tests.

Best last out, last 3, last 5, 2 of last 5, last 10, avg best 2, avg last 3, avg last 5, avg last 10, etc.

Lowest last out, last 3, last 5, 2 of last 5, last 10, avg lowest 2, avg last 3, avg last 5, avg last 10, etc.

Stuff like the above for categories like overall, surface, distance, track condition, track code, etc.

But if you start adjusting raw figs for things like rider, trainer, post position, days since fig was earned, workout patterns leading up to today's race, etc. --

It hits you that a project that looks simple at first (for example comparison of best last out among two or three different fig makers) can really explode in scope.



-jp

.

[Tom]

Best usable last race figure using the LASST method:

LA - Layoffs
S - Surface (Turf, Sloppy Etc.)
S - Structure of Distance (Sprint, Route Etc.)
S - Slow Paced Race
T - Trouble in the race

[cj]

Originally Posted by Tom

Best usable last race figure using the LASST method:

LA - Layoffs
S - Surface (Turf, Sloppy Etc.)
S - Structure of Distance (Sprint, Route Etc.)
S - Slow Paced Race
T - Trouble in the race

I still use this today and even add a B to the beginning occasionally for Bias.

[classhandicapper]

I have Beyer figures and very basic class ratings for stakes races in my database going back to late 2014.

I just look at the last race in various categories (dirt, turf, sprint, route etc..) to see which performs better in each category. There are a lot of races where it's obvious the last race wasn't the best one in the PPs to use, but both figures suffer equally from that. So it probably all evens out in a large sample for just testing purposes.

I put more focus on trying to combine them because each suffers from weaknesses and accuracy issues, but different ones.

I've been able to prove that combining them works better than either does alone.

I've been able to prove that looking at multiple races in a weighted way (more recent gets more weight) outperforms looking at just one race.

I wish I had time to make manual class ratings of a higher quality than I can generate systematically (kind of like Timeform Europe for races outside the US). It's hard for me to know how much value I add with a lot of my subjective class analysis. For I know, it's not very much.

I'd also like to code some running line selection rules, but it seems like it would be tough for me to get right because it's kind of subjective and depends on the case.