Top Horse Analytics: 12/05/2018 Free Pick of Day - GP, TAM

[Augenj]

Link only works for day of race.

http://www.equibase.com/products/freepicks2.cfm#GP_th
http://www.equibase.com/products/freepicks2.cfm#TAM_th

[VigorsTheGrey]

When the stats for THA are compiled for earnings, Trainer, jockey and others...Do the flags represent Data from the “logical start date” for the current meeting, or from the entire year to date...?

[Augenj]

Originally Posted by VigorsTheGrey

When the stats for THA are compiled for earnings, Trainer, jockey and others...Do the flags represent Data from the “logical start date” for the current meeting, or from the entire year to date...?

OK, hold on tight here.

A "meet" for me means racing days, not whatever the track calls it or changes the name of it during the year. Also, a gap of two weeks or so between racing days doesn't reset the start date of the logical year.

* If the track has only one meet within a year of the start of the meet (think AQU), the date the meet starts is the start of the logical year even if it extends into the next calendar year.

* If the track has two meets within a year of the start of a meet (think BEL), the date the first meet starts is the start of the logical year.

* Now it gets a little tricky when the track has three meets of contiguous racing days within a year of the start of a meet (think CD and SA). The start of the logical year would be the day after the longest gap between the three meets. I do this to try and stack race days as close as possible in my data bases.

On another note, GP runs a long meet with roughly 2,000 contiguous races. I break this up into two data bases, GP (winter) and GPS (spring/summer). Gulfstream even calls it their Spring and Summer meet. Each data base is around 1,000 races, plus or minus some. The start of GPS is when the current jockeys, trainers, and horses are heading north and this affects the THA analysis runs that create algorithms for rating horses in a race. Therefore, GPS gets a separate analysis from GP.

Thanks for asking. Nobody else has before. Trouble is that my explanations make sense to me but not so much to others.

[VigorsTheGrey]

Racing dates per DRF: https://www1.drf.com/Entries/RacingDates.do


BELMONT PARK BEL
09/07/2018 - 10/28/2018 (TB)
04/27/2018 - 07/15/2018 (TB)

You’re right, it takes a little thinking to unravel what you are saying here.

So in the case of Belmont, ARE the numbers used during the Fall meet including numbers starting from the “logical year date” of 04/27/2018...?

Now in the case of Tampa Bay: Per DRF,
TAMPA BAY DOWNS TAM
07/01/2018 - 06/30/2019 (TB)

Yet, on THA, the start date is shown as 11/24/18...?

Anyway, just for clarifications, in the case of Tampa Bay, are the compilations beginning fresh as of the most recent start a few days ago and moving forward from there...? So that the jockey flag, for example, goes to the jockey with the most wins (or is it earnings) since a few racing days ago, AND NOT compiled from the whole years’ dates prior to the most recent start on this new meeting...?

[VigorsTheGrey]

GULFSTREAM PARK GP
01/01/2019 - 03/31/2019 (TB)
12/01/2018 - 12/30/2018 (TB)
07/01/2018 - 09/30/2018 (TB)
04/05/2018 - 06/30/2018 (TB)

http://www.paceadvantage.com/forum/a...6&d=1543845337

Logical start date shown as 12/01/18.

[Augenj]

With BEL, logical year start date is 4/18/2018 and does include the Fall meet.

TAM doesn't race 365 days a year so the logical year start date is the first day they raced (11/24/2018) after a long break from June of no racing.

Jockeys, Trainers, etc stats are calculated with their history of wins for the past year as a win percent to give them a ranking.

GP just started their logical year on 12/01/2018 and will run until April of 2019. Then Gulfstream will start their Spring/Summer season (GPS) and run until GPW starts. GPS is not a track code but one I created to split the winter/spring-summer seasons.

[VigorsTheGrey]

Thank you for the info...so the jockey, trainer, and earnings are cumulative from January 1st on, then imported into the program...so that all wins from all tracks comprise the rankings, which then are updated each day for each category so that even interloping jockeys and trainers and earnings are recorded in every case...the jockey / Trainer ratings and earnings are not necessarily venue-exclusive in all cases...?

These particular ratings are carried over into the ‘logical start’ date for such and such track...? And then begin fresh again on Jan 1st of the new year...? Or do they just continue to be updated...?

[Augenj]

Originally Posted by VigorsTheGrey

Thank you for the info...so the jockey, trainer, and earnings are cumulative from January 1st on, then imported into the program...so that all wins from all tracks comprise the rankings, which then are updated each day for each category so that even interloping jockeys and trainers and earnings are recorded in every case...the jockey / Trainer ratings and earnings are not necessarily venue-exclusive in all cases...?

These particular ratings are carried over into the ‘logical start’ date for such and such track...? And then begin fresh again on Jan 1st of the new year...? Or do they just continue to be updated...?

There is no starting "reset" for any of the JTSFEWLC flags. These flags are calculated using the latest DRF(Equibase) data. The best in the race are calculated and flagged then. As time goes on, the data will be updated at Equibase and that data will be used to calculate the flags for that day.

[VigorsTheGrey]

Originally Posted by Augenj

There is no starting "reset" for any of the JTSFEWLC flags. These flags are calculated using the latest DRF(Equibase) data. The best in the race are calculated and flagged then. As time goes on, the data will be updated at Equibase and that data will be used to calculate the flags for that day.

So for the Earnings flag, for example, is the data then for the earnings this year only, or for total earnings...the Equibase data seems to have 2 categories, one for total(lifetime) and one for year by year (which I assume starts and ends January 1st each year, and may not align, necessarily, with the ‘logical’ starts of THA...

[Augenj]

Originally Posted by VigorsTheGrey

So for the Earnings flag, for example, is the data then for the earnings this year only, or for total earnings...the Equibase data seems to have 2 categories, one for total(lifetime) and one for year by year (which I assume starts and ends January 1st each year, and may not align, necessarily, with the ‘logical’ starts of THA...

A computer analysis run is made on each track using last year's data base. This happens before the start of every logical year. Earnings are calculated several ways in the analysis. The best "way" is used for this track the rest of the logical year.

The "ways" are:
lifetime earnings only
lifetime earnings + 0.5 times the current year
lifetime earnings + 1.0 times the current year
on up to
lifetime earnings + 5.0 times the current year

Some tracks overlap the start of the calendar year which affects the current year values. One track may be biased heavily with the current year and others may not.

[VigorsTheGrey]

Originally Posted by Augenj

A computer analysis run is made on each track using last year's data base. This happens before the start of every logical year. Earnings are calculated several ways in the analysis. The best "way" is used for this track the rest of the logical year.

The "ways" are:
lifetime earnings only
lifetime earnings + 0.5 times the current year
lifetime earnings + 1.0 times the current year
on up to
lifetime earnings + 5.0 times the current year

Some tracks overlap the start of the calendar year which affects the current year values. One track may be biased heavily with the current year and others may not.

Interesting stuff, I can only wonder what the rationale is behind each of the ‘ways’ ...no doubt optimization going on...

...for myself, I like the metric of THIS YEAR’s earnings to compare the runners with, but ever since this thread started I can now see how imbalances in earnings reports could occur at different stages of the year...I wish there was something like giving weighted points accorded with purse disbursement standards for each 1st, 2nd, third placing for a given period...then add those up...then take the total purses earned number and divide by that placing number...would that make sense at all...?

[Augenj]

Originally Posted by VigorsTheGrey

Interesting stuff, I can only wonder what the rationale is behind each of the ‘ways’ ...no doubt optimization going on...

...for myself, I like the metric of THIS YEAR’s earnings to compare the runners with, but ever since this thread started I can now see how imbalances in earnings reports could occur at different stages of the year...I wish there was something like giving weighted points accorded with purse disbursement standards for each 1st, 2nd, third placing for a given period...then add those up...then take the total purses earned number and divide by that placing number...would that make sense at all...?

It would but making sense of horse racing is sometimes like a dog chasing its tail and never catching it. Handicappers tend to look for profitable conditions that can be applied track to track. In all the data mining I've done over thousands of races at eleven major tracks, I haven't found very many factors that can be useful that way. Each track seems to have its own ecosystem of jockeys, trainers, surfaces, etc and is quite different if you try to use SA profitable conditions at BEL, for example. That's why I make analysis runs to develop a unique profile of each track and use that profile to calculate races at that track only.

[VigorsTheGrey]

Originally Posted by Augenj

It would but making sense of horse racing is sometimes like a dog chasing its tail and never catching it. Handicappers tend to look for profitable conditions that can be applied track to track. In all the data mining I've done over thousands of races at eleven major tracks, I haven't found very many factors that can be useful that way. Each track seems to have its own ecosystem of jockeys, trainers, surfaces, etc and is quite different if you try to use SA profitable conditions at BEL, for example. That's why I make analysis runs to develop a unique profile of each track and use that profile to calculate races at that track only.

That makes total sense to me...each track needs to be evaluated by its own metrics and conditions...just like each race needs to be evaluated as the unique event that it will eventually become...this requires enormous effort and concentration, memory, imagination, etc...

..I think systematizers definitely understand that and invoke the use of systems for that very reason...systems are, in effect, short-cuts of sorts that increase efficiency in processing data...but as you know, they have their own limitations and are susceptible as well...

But did you ever wonder, for example, what would result if you begin each logical year “blank slate” such that the first race of the meeting HAD NO DATA...FOR JOCKEYS and TRAINERS, yet, and that from then on...the JOCKEY stats...would begin compiling....in a sort of “what have you done for me lately” method...right until the end of the meeting.

[Augenj]

Originally Posted by VigorsTheGrey

But did you ever wonder, for example, what would result if you begin each logical year “blank slate” such that the first race of the meeting HAD NO DATA...FOR JOCKEYS and TRAINERS, yet, and that from then on...the JOCKEY stats...would begin compiling....in a sort of “what have you done for me lately” method...right until the end of the meeting.

For me that would require running an analysis every week which I have on my to do list to check out. Analysis runs are just for calculating ratings for the horses. I'm already doing this for the conditionals in a private system, conditionals being how does the top horse do in a dirt maiden claimer, and about 800 more. I'm not sure how the statistics would look but that's another issue.

[VigorsTheGrey]

Originally Posted by Augenj

For me that would require running an analysis every week which I have on my to do list to check out. Analysis runs are just for calculating ratings for the horses. I'm already doing this for the conditionals in a private system, conditionals being how does the top horse do in a dirt maiden claimer, and about 800 more. I'm not sure how the statistics would look but that's another issue.

Thank you for your responses...I enjoy conversing with you...I learn a lot...you are very kind...talk to you some time again soon...good night...!