Conflicting Trainer Stats

[upthecreek]

Yesterday I got into a discussion with TVG's Rich Perloff(actually he gave me an on air lecture),but anyway to the point Parx R6 he had a "mutual fund" play #3 trained by John Servis The DRF had Servis 29%/roi of $5.60 when going from sprint/rt I pointed out that Brisnet had Servis only 9% 1st at a route Now what's the difference? The only way the horse can be "1st rt" is if he ran in a sprint last out I suppose it could include 1st time starters Why such a difference? Maybe somebody can explain

[Elliott Sidewater]

Simple, first route is the first time EVER at a route. Sprint to route includes all incidences of a horse routing today whose previous race was a sprint. First route could be a first time starter, but also counts horses with all sprints on their record competing in a route today. First route will have a smaller sample size than sprint to route if compared over the same time period.

[upthecreek]

Thnx I was sitting here thinking about it and that dawned on me that it didn't have to be the horses first rt attempt It was funny though that Bris didn't list the sprint/route category,only 1st route

[camourous]

Servis has one in the 6th race today at PARX that is going Sprint-Route and Bris doesn't give the Sprint-Route stat either

[eqitec]

The various data providers use differing time frames when presenting their statistics, presumably due to contractual issues with Equibase, which is one of the flaws of Equibase being a profit-making service.

Also, there is a lot of unnecessary nonsense presented in the stats, such as presenting a trainer stat for "repeat winners" of the new trainer when a horse is a 1st after claim and the new trainer's repeat winner % is not relevant to the horse. There are many such examples.

[VigorsTheGrey]

The various data providers use differing time frames when presenting their statistics, presumably due to contractual issues with Equibase, which is one of the flaws of Equibase being a profit-making service.

Also, there is a lot of unnecessary nonsense presented in the stats, such as presenting a trainer stat for "repeat winners" of the new trainer when a horse is a 1st after claim and the new trainer's repeat winner % is not relevant to the horse. There are many such examples.

I have a side question regarding your website: Is there anyway to see the trainer stats you show on a daily basis? Do you have an archive? Otherwise I'll just print them up daily and keep them in a binder....Vig.

[cj]

The various data providers use differing time frames when presenting their statistics, presumably due to contractual issues with Equibase, which is one of the flaws of Equibase being a profit-making service.

Also, there is a lot of unnecessary nonsense presented in the stats, such as presenting a trainer stat for "repeat winners" of the new trainer when a horse is a 1st after claim and the new trainer's repeat winner % is not relevant to the horse. There are many such examples.

I don't think this is the case. Pretty sure the provider can pick any time frame they like. It is a choice by the provider, not Equibase. Ubercapper from the board here can correct me if I'm wrong.

[eqitec]

Assuming you are correct, what would their logic be in using different time frames? It only creates confusion when players see different statistics for the same parameter. Do they think their time frames are superior to their competitors?

[RonTiller]

We can publish whatever statistical time period for which we have Equibase data. The developers HDW works with want a different mish mash of statistics for various time periods so we make statistics every night for everything from last 7 years to last 365 days (the default time period we use on many of our own reports) to last 180 days to last 90 days to last 30 days to even last 7 days (hot jockey or hot trainer) to current meet to previous meet and more in between.

I am not making any pronouncements concerning which time periods are "valid" or "representative" or most predictive for today's race. That itself is a big can of worms everybody here can thrash around.

Regarding the time period for statistics, here is a summary of the sprint to route statistics for Jason Servis for various time periods.


Period Starts Wins Win %
Last 60 days 5 0 0%
Last 180 days 15 0 0%
Last 240 days 22 3 13.64%
Last 365 days 35 6 17.14%
Last 730 days 69 9 13.04%
Last 1000 days 92 12 13.04%
Last 2000 days 162 31 19.14%
Last 3000 days 220 41 18.64%
Last 4000 days 287 49 17.07%
Last 5000 days 333 57 17.12%


All these statistics are as of right now. Suppose we ran these statistics going into the Sep 12, 2015 race at MTH, the start of a 16 race losing streak with sprint to route. Pick some different time periods and you get a smorgasbord of results:


Period Starts Wins Win %
Last 30 days 3 2 66.67%
Last 60 days 6 2 33.33%
Last 90 days 9 3 33.33%
Last 365 days 38 7 18.42%
Last 1000 days 93 18 19.35%


Divide it up further by track and you get more variation. Here's a selection going back from today's date:


Period Track Starts Wins Win %
Last 365 days AQU 9 0 0%
Last 365 days MTH 12 4 33.33%
Last 1000 days AQU 29 2 6.9%
Last 1000 days MTH 23 6 26.09%
Last 1000 days GP 11 1 9.09%


Make of this what you will.

Ron Tiller
HDW

[cj]

Assuming you are correct, what would their logic be in using different time frames? It only creates confusion when players see different statistics for the same parameter. Do they think their time frames are superior to their competitors?

I would guess the majority of people are using only one provider. I have no idea how the parameters are chosen or why, even with my employer TimeformUS. I never bothered to ask. I think the numbers we use are fine. If you are getting vastly different numbers using different parameters, maybe, just maybe, the stats don't have much validity in the first place.