Trainer Statistics

[QuarterCrack]

To those of you who compile your own trainer stats, what things in particular do you keep track of? I've started keeping my own records of trainers at MNR and I basically just take the stats from the winners of each race (do others use races besides wins? I don't know - I'm sure there's some valuable info beyond just the winner...) and I make note of such things as number of days since last race, class of previous race, class of current race, workout info since last race, jockey, odds, equipment, medication, and owner. I was curious if there were other specific things that other people keep track of when coming up with their own trainer stats. It's always interesting to hear others' take on things.

[chrisg]

First off, always use abbreviations! :D

For winners I tend to note-

Layoff 1-4(How many days to determine that is up to you)

Claim 1-4

Ship 1-4

Non-claim Trainer Change 1-4

If I see it was a pattern of sprint-sprint>route or route>sprint-sprint or turf route> dirt sprint, I'll note that.

I'll also write down the figures earned for the above mentioned pattern.

And of course, days since last raced.

Also, I've found it's not worth my time to keep track of every winner. It's best for me to focus on 1-3 trainers or certain types of race or patterns.

[Turfday]

I'm new to the board and perhaps the site that I run can be of some assistance. Using Equibase data from Jan. 1, 1995 forward, Turfday.com offers trainer, jockey, trainer-jockey combination, sire and posts & pace information for every track in the United States and Canada.

Our database covers every single race run in North America and Canada from that year forward.

Your question was on trainer stats. The following trainer stats are readily available with the following search options: since Jan. 1, 1995, last 1,095 calendar days (three years) from the last update of our data (updated weekly), last 365 days (one year) from last update of our data.

Another search option is "trainer-track" so you can search how any one trainer's horses perform at a specific track using the above time period search options.

Here are the categories of stats provided under trainer:

OVERALL
SPRINTS
ROUTES
TURF
OFF TRACKS
1ST-TIME STARTERS
45+ DAY LAYOFF
120 + DAY LAYOFF
SECOND START OFF LAYOFF
FIRST OFF A CLAIM
FIRST START WITH TRAINER
FIRST TIME ON TURF
OVERALL WITH FAVORITES
SPRINT FAVORITES
ROUTE FAVORITES
TURF FAVORITES
SPRINT TO ROUTE
SPRINT TO ROUTE (60 DAY CYCLE)
ROUTE TO SPRINT
ROUTE TO SPRINT (60 DAY CYCLE)
DIRT TO TURF
DIRT TO TURF (60 DAY CYCLE)
TURF TO DIRT
TURF TO DIRT (60 DAY CYCLE)
FIRST TIME LASIX
SECOND TIME LASIX
BLINKER CHANGE
BLINKERS ON
BLINKERS OFF
WITH 2-YEAR-0LDS
WITH 3-YEAR-OLDS
WITH 4-YEAR-OLDS & UP
WITH FILLIES & MARES
WITH COLTS & GELDINGS
HOT OR COLD LAST 30 CALENDAR DAYS
HOT OR COLD LAST 90 CALENDAR DAYS

I hope this helps. There are a handful of paceadvantage.com posters currently using the site and I hope they will post a report in the next couple of weeks.

[schweitz]

Quartercrack, Get a copy of "Handicapping Trainers" by John Whitaker and prepare to do alot of work.

[Turfday]

Using a national bell curve, we also provide an interesting search feature where users can look up the BEST and WORST in each of those 36 categories....all the trainers that graded A- or better and all the trainers that graded D or worse. The rankings are most enlightening.

As one minor example....using a minimum sample size of 30 turf favorites in the last 1,095 calendar days since Feb. 14 (date of our last weekly update), the following trainers were at the bottom (worst) when there horses started as a turf favorite:

Eric J. Kruljac (5/31) 16%
Mark Casse (12/75) 16%
John C. Kimmel (10/58) 17%
William P. White (9/50) 18%
Layne S. Gilforte (6/34) 18%
Mark Hennig (10/51) 20%

Hennig (45%) and Casse (52%) also have a particularly LOW in-the-money percentage with turf favorites, a great stat to know as a possible "bet against" when they start a turf favorite.

Conversely, as a comparison, the best trainers with turf favorites (minimum sample size 30) are winning 40-45% of those races topped by Wally Dollase with a whopping 53% (21/40) with a get this..85% in-the-money pct.

[MV McKee]

Quartercrack,

It looks like you are playing just one track (Mnr), so my advice to you would be to be subjective in your trainer analysis.
Rather than simply track the attributes of winners, I look for characteristics of horses that "fooled" me, or ran much better or worse than I expected he would. I have had horses that ran last run much better than I expected them to (for instance dueled on the pace with a horse I could not envision them warming up), and I have had horses who finished in the money who ran much poorer than I expected. I try and notate each of these efforts, and then notate anything indicated in the available pre-race information that may have been a harbinger of the resultant good/poor effort.
My personal opinion is that trainer data is a poor candidate for traditional statistical analysis, there are simply too many dynamics involved, and many of these lurking variables are simply too influential to ignore.
But if you opt to go that route, there are a couple of important things to note when you are looking at trainer statistics, both overall and in catgories.

If you are looking at trainer performance in a particular category (for example favorites) you really need to go well beyond simply looking at the trainers win percentage. Take the following scenario:
Imagine that all trainers are homogenous, and that over eternity all of them win with favorites at the standard 32% rate. Now lets say that you are looking at 1000 of these homogenous trainers nationwide, and lets pretend that each of these trainers has started 50 favorites. Out of that 1000 trainers, it would actually be highly UNUSUAL if at least 15 of them did NOT win with at least 50% of their favorites, and it would be even more UNUSUAL if at least 20 of them did NOT won with less than 17% of their favorites.
The point here is that it is difficult to strike a balance between creating a category broad-based enough to get a sufficient sample and still be meaningful. In the case of a favorite (where the expected win rate would be around 32%) a win percentage above 50% or below 17% would be expected in about 4% of the trainers sampled.

This leads me to overall trainer stats (not categorized). In this case you are dealing with an expected win rate of 12% (probably less at Mnr since I believe thay have larger avergae fields).
One thing to note here is that a trainer above 20% is almost certainly unusually good (provided you are looking at 80 races or more). However, a trainer a 7% or so is not necessarily bad. Because the success rate (a win) is so far away from the failure rate (a non-win) it would be not be highly unusual for a true 12% (average) trainer to only win 6 races out 100. However, it would be highly unusual for that same trainer to win 24 out of 100 races.

I probably got a little wordy, but my point is that you should be wary of an apparent indication if it is negative. By the time a trainer is 1 for 70 or even 25 for 90, everyone has pretty much caught on. Categorized stats are fairly common currency, and maybe 1 out of 10,000 of them are actually statistically significant. If you were playing upwards of 4 or 5 tracks a day, it may be the way to go. But if you are playing just one or two tracks, my suggestion would be to notate any horse that ran better or worse than you expected it to..simple as that.

[Turfday]

To follow up on your excellent post, and using the gauge of "80" minimum samples as you mentioned, I have provided below a listing of the trainers w/favorites who performed the WORST with data from Feb. 14, 2001 thru Feb. 14, 2004. Generally, it's harder to find large sample sizes with trainers who perform poorly with favorites since, as said, the "public is on to them." However, as you can see below, several of these have pretty big sample sizes.


--------------------------------------------------------------------------------
Bottom Trainer Records
Bottom 3 Year (sprints, routes, turf favorites combined)

Trainer Starts Wins Win%

Sanders Brent 80 13 16%

Doyle Michael J. 89 15 17%

Attard Tino 105 25 24%

Randazzo Teddy C. 92 20 22%

Vivian David A. 87 17 20%

Donk David G. 98 21 21%

Semer John R. 150 36 24%

Shanyfelt Douglas E. 153 35 23%

Tanner Lynda R. 89 18 20%

Sheppard Jonathan E. 242 57 24%

Conversely, using your gauge of "80" sample sizes in that same time frame, here are the best trainers with favorites (sprints, routes, turf favorites combined):

Trainer Starts AOPS Wins Win%

Hanson Jim 97 55 57%

Morgan James E. 119 56 47%

Von Hemel Donnie K. 299 143 48%

Diaz, Jr. Julio 148 67 45%

Bennett Keith 143 74 52%

Payne Felix G. 110 54 49%

Engel Roger F. 108 53 49%

Allen, Sr. Ronald D. 141 68 48%

Hellman Leroy 181 84 46%

Wolfson Martin D. 126 58 46%

[andicap]

In NY,

I keep track of only certain wins, looking for short-term trends at a particular track and long-term ones.

Using a template I can just fill in (with abbreviations ike Chris said), I keep track of every trainer winner in detail

1) off layoffs
2) higher than $15 (lower in short fields)
3) shipping in

I should probably add layoffs plus 1,2,3, races, but my time is short. One day.

In another notebook, I keep the following categories to notice
trends at the track

1. Trainers that lose races at short odds (8-5 or so)
2. Trainers that lose off layoffs at 4-1 or less
3. Trainers that lose shipping in at 4-1 or less
4. Trainers that win off layoffs
5. Trainers that win shipping in.

For the latter, the level of detail is much less -- a timesaver. Again I wish I had more time (and the computer expertise to use a database) to get into more detail, especially with trainers that lose at short odds.

[SAL]

Interesting thread.... I've always been interested in trainer stats. The new version of DRF Formulator looks like it might be able to dig up a lot of different kinds of trainer stats. I'm looking foward to seeing what it can do.

[delayjf]

I think its useful to know when a trainer is "sending" and to echo MVMCEE I think it is important to look at a trainer when his horses run unexpectedly well. One problem I see is that valuable information is lost when looking a win stats only. If a trainer prepares a horse for an all out try off a layoff, but loses the race by a nose at 60-1, that fact is often overlooked by most handicappers.

When looking at a specific trainer pattern, ie like layoffs, keep in mind that he will lose more of these types of races than win, but take a look at the PP's for both his winner (including close up finishes) and his non-competitive runners to determine if his competitive horses show something (perhaps a workout pattern) that his non competitive horses don't. This can be a lot of work.

Since I've found that most longshots show a race in its PPs that would make it competitive in todays race, you still have to answer the question " if he runs his race, can he win". This is where speed/pace handicapping come in. You still need some process to determine how good the horse is.

This is often impossible and a lot of times you won't have all the answers as with the case of lightly raced horses and FTS. If the price suits you, sometimes you just have to bet on faith.

[Turfday]

You said, " One problem I see is that valuable information is lost when looking a win stats only. If a trainer prepares a horse for an all out try off a layoff, but loses the race by a nose at 60-1, that fact is often overlooked by most handicappers."

You're exactly right. Years back, I thought of that exact scenario: how do we handicappers give credit where credit is due to trainers (even jockeys, or even sires) when their horses OUTRUN their odds and perhaps with some frequency. Conversely, how about just the opposite?

I created the formula called "Betting Value Averages" that ran for years in Daily Racing Form's Simulcast Weekly edition and is available on www.turfday.com (sorry for the spam).

Using Equibase data, search options are available for last 365 days, last 1095 days or since Jan. 1, 1995.

My formula judges trainers in 36 categories mentioned in an above post in this thread. The formula gives credit where credit is due. Consistency BVA is based on finish position and field size. The Value BVA is based on closing odds, finish position and field size. The Overall BVA combines Value and Consistency.

Even better, the BVA's are formulated on a "rolling average" whereby what's happening now and most recently to identify current trends is weighed more heavily than what happened a year ago. However, to get a relevant sample size in the various categories we use the above search options. Grades are given on a national bell curve when 30 or more samples are available.

The end result is much like a baseball player's "slugging percentage."

The BVA formula accounts for the trainer (or jockey, or trainer-jockey combo) finishing second at a big price (60-1 as you used) in a relatively large field would get more credit than a trainer winning a race at 3/5 in a small field. Hence, it is more likened to a slugging percentage.

[delayjf]

I'm not saying long term trainer studies are bad. But it has occurred to me that perhaps the Trainer are reading the studies as well. More and more I'm finding that trainers who once their special methods becomes know, alter their methods.

I'm also seeing simular manipulations with horses with obvious well published bad trips. The next race is sub par at a low price, followed by a improved effort or win the second race back at a bigger price. Of course we may be looking at a performance bounce if the horse in question busted his butt attempting to over come the bad trip.

A few years back, I purchased Jim Mazurs trainer studies for Saratoga and DeLmar. Personally, I think his stuff is top rate, very detailed and documented, but I personally didn't porfit from them. On the day's I played, his angles simply didn't work. Perhaps others have had better luck. Perhaps trainers realize that they need to shift gears in order to bring home the Bombs.

[Brian Flewwelling]

Talk about the good timing! I logged on to take a break from revising my Trainer analysis routines :D

I don't have a fancy stat like BVA (yet) but i do one thing for all conditions analyzed (much lke above lists --- from 1999 only)

for each factor, i sum the number of races, n wins, nITM, and the Winning Payouts. I also sum the probability of winning based on both the Odds at Post and Number of Runners. The latter allow me to determine Impact or, as i call them, Efficiency stats to go along with Win% etc

What i get is a reading of true efficiency in terms of how the betting public percieves that trainer. Looking at the number of runners gives credit for field size, but also tends to tell the most about quality of stock.

Things i look for is the Eff wrt odds (abbrevs ok chrisg?) for favs i am trying to bet against, and long-shots i would like to bet.

Thanx for the ideas. Keep this thread moving

Brian

[Turfday]

We are on the same track here. Obviously, the public's perception of certain trainers, trainer-jockey combos, certain sires, perhaps even certain posts (depending on a possible track bias) can UNDULY influence the odds thereby creating overlays on other horses. The public loves to overbet the obvious.

Right now, in Southern California, the "flavor of the month" which has dragged on now for months, is overbetting trainer Jeff Mullins horses, as an example.

Problem is, he's winning (or on the board) at such a high clip and, except for first-time starters, he has no real weaknesses. So even the contrarians are finding it tough to play against him without getting burned.

My BVA formula is a direct attempt to "ferret out" the value trainers, jockeys and trainer-jockey combinations and sires by giving credit for in-the-money finishes using odds, field size and finish position within a given race. Conversely, it also will point out the undervalued trainers, jockeys, trainer-jockey combinations and sires.

I will point out a remarkable example in this thread a little later on so stay tuned.

[LOU M.]

Take the top dozen or so trainers at your track which should account for a fair amount of all winners ,spread sheet them to find the differences between their winners and losers and find the best way to bet them.The variables are endless,both their methods and how to bet them, but they're there.

Lou M.

[Speed Figure]

Bob Selvin,

Are you a software or drf user?

[Turfday]

Equibase is our data provider. We get continual feeds throughout the day and we update the voluminous categories on a weekly basis. We also post race results "as they happen" direct from Equibase. Take the FREE tour of the site. Check out our ad in this Saturday's DRF. Your comments would be appreciated.