Congratulations to Thoro-Graph customer Steve Wolfson, Jr. for his win in the DRF/NTRA National Handicapping Championship. Also to Rick Lang, who was in second place going into the last few races.
A tip of the hat to Friedman, not just for his win in the other tournement, but for having the good grace to say that luck played a part-- which it always does in contests using small samples, no matter who wins.
Friedman\'s comments on the Ragozin site about use of trainer profiles make it clear that he (and undoubtedly others) misunderstand how these tools should be used, and that this is a subject that has been woefully neglected here. There are some pretty sophisticated handicappers out there, so who wants to get the discussion going?
Going from an earlier string, if you attack a race by asking \"what do I like about this race?\", I use the trainer profiles by asking \"what does this guy do well?\"
The limits that Friedman alluded to in his post are obvious. Trainer angles are not a factor in every race. You have to use your judgement in weighing sample size, and the stats may not be as specific to a situation as you would like. However, there are situations where you can\'t help noticing that a 15-1 borderline contender is second-off-a-layoff for a very strong second-off-a-layoff trainer.
I look for areas where the win/ITM% and ROI substantially exceeds the trainer\'s overall win/ITM% and ROI. I\'m not looking for a $0.20 increase in a category on overall ROI; I\'m looking for categories that are at least $1.00 over.
One comment on sample size. If the sample category includes 50 starts, and overall the stats are for 500+ starts, that\'s a small sample. If the guy has 12 wins in that category, and his overall win% is 10%, that small sample indicates a strong category for that trainer in my book. The situation may not arise every day, but when it comes up, the guy really hammers it. If the price is right and the horse has a chance with a forward move, I\'m not going to say \"that sample\'s too small\", I\'ll take a swing on the chance he\'s going to hammer it again.
On the other hand you also have to be careful in giving a trainer too much credit for being a LITTLE bit stronger in a category. Sometimes the price is right and this will lead you to talk yourself into something that isn\'t there. Trainer stats always work better if applied to horses that have a realistic chance on the figs; I don\'t use them to bet a horse that needs a 10 point jump. HP
I sometimes use them when a horse needs a 10 pt jump in one of the categories they are most useful for, namely 2nd lifetime start. There seem to me to be an awful lot of good, underbet trainers who are exponentially better 2nd time out at very playable odds. I also rely on them for 1st & 2nd time lasix, as well as when a horse is running back on 10 or less days rest, as the stats show that some trainers excel in these categories. Otherwise, I pretty much focus on the categories which apply to the particular horse & race, & those percentages influence the percentage chance I assign to the horse running a particular number.
Very sorry to report that OPM suffered an untimely death in the family & couldn\'t make it to GG last Sat. My sincere condolences, OPM, in this difficult time.
Sans OPM, I had to rely on a taxi driver to try to find the pic n pac close to the track, which seemed to both of us to be located between an expressway & the water. We didn\'t find it, but we did run into the very large Berkley contingent from the 200k anti-war protesters, none of whom was willing to also take up the cause I suggested, namely a reduction in confiscatory takeout rates. For some reason, they seemed to think that the issue was not that important given everything else that is going on in the world right now.
The mention of Pic N Pac brought back fond memories of a misspent youth cutting school and heading to GGF. The anti war march reference made me think it was 1971 all over again.
Next time you\'re there, Pic N Pac is on the corner of San Pablo & Gilman, at least it still was there the last time I visited the old home town @5 years ago.
There is a lot of trainer information provided by TG and many ways to use it but here are some very basic ways I use it.
I only play on the weekends and I like play a lot of different tracks from across the country depending on the cards and the weather. There is no way I can be familiar with all the trainers. With the TG trainer stats, I can quickly tell if a trainer seldom wins or conversely if a trainer frequently wins. I typically steer clear of the former and often try to beat the latter (non-contenders or marginal contenders from top barns are typically overbet and can help make a race playable).
At the start of a new meet, I like to look at how the trainer has done at that particular track in the past relative to the overall stats. Some trainers point to certain meets and others use some meets just to \"condition\" their horses for another meet down the road.
For special applicable circumstances, such as layoffs or first-time-turf, I like to see how the trainer does for that category relative to their overall percentages. A big relative difference can be meaningful even if the absolute percentage is not that high. For example, an overall 5 pct trainer (very bad) that wins 15 pct of the time off a layoff (not great but not bad either) might provide some \"hidden\" value with such horses. In particular, if it is up against another layoff horse from a top barn (say 28% overall winners) that has a poor relative percentage on layoffs (say 15%). Both have the same absolute percentage with layoff horses but one tends to move them up while the other does the opposite. This all assumes that decent sample sizes are involved.
Speaking of sample sizes, they can be somewhat problematic; however, a sample size of 20 or more is fairly reliable, particularly if you look at ITM as well as win pct. Also, even small samples can sometimes be meaningful. For example, if a trainer has started only 10 horses \"First Lasix\" and had only 1 winner and 3 ITM, it would not tell you very much but, if that same trainer had 8 winners and 10 ITM out of 10 starters, you might legitimately infer a strong trend despite the small sample size.
I have to close by getting on my soap box for a few sentences. I would much rather see how often the trainer gets a new top or has his horse run an X off a layoff than how often his horse wins or runs ITM. The former tells me if the trainer tends to send them out off the bench ready to fire while the latter is as much about placement as fitness. The same goes for many of the other categories. For example, if I knew that off layoffs a trainer\'s horses run a top or better 50% of the time, less than 5 pts worse than their top 45% of the time and X only 5% of the time (50/45/5) I could easily factor that into my analysis of a particular race. Something like that would be a nice enhancement to the already helpful TG trainer stats.
Chris
P.S. Congrats to both Steve Wolfson and Len Friedman. I Also send my sincere condolences to OPM.
Finally the board has gotten around to discussing the much dismissed trainer factor and how to use trainer info with thorograph. Let me first say, and Jerry knows this, that I use Thorograph and have used them for quite awhile now. Last year I also started incorporating a set of proprietary stats that Ed Bain sells. After struggling with the new information and melding it with speed figures, I put down the Thorograph numbers and concentrated solely on trainer statistics. After learning how to winnow down the automatic Ed Bain 4 + 30 (at least 4 wins and a 30% win percentage in say, 1st start off a layoff in a sprint) I then incorported back the Thorograph figures. A few months later, actually fall of 2002, my win % has never been better. I use the Thorograph to eliminate the losers from Bain and conversely use Bain to eliminate the losers from Thorograph. I\'m not suggesting this for anyone else. I just know it works for me and the trainer stats that Jerry provides with his product make Thorograph a much better product on the whole than the competition. The Bain stats give the insight that Chris was talking about that helps point out many, many live longshots. It\'s not always a happy marriage between the 2 stat packages but when they are good together, they are very good.
Michael
Thanks you guys for thinking about me. She will be missed. Now, what are we going to do about the million?
Jerry:
Your trainer profiles are GREAT! But they would be even greater if you programmed the following queries:
1. MSW to mclm. That is... dropping from Maiden Special Weight races into maiden claimers.
2. First start at route. That is... debuting at 1 mile or over.
John T Chance
couple comments:
1. Trainer A has 111 starts this past year and is winning at a 16% rate. For turf races over the last several years, she has 76 starts with a 8% win rate. Obvious right?
Surprise! This is NOT a statistically significant difference at a 90% confidence level. [difference in proportions test]
2. You can increase the value of trainer trends by keeping a notation of significant trends with the past performance. As an somewhat obvious example: We have two second time starters that both ran the same effort last out (however you measure effort.) One trained by Zito. One trained by Pletcher. Let\'s say they have similar 2nd out statistics. It might be beneficial to know that Zito is 4% 1st out and Pletcher is 22% first out. That is something don\'t get with Bris, Drf, or Rags.
3. ROI, as it relates to trainer stats, is worthless.