Performance Indicators

[traynor]

Most bettors tend to focus too much on the finish in the last race, and discount the chances of entries who appear to have "faded in the stretch" or "quit late" in their last race. Research indicates it is far more likely that a poor finish in its last race may be a strong indicator of improvement if the entry ran well early, but was eased in the final stages of its last race to "save it for the next race." In short--the race in which an entry runs well to the three-quarter mile position (3C) or stretch position (Str) in its last race, but finishes up the track, is a stronger indicator of improving performance than an entry that held on to be close up at the finish or in the money in its last race.

PACE RACES ALL TRACKS CloseToPace3CLast 11225 3772 33.60 %
PACE WINNERS ALL TRACKS CloseToPace3CLast 1368 627 45.83 % IV 1.36
PACE PLACERS ALL TRACKS CloseToPace3CLast 1397 559 40.01 % IV 1.19

TROT RACES ALL TRACKS CloseToPace3CLast 3814 1295 33.95 %
Trot WINNERS ALL TRACKS CloseToPace3CLast 520 230 44.23 % IV 1.30
Trot PLACERS ALL TRACKS CloseToPace3CLast 518 197 38.03 % IV 1.12

PACE RACES ALL TRACKS CloseToPaceStrLast 11225 3888 34.64 %
PACE WINNERS ALL TRACKS CloseToPaceStrLast 1368 652 47.66 % IV 1.38
PACE PLACERS ALL TRACKS CloseToPaceStrLast 1397 584 41.80 % IV 1.21


TROT RACES ALL TRACKS CloseToPaceStrLast 3814 1306 34.24 %
Trot WINNERS ALL TRACKS CloseToPaceStrLast 520 238 45.77 % IV 1.34
Trot PLACERS ALL TRACKS CloseToPaceStrLast 518 204 39.38 % IV 1.15

PACE RACES ALL TRACKS CloseToPaceFinLast 11225 3405 30.33 %
PACE WINNERS ALL TRACKS CloseToPaceFinLast 1368 454 33.19 % IV 1.09
PACE PLACERS ALL TRACKS CloseToPaceFinLast 1397 442 31.64 % IV 1.04

TROT RACES ALL TRACKS CloseToPaceFinLast 3814 1182 30.99 %
Trot WINNERS ALL TRACKS CloseToPaceFinLast 520 162 31.15 % IV 1.01
Trot PLACERS ALL TRACKS CloseToPaceFinLast 518 173 33.40 % IV 1.08

What the numbers mean: In 11,225 Pace races at all tracks, 33.60 % of all entries were in the lead or close to the lead at the three-quarter mile position in their last race. However, 45.83 % of the winners of their next race were in the lead or close to the lead at the three-quarter mile position in their last race (IV 1.36).

However, In 11,225 Pace races at all tracks, 30.33 % of the entries were close to the pace at the finish of their last race, while 33.19 % of the winners of their next race were in the lead or close to the lead at the finish in their last race (IV 1.09).


("CloseToPace" is defined as being in the lead or less than 3 lengths behind the leader at the positions indicated.)

[pandy]

Good stuff. I would think thoroughbreds have similar stats.

[traynor]

Good stuff. I would think thoroughbreds have similar stats.

They do. The problem is that the rights to the modeling software I use for thoroughbreds was purchased by a private group of bettors, and I am prevented by the terms of that sale from publicly posting the output. I can use it for my own wagering, but that is it.

If anything, the stats for thoroughbreds even more strongly suggest that the finish in the last race--whether "up close" or a finish in-the-money--is a poor indicator of winning potential. That is, there is not much difference between the finish of "all entries" and the finish of "winners" in the next outing.

One of the things I have found intriguing (following your recent postings on the emphasis of early speed in harness races) is that it is possible to handicap subsequent performance strictly on the basis of early speed in previous races. That is, handicapping a race almost as if it ended at the three-quarter mile position or stretch position, rather than the finish.

That is consistent with the statistics above, that indicate tossing the finish position into the mix when analyzing pace, speed, and potential improvement may be misleading, and better comparisons can be made by ignoring the finish. The same is true of thoroughbreds.

[traynor]

I want to be sure to credit Bob Heyburn, who wrote Fast and Fit Horses, for the original idea of handicapping races by ignoring the finish position or beaten lengths. It is every bit as applicable to harness races as Heyburn found it to be for thoroughbreds.
http://www.amazon.com/Fast-Fit-Horses-Bob-Heyburn/dp/089709168X

[traynor]

An interesting corollary to the above is the failure of many Sartin methodolgists to apply their pace methodology effectively to harness races (as well as failing to apply it effectively to thoroughbred races). Specifically, the use of "pace lines" to predict subsequent behavior is misleading when taken from "good races" (in which the entry won or finished close up).

It is more useful to consider pace lines as fitting the profile of the last race or next to last race of entries that won their next race, rather than profiled from winners. When so considered, it makes the (Sartin) percent early calculation in pace handicapping much more predictive (than when the percent early calculation is based on a profile of winners in winning races).

Simply stated, for comparison purposes, the important race is the race (or races) preceding a win--not the winning race itself. That presents a further advantage of being counter-intuitive. The mutuel prices on entries that seem to have faded (or quit) in the stretch are often generous.

Anyone can justify a winner after it has won. The trick is to locate that winner before the race is won. Paying less attention to finish position and more attention to the three-quarter and stretch positions may help do just that.

[badcompany]

What the numbers mean: In 11,225 Pace races at all tracks, 33.60 % of all entries were in the lead or close to the lead at the three-quarter mile position in their last race. However, 45.83 % of the winners of their next race were in the lead or close to the lead at the three-quarter mile position in their last race (IV 1.36).


("CloseToPace" is defined as being in the lead or less than 3 lengths behind the leader at the positions indicated.)

As Pandy noted, good stuff.

I would think that position, at least on 1/2 mile speed favoring tracks, plays a big role in these stats.

Many of the horses that faded did so as a result of being 1st, 2nd or 3rd over, but subsequently drew a better post or had a better position in their next start.

[pandy]

In thoroughbred racing in particular, you can make a solid case that the best ROI comes from handicapping races with little or no regard to where a horse finished. I agree about Heyburn's book, which I have.

[am1947]

Hello Traynor

Very Interesting numbers!
Some questions if you do not mind.

Using three lengths have you tried 2 lengths or 4 lengths?

I would think the field size would influence outcome ie 5 or 6 horse field
do the numbers include all size fields or a 5 or 6 or 7 size min or a 10 max?

Are these numbers for 1/2 , 5/8 , 7/8, and 1 mile tracks sizes combined?
Any/same significance if using just 1/2 or 5/8 or 7/8 or 1 mile?

Any filters for stating PP influence on the numbers? ie 7,8 on a 1/2 are poor
win %...

Any filters for 3 wide horses or parked out or 1st over ?

Have to say your posts are definitely food for thought. Always enjoy them
Hope your Holidys go well

Regards,
AM

[mrroyboy]

Guys I hate to disagree with all this but Harness Racing can not be compared with t breds.

A good race or a fit horse is the same today as it was in Al Stanley's day. Horse has to do something early, maintain that something through the middle of the race and then finish competitively. Now many horse used early with have an excuse for fading in the stretch. But that means it is an excusable race not a good race.

Also, there is a huge difference in harness racing from being 1st at any call or within 3 lenghts etc. You can't use the t breds upclose definiation because the sports are very different.

[1st time lasix]

imagine this theory coorelates well to Steve Klein's work in his speed analysis. It also might suggest to some that because there are so many races for cheaper inferior stock - that speed is predominant variable in the handicapping process when track conditions favor it. No surprise there I guess... I will suggest that on turf it is possibly less so....

[traynor]

As Pandy noted, good stuff.

I would think that position, at least on 1/2 mile speed favoring tracks, plays a big role in these stats.

Many of the horses that faded did so as a result of being 1st, 2nd or 3rd over, but subsequently drew a better post or had a better position in their next start.

I tried it, because it seems perfectly logical. However, the numbers don't hold up to the "logic." That means, when the data is layered down to compensate for (or to look for) park outs or post position discrepancies, it has less predictive value than when ignoring those factors. That is, the entries with excuses perform no better than "all horses"--they do not have the advantage of the entries that were eased.

What it seems to indicate is that a horse eased after being parked a couple of quarters, making a big move at some point, or overcoming an outside post (especially on a half-mile track) tried its best and came up short. Conversely, entries without those indicators that ran well to the three-quarter and/or stretch position and were then eased (without being overly exerting in the final run) seem to benefit considerably.

[traynor]

In thoroughbred racing in particular, you can make a solid case that the best ROI comes from handicapping races with little or no regard to where a horse finished. I agree about Heyburn's book, which I have.

I was actually surprised to see how consistently it holds up at different tracks. I have used it at The Meadows for quite some time, but it is not an anomaly specific to Mea.

[traynor]

Hello Traynor

Very Interesting numbers!
Some questions if you do not mind.

Using three lengths have you tried 2 lengths or 4 lengths?

I would think the field size would influence outcome ie 5 or 6 horse field
do the numbers include all size fields or a 5 or 6 or 7 size min or a 10 max?

Are these numbers for 1/2 , 5/8 , 7/8, and 1 mile tracks sizes combined?
Any/same significance if using just 1/2 or 5/8 or 7/8 or 1 mile?

Any filters for stating PP influence on the numbers? ie 7,8 on a 1/2 are poor
win %...

Any filters for 3 wide horses or parked out or 1st over ?

Have to say your posts are definitely food for thought. Always enjoy them
Hope your Holidys go well

Regards,
AM

The "AllTracks" designation is just that--it includes every harness meet in the US and Canada in the last whatever number of months (it is a current model) without regard to track size. I have it broken down by track, and there seems very little difference in the results for different track sizes.

I use other filters, and overfiltering tends to generate less useful numbers. As I mentioned above, compensating for park outs, PP changes, etc. seems to alter the results enough to diminish the value as a predictor. That may be because so many bettors look for "excuse races" that such are overbet in subsequent starts.

The "less than three lengths" designation is not arbitrary. I have used various values, and that is the most predictive. The "less than" part is important--it excludes entries three full lengths back at whatever position.

I ran similar tests using running positions, and the numbers are very close to those above. Specifically, an entry that is 3rd or better at the three-quarter or stretch position--ignoring the finish position--is much more likely to win the next race than an entry that finished in the money.

I didn't consider field size, because I have never found it to be especially useful as a factor.

[traynor]

Guys I hate to disagree with all this but Harness Racing can not be compared with t breds.

A good race or a fit horse is the same today as it was in Al Stanley's day. Horse has to do something early, maintain that something through the middle of the race and then finish competitively. Now many horse used early with have an excuse for fading in the stretch. But that means it is an excusable race not a good race.

Also, there is a huge difference in harness racing from being 1st at any call or within 3 lenghts etc. You can't use the t breds upclose definiation because the sports are very different.

Your disagreement is noted. However, the factors used to compare harness race performance and thoroughbred race performance are remarkably similar in results. They are both horses guided by humans. One walks fast, the other runs. Aside from that, they are much the same.

[traynor]

imagine this theory coorelates well to Steve Klein's work in his speed analysis. It also might suggest to some that because there are so many races for cheaper inferior stock - that speed is predominant variable in the handicapping process when track conditions favor it. No surprise there I guess... I will suggest that on turf it is possibly less so....

I have no idea who Steve Klein is. I think speed is important at all grades of race. All the class in the world will not get a horse to the wire in time if a faster horse gets there first. That is not an arbitrary statement--there are many. many examples of "cheaper, inferior stock" (according to some human definition) breezing past "higher class" (according to some human definition) entries to win. I don't think speed is the only factor (or pace, class, or quite a few other things are the only factors) in a horse race. I think speed is important at all grades of race, including turf races, along with a number of other factors.

[Maximillion]

They do. The problem is that the rights to the modeling software I use for thoroughbreds was purchased by a private group of bettors, and I am prevented by the terms of that sale from publicly posting the output. I can use it for my own wagering, but that is it.

If anything, the stats for thoroughbreds even more strongly suggest that the finish in the last race--whether "up close" or a finish in-the-money--is a poor indicator of winning potential. That is, there is not much difference between the finish of "all entries" and the finish of "winners" in the next outing.

One of the things I have found intriguing (following your recent postings on the emphasis of early speed in harness races) is that it is possible to handicap subsequent performance strictly on the basis of early speed in previous races. That is, handicapping a race almost as if it ended at the three-quarter mile position or stretch position, rather than the finish.

That is consistent with the statistics above, that indicate tossing the finish position into the mix when analyzing pace, speed, and potential improvement may be misleading, and better comparisons can be made by ignoring the finish. The same is true of thoroughbreds.


Very interesting stuff....gonna check out the book.

[traynor]

In thoroughbred racing in particular, you can make a solid case that the best ROI comes from handicapping races with little or no regard to where a horse finished. I agree about Heyburn's book, which I have.

There is an exception that you have probably noted in thoroughbred races, but I will add it for the benefit of those who may spend less time analyzing results. The exception to the general trend of ignoring the finish is the opposite--the "run out," followed by a stretch in distance in the subsequent start. For example, an entry finishes fairly well, but is not eased at the finish, continuing on for another furlong or so at a good clip. Really easy to spot if you are ontrack, or watch reruns carefully.

It is a common occurrence in six furlong races used as tighteners for a seven furlong race. The finish position or lengths behind is not as indicative as that the horse "continues running after the race is over."

[traynor]

For comparison purposes, these are the stats for The Meadows, current wagering model (12/19/2012):

Mea Pace CloseToPace3CLastAll 540 36.11 %
Mea Pace CloseToPace3CLastWinners 57 50.88 % IV 1.41
Mea Trot CloseToPace3CLastAll 406 35.47 %
Mea Trot CloseToPace3CLastWinners 50 38.00 % IV 1.07

Mea Pace CloseToPaceStrLastAll 540 35.56 %
Mea Pace CloseToPaceStrLastWinners 57 52.63 % IV 1.48
Mea Trot CloseToPaceStrLastAll 406 33.74 %
Mea Trot CloseToPaceStrLastWinners 50 32.00 % IV 0.95

Mea Pace CloseToPaceFinLastAll 540 27.04 %
Mea Pace CloseToPaceFinLastWinners 57 22.81 % IV 0.84
Mea Trot CloseToPaceFinLastAll 406 29.80 %
Mea Trot CloseToPaceFinLastWinners 50 24.00 % IV 0.81

Note the sharp drop off in predictive quality when the finish position is factored in. This same principle applies to thoroughbred races as well.

[am1947]

Traynor

Thanks for answering..

"I have it broken down by track, and there seems very little difference in the results for different track sizes"
I assumed you had but wanted to be sure I understood exactly what you were stating.

"That may be because so many bettors look for "excuse races" that such are overbet in subsequent starts."
I tend to agree with that as well ie cappin 101.

"The "less than three lengths" designation is not arbitrary"
Just wanted to sure of the significance


"I didn't consider field size, because I have never found it to be especially useful as a factor"
Not sure I agree about this. Was asking to see what size ot the data comes from the smaller fields . ie less then 10% 20% etc. I suspect that the lower the number of entrants the more chance to be within the less then 3 lengths.
Also would think the driving strategy is different then normal sized fields.
NOt sure if you ran just the smaller field sizes by themselves to see if the data is similar. Maybe you have and that is why you have never found it useful. Just trying to clarify a bit.


Again thanks for posting this stuff. It is always good to help with aiding the cappin thought process.

Regards
AM

[badcompany]

Here's another take. In a cookie cutter 1/2 mile track race, where the speed doesn't get harasssed early, the first over horse will pull at about the 1/2, and by the 3/4 pole will be about a 1/2 length off the leader. Provided the other horses on the outside keep up, the second over will be about 1 1/2 out; the 3rd over, 2 1/2. So, according to your criteria the third over horse will be "Close to the pace," but he really isn't. Why? Because for the 3rd over horse to close, it will have to go 4 wide and run a much greater circumference than the rail horses. From an mph standpoint, the 4 wide horse could be running as fast or faster than the inside horses; yet, still lose ground.

So, it might not be that the horse is being eased. It's just running a longer race.

Either way, I agree with your premise. Lengths beaten should be viewed within the context of the race, not just as raw numbers.

[traynor]

Here's another take. In a cookie cutter 1/2 mile track race, where the speed doesn't get harasssed early, the first over horse will pull at about the 1/2, and by the 3/4 pole will be about a 1/2 length off the leader. Provided the other horses on the outside keep up, the second over will be about 1 1/2 out; the 3rd over, 2 1/2. So, according to your criteria the third over horse will be "Close to the pace," but he really isn't. Why? Because for the 3rd over horse to close, it will have to go 4 wide and run a much greater circumference than the rail horses. From an mph standpoint, the 4 wide horse could be running as fast or faster than the inside horses; yet, still lose ground.

So, it might not be that the horse is being eased. It's just running a longer race.

Either way, I agree with your premise. Lengths beaten should be viewed within the context of the race, not just as raw numbers.

I agree--and your observation tends to validate my point. It is at a critical state of the race (the three-quarter or stretch position) that a competent driver (or jockey) knows how much horse he or she has left, and what can be done with it.

In the scenario above, because the 3rd over horse has so much ground to make up to be in contention for the win, it would not be surprising if the driver eased in the stretch run, rather than trying to pick up nickels and dimes for possible place or show, using up any reserves the horse might have, and tossing the chance of a possible better opportunity in the next race.

My reason for emphasizing the beaten lengths at the three-quarter and stretch positions is that it is a much stronger predictor of a next-race win than beaten lengths at the finish. Understand that it is not a standalone factor, other than its use in qualifying contenders. Considerable other analysis takes place, including FPS rates-of-speed in each quarter, with consideration of the effects of post position change, park outs, pace advantage, running styles, etc.--all the standard components of a comprehensive race analysis.

While the scenario you described is a good reason to excuse a poor finish, it is not really excuses that I am looking for. The best prices seem to be on those entries that were not obviously "used hard early" but that ran well to the late stages of a race, then seemed to fade. I understand that everything needs to be considered in context, but as a quick-and-easy qualifier of a contender that may substantially improve in its next race (performing better than its numbers indicate), considering beaten lengths at the three-quarter or stretch position in the last race is a strong indicator.

[traynor]

To give you an idea of how specific factors can be effectively combined to point out higher priced winners (for those who like to avoid "choking on chalk"):

PACE RACES ALL TRACKS OddsLessThan3Last 11225 1898 16.91 %
PACE WINNERS ALL TRACKS OddsLessThan3Last 1368 285 20.83 % IV 1.23
PACE PLACERS ALL TRACKS OddsLessThan3Last 1397 265 18.97 % IV 1.12

TROT RACES ALL TRACKS OddsLessThan3Last 3814 694 18.20 %
Trot WINNERS ALL TRACKS OddsLessThan3Last 520 108 20.77 % IV 1.14
Trot PLACERS ALL TRACKS OddsLessThan3Last 518 95 18.34 % IV 1.01

PACE RACES ALL TRACKS Odds3To6Last 11225 7059 62.89 %
PACE WINNERS ALL TRACKS Odds3To6Last 1368 605 44.23 % IV 0.70
PACE PLACERS ALL TRACKS Odds3To6Last 1397 741 53.04 % IV 0.84

TROT RACES ALL TRACKS Odds3To6Last 3814 2310 60.57 %
Trot WINNERS ALL TRACKS Odds3To6Last 520 238 45.77 % IV 0.76
Trot PLACERS ALL TRACKS Odds3To6Last 518 256 49.42 % IV 0.82

PACE RACES ALL TRACKS OddsMoreThan6Last 11225 2295 20.45 %
PACE WINNERS ALL TRACKS OddsMoreThan6Last 1368 391 28.58 % IV 1.40
PACE PLACERS ALL TRACKS OddsMoreThan6Last 1397 346 24.77 % IV 1.21

TROT RACES ALL TRACKS OddsMoreThan6Last 3814 776 20.35 %
Trot WINNERS ALL TRACKS OddsMoreThan6Last 520 157 30.19 % IV 1.48
Trot PLACERS ALL TRACKS OddsMoreThan6Last 518 132 25.48 % IV 1.25

Using a "double qualifier" of an entry that was less than three lengths back at the three-quarter or stretch position in its last race AND that went off at odds of EITHER less than 3 to 1 OR greater than 6 to 1 in that race can put you on to some solid choices at decent mutuels. Similarly, avoiding those entries with odds from 3 to 1 to 6 to 1 in the last race who would otherwise qualify by lengths back at the three-quarter or stretch positions increases the return.

[mrroyboy]

Sure If a driver has a fit horse he will usually try for the lead or 1st over etc. Horses that just sit on the rail are always suspect. Pandy's book calls these form points. I agree.

[dkithore]

Simply stated, for comparison purposes, the important race is the race (or races) preceding a win--not the winning race itself. That presents a further advantage of being counter-intuitive. The mutuel prices on entries that seem to have faded (or quit) in the stretch are often generous.


----------------------------------------ST Turn FP
24Nov12 7 f :472 :352 1:23 43 16/ 11- - - 10 1
3Nov12 6 f :351 :343 1:094 40 13/ 8- - - 8 2 - 2 1/4

Your data is saying, the 3Nov race is more imp. than the winning race?
But isn't this line predictive (up close within 3L) of the upcoming win outcome? Can you clarify?

[traynor]

Again, individual factors cannot necessarily be combined usefully. What happens in the real world is that putting two (or more) factors together essentially creates a third (or hundredth or whatever) factor in which both of the factors studied exist. In some cases, results are improved, but in others results are diminished.

In the particular case of combining a factor of close to the pace at the three-quarter or stretch position in the last race, combined with odds in that race of less than 3 to 1, the combination easily outperforms either of the factors considered in isolation.

PACE RACES ALL TRACKS CloseToPace3CLast+Odds 13116 1816 13.85 %
PACE WINNERS ALL TRACKS CloseToPace3CLast+Odds 1595 392 24.58 % IV 1.78
PACE PLACERS ALL TRACKS CloseToPace3CLast+Odds 1630 311 19.08 % IV 1.38

TROT RACES ALL TRACKS CloseToPace3CLast+Odds 4403 600 13.63 %
Trot WINNERS ALL TRACKS CloseToPace3CLast+Odds 598 133 22.24 % IV 1.63
Trot PLACERS ALL TRACKS CloseToPace3CLast+Odds 596 111 18.62 % IV 1.37

PACE RACES ALL TRACKS CloseToPaceStrLast+Odds 13116 1723 13.14 %
PACE WINNERS ALL TRACKS CloseToPaceStrLast+Odds 1595 382 23.95 % IV 1.82
PACE PLACERS ALL TRACKS CloseToPaceStrLast+Odds 1630 302 18.53 % IV 1.41

TROT RACES ALL TRACKS CloseToPaceStrLast+Odds 4403 580 13.17 %
Trot WINNERS ALL TRACKS CloseToPaceStrLast+Odds 598 135 22.58 % IV 1.71
Trot PLACERS ALL TRACKS CloseToPaceStrLast+Odds 596 110 18.46 % IV 1.40

PACE RACES ALL TRACKS CloseToPaceFinLast+Odds 13116 4259 32.47 %
PACE WINNERS ALL TRACKS CloseToPaceFinLast+Odds 1595 623 39.06 % IV 1.20
PACE PLACERS ALL TRACKS CloseToPaceFinLast+Odds 1630 614 37.67 % IV 1.16

TROT RACES ALL TRACKS CloseToPaceFinLast+Odds 4403 1264 28.71 %
Trot WINNERS ALL TRACKS CloseToPaceFinLast+Odds 598 209 34.95 % IV 1.22
Trot PLACERS ALL TRACKS CloseToPaceFinLast+Odds 596 186 31.21 % IV 1.09

And--again--focusing on the three-quarter positon and stretch position easily outperformed the finish position.

[traynor]

----------------------------------------ST Turn FP
24Nov12 7 f :472 :352 1:23 43 16/ 11- - - 10 1
3Nov12 6 f :351 :343 1:094 40 13/ 8- - - 8 2 - 2 1/4

Your data is saying, the 3Nov race is more imp. than the winning race?
But isn't this line predictive (up close within 3L) of the upcoming win outcome? Can you clarify?

My apologies for the confusion. When studying past results, many try to create a "profile" of winners, using the winning races. For example, Sartin advocates (and most other pace handicappers), create "energy distribution" figures that supposedly represent the "ideal race" for winners. Good idea, but they are looking at the wrong race. The profiles should be of the race preceding the winning effort, not the winning effort itself.

The whole notion of selecting "representative pace lines" is great, as long as there is an awareness that what one is looking for is the pattern of the race preceding the win, not the pattern of the winning race itself.

Example race:

Dec04 Mea 14000 4 3 3 2 2 1 113.21 28.71 27.70 56.40 83.90 113.21 32.50 67.00 14000

Nov19 Mea 14000 7 3 3 2 8 8 115.65 29.47 27.25 57.12 85.78 113.63

Nov13 Mea 12800 3 3 3 3 3 5 113.75 29.24 27.04 55.67 84.11 113.15

4 Mattacardle Wilbur Yoder 67.00 18.00 10.40
1 Caviart Key Aaron Merriman 4.20 3.60
9 Babe`s I Scoot Dave Palone 3.20

I am interested in the race of 12/04 because I bet on it. I care not one bit what kind of "energy distribution" or "percent early" or "pace profile" would be appropriate for that race. I only care that Mattacardle won. My interest--because I have to predict the finish to bet on it--is on the 11/19 race--in which it ran well to the three-quarter position, then seemed to "fold in the stretch." Getting odds of 32 to 1 on a horse that is a standout is something I wish would happen to me 20 times a day, and 50 times a day on the weekend. I'm working on it.

[eurocapper]

How is it different from usual trouble analysis of the last race (watching the race, which even Beyer advocates and copied from harness handicappers)? If there wasn't real trouble it may be an overlay.

For combining handicapping factors I don't see how the weighting can be done better than empirically, using for instance 1st and 2nd finishers of the races.

[dkithore]

My apologies for the confusion. When studying past results, many try to create a "profile" of winners, using the winning races. For example, Sartin advocates (and most other pace handicappers), create "energy distribution" figures that supposedly represent the "ideal race" for winners. Good idea, but they are looking at the wrong race. The profiles should be of the race preceding the winning effort, not the winning effort itself.

The whole notion of selecting "representative pace lines" is great, as long as there is an awareness that what one is looking for is the pattern of the race preceding the win, not the pattern of the winning race itself.

Example race:

Dec04 Mea 14000 4 3 3 2 2 1 113.21 28.71 27.70 56.40 83.90 113.21 32.50 67.00 14000

Nov19 Mea 14000 7 3 3 2 8 8 115.65 29.47 27.25 57.12 85.78 113.63

Nov13 Mea 12800 3 3 3 3 3 5 113.75 29.24 27.04 55.67 84.11 113.15

4 Mattacardle Wilbur Yoder 67.00 18.00 10.40
1 Caviart Key Aaron Merriman 4.20 3.60
9 Babe`s I Scoot Dave Palone 3.20

I am interested in the race of 12/04 because I bet on it. I care not one bit what kind of "energy distribution" or "percent early" or "pace profile" would be appropriate for that race. I only care that Mattacardle won. My interest--because I have to predict the finish to bet on it--is on the 11/19 race--in which it ran well to the three-quarter position, then seemed to "fold in the stretch." Getting odds of 32 to 1 on a horse that is a standout is something I wish would happen to me 20 times a day, and 50 times a day on the weekend. I'm working on it.

Traynor

Thanks for clarification. That makes sense. I recall Jim Lehan refers to this pattern except I haven't used it lately and forgot about it.

[traynor]

How is it different from usual trouble analysis of the last race (watching the race, which even Beyer advocates and copied from harness handicappers)? If there wasn't real trouble it may be an overlay.

For combining handicapping factors I don't see how the weighting can be done better than empirically, using for instance 1st and 2nd finishers of the races.

Trouble analysis focuses on a specific race. The factor(s) listed are generic, rather than specific to a given race. Rather than agonizing over the causes and effects (that one can get totally wrong more often than right) in a specific race, the advantage is in volume. I do both, but I am leaning more and more to ignoring the specifics of a given race, and instead relying increasingly on the patterns present in the various entries in the race. It takes some doing, having been immersed in trip handicapping and watching races for a number of years.

The basic problem with trouble analysis is that it is only really useful when you are the only one who sees it. That doesn't happen much any more, even at minor tracks.

Empirically is the only real way to determine the value of combined factors. Crunching numbers and applying "sophisticated statistical techniques" are pretty much useless. It makes analysis way more complex, especially because layering and chunking create new segments and groupings of races. And there is always the problem of a third or fourth or nineteenth variable that is not considerd that is a major influence on the results. Fortunately, I like data analysis, so it it is good fit for me.

[traynor]

Same pattern, at Dover Downs. Note that both the three-quarter position and stretch position are more predictive of subsequent winners in pace races than the finish position.

DD Pace CloseToPace3CLast+OddsAll 1111 15.48 %
DD Pace CloseToPace3CLast+OddsWinners 120 29.17 % IV 1.88
DD Trot CloseToPace3CLast+OddsAll 197 14.21 %
DD Trot CloseToPace3CLast+OddsWinners 27 18.52 % IV 1.30

DD Pace CloseToPaceStrLast+OddsAll 1111 14.94 %
DD Pace CloseToPaceStrLast+OddsWinners 120 27.50 % IV 1.84
DD Trot CloseToPaceStrLast+OddsAll 197 12.69 %
DD Trot CloseToPaceStrLast+OddsWinners 27 14.81 % IV 1.17

DD Pace CloseToPaceFinLast+OddsAll 1111 30.51 %
DD Pace CloseToPaceFinLast+OddsWinners 120 36.67 % IV 1.20
DD Trot CloseToPaceFinLast+OddsAll 197 29.95 %
DD Trot CloseToPaceFinLast+OddsWinners 27 37.04 % IV 1.24

[Ray2000]

Interesting IVs, thx for posting this

Not that you or anyone would use a single factor in making a bet, but if one did, do have any info on the return for these ClosetoPace winners?

[traynor]

Interesting IVs, thx for posting this

Not that you or anyone would use a single factor in making a bet, but if one did, do have any info on the return for these ClosetoPace winners?

Yes, but I think if I posted it, too many people would jump on it and ruin the value. There is enough information posted that anyone interested can work out the details for whatever tracks they play fairly easily.

[Ray2000]

thanks for the reply, sounds like a positive ROI and that's what I was curious about.

[Capper Al]

Interesting IVs, thx for posting this

Not that you or anyone would use a single factor in making a bet, but if one did, do have any info on the return for these ClosetoPace winners?

Don!t knock single factor horses. It can be a way to pick long shots.

[imofe]

Traynor,

It seems to me that you can improve your results even more if you only apply those rules ( close up and odds) to horses that you consider contenders in the race. For example, you have a horse that has little gate speed racing from the 8 hole tonight coming off a race where he drew inside and raced close to the pace at less than 3-1.

[mrroyboy]

I agree. You need to consider other factors especially at 1/2 mile tracks like Yonkers but other wise you did a great job Tray. Hope to see more

[traynor]

Traynor,

It seems to me that you can improve your results even more if you only apply those rules ( close up and odds) to horses that you consider contenders in the race. For example, you have a horse that has little gate speed racing from the 8 hole tonight coming off a race where he drew inside and raced close to the pace at less than 3-1.

That is where the rub comes in. I am currently testing various combinations of factors and pre-qualification of contenders. Initial results indicate that the more picky, the lower the return. Only considering solid choices generates red ink (everybody sees the advantage of that entry).

What I am actually looking for is something (or a number of somethings, preferably) that are counter-intuitive, and generate a solid profit. Something that most bettors would skip over looking for validation with multiple factors indicating a good performance. Multiple factors usually mean much lower mutuels--higher win%, but lower $$$. Because I do this for money (rather than bragging rights) I want the money.

An example--there are cases in which things that look "good" are really not. One of the most surprising is a change from a top trainer to a lesser trainer. That is actually profitable, most likely because many bettors would consider it a "negative."

Similarly, the fact that position/beaten lengths at the three-quarter and stretch position is much more predictive of a subsequent win than a close up or finish in the money.

I am discovering some interesting things. I will post them as I have time.

[traynor]

I agree. You need to consider other factors especially at 1/2 mile tracks like Yonkers but other wise you did a great job Tray. Hope to see more

Yonkers is tough. In some ways, and in others easier, because the drivers seem to have more skill and good sense than to make moves that will not pan out.

YR Pace LossInStretchLastAll 1105 24.89 %
YR Pace LossInStretchLastWinners 140 31.43 % IV 1.26
YR Trot LossInStretchLastAll 240 26.67 %
YR Trot LossInStretchLastWinners 33 27.27 % IV 1.02

YR Pace NegativeTrainerChangeAll 1105 20.36 %
YR Pace NegativeTrainerChangeWinners 140 28.57 % IV 1.40
YR Trot NegativeTrainerChangeAll 240 22.50 %
YR Trot NegativeTrainerChangeWinners 33 30.30 % IV 1.35

A good spot play for Yonkers. It doesn't come up often, but it hits a lot when it does:
YR Pace 3rdOB3C2Back+3rdOBStrLast+Gain2CToFin+ImprovedART All 1105 3.08 %
YR Pace 3rdOB3C2Back+3rdOBStrLast+Gain2CToFin+ImprovedART Winners 140 6.43 % IV 2.09

All the notations indicate "speed carried further" similar to the diagonal improvement lines recommended by Mike Fiore for thoroughbreds. Basically, an entry that was 3rd or better at the three-quarter position in the second race back, 3rd or better at the stretch position in the last race, gained position or lengths from the half-mile to the finish last, and had a better final time ("ART") time in the last race than in the second race back. Using "less than 3 lengths back" (rather than runing position) works as well, and pays even better.

As odd as it may seem, some pay well, especially if they slack off a bit in the stretch run. Note the "gain" is from the half-mile position, not the three-quarter. It points out a horse making a major move in the latter part of the race, but possibly eased in the stretch run. It seems to be a favored conditioning technique at Yonkers.

[traynor]

The spot play for Yonkers mentioned above:

YR Pace CloseToPace3C2Back+CloseToPaceStrLast+Gain2CToFin+ ImprovedART All 1105 11.76 % ROI 1.06
YR Pace CloseToPace3C2Back+CloseToPaceStrLast+Gain2CToFin+ ImprovedART Winners 140 17.86 % IV 1.52

YR Trot CloseToPace3C2Back+CloseToPaceStrLast+Gain2CToFin+ ImprovedART All 240 11.67 % ROI 0.31
YR Trot CloseToPace3C2Back+CloseToPaceStrLast+Gain2CToFin+ ImprovedART Winners 33 12.12 % IV 1.04

YR Pace 3rdOB3C2Back+3rdOBStrLast+Gain2CToFin+ImprovedART All 1105 3.08 % ROI 1.80
YR Pace 3rdOB3C2Back+3rdOBStrLast+Gain2CToFin+ImprovedART Winners 140 6.43 % IV 2.09

YR Trot 3rdOB3C2Back+3rdOBStrLast+Gain2CToFin+ImprovedART All 240 2.92 % ROI 0.21
YR Trot 3rdOB3C2Back+3rdOBStrLast+Gain2CToFin+ImprovedART Winners 33 3.03 % IV 1.04

Obviously, it is best used on pace races. I don't know if I would call it "hugely profitable," but an ROI of 1.80 at Yonkers is not too shabby. That is ROI, NOT "dollars returned for $2.00 wagers." Happy Holidays! (And don't say bad things about statistics.)

[traynor]

I should probably mention that the spot play mentioned is not wishin' and hopin' and chasing rainbows. Specifically, all mutuels have been truncated to the mean of the interquartile range. In plain English, that means all the outliers that might distort the results have been corrected. They still exist in play, but for modeling purposes, the data has been cleaned using standard statistical modeling processes and procedure to prevent possible distortion by anomalies and outliers.

The reason is simple--by correcting for outliers and anomalies, the results in the real world are usually better than the model indicates. Results that fail to correct for outliers and anomalies tend to produce results in the real world that are worse than the model indicates.

[traynor]

Error:
The second sentence above should read:
"Specifically, all mutuels have been truncated to 1.5 times the mean of the interquartile range."

My apologies for the error. It is, after all, Christmas.

[imofe]

Good stuff traynor. Not sure why anyone would not clean up the results, but I have read amazing things on other boards.

I think we had a little confusion when I wrote about using contenders as opposed to the whole field. Negatives in a horse does not automatically make it a non contender. A trainer change to a weaker trainer alone would definitely not eliminate a horse from being a contender. Many times the good trainer has done so many things that it takes a while for the horse to go off form. All I am saying is that in every test of a spot play or methodology that I have done with long term results, the results were always better when applied to contenders only. This is different from saying that each horse must have positive upon positive from form to driver to money box, etc. I guess it can also be how someone determines the contenders.

Anyway, thanks for all the thought provoking posts and happy holidays to all.

[traynor]

Good stuff traynor. Not sure why anyone would not clean up the results, but I have read amazing things on other boards.

I think we had a little confusion when I wrote about using contenders as opposed to the whole field. Negatives in a horse does not automatically make it a non contender. A trainer change to a weaker trainer alone would definitely not eliminate a horse from being a contender. Many times the good trainer has done so many things that it takes a while for the horse to go off form. All I am saying is that in every test of a spot play or methodology that I have done with long term results, the results were always better when applied to contenders only. This is different from saying that each horse must have positive upon positive from form to driver to money box, etc. I guess it can also be how someone determines the contenders.

Anyway, thanks for all the thought provoking posts and happy holidays to all.

The reason for NOT cleaning results is that it creates the illusion of profitability. Good ego boost for those who don't bet much. Correcting for outliers usually has the effect of making (what the model maker believes to be) profitable models non-profit. Hard on the ego. Fortunately, I am far more interested in profit than in stroking my own ego.

Specifically regarding contenders, I will have to test that. The problem is that it tweaks all the other results unless specifically coded as an alternative factor/attribute. Meaning, XYZ is different than XYZ iff (if and only if) contenders are considered. It would represent a subcategory of XYZ that might have better or worse values than the category XYZ. When I have more time, I will definitely look into it. Thanks for the suggestion.

[traynor]

Why, you may ask, would anyone possibly care about statistics? Because poking around with statistics turns up stuff like:

Nfld Pace DriverWPercent20PlusAll 1401 23.34 % ROI 0.67
Nfld Pace DriverWPercent20PlusWinners 171 23.98 % IV 1.03
Nfld Trot DriverWPercent20PlusAll 470 18.09 % ROI 0.76
Nfld Trot DriverWPercent20PlusWinners 66 18.18 % IV 1.01

Nfld Pace DriverWPercent15To20All 1401 47.32 % ROI 0.80
Nfld Pace DriverWPercent15To20Winners 171 47.37 % IV 1.00
Nfld Trot DriverWPercent15To20All 470 45.32 % ROI 0.94
Nfld Trot DriverWPercent15To20Winners 66 43.94 % IV 0.97

Nfld Pace DriverWPercent10To15All 1401 6.71 % ROI 1.23
Nfld Pace DriverWPercent10To15Winners 171 6.43 % IV 0.96
Nfld Trot DriverWPercent10To15All 470 13.40 % ROI 1.34
Nfld Trot DriverWPercent10To15Winners 66 13.64 % IV 1.02

Nfld Pace DriverWPercentLessThan10All 1401 45.90 % ROI 0.62
Nfld Pace DriverWPercentLessThan10Winners 171 45.61 % IV 0.99
Nfld Trot DriverWPercentLessThan10All 470 41.06 % ROI 0.58
Nfld Trot DriverWPercentLessThan10Winners 66 40.91 % IV 1.00

The best drivers tend to be overbet, simply because they are the best drivers. They win more races, but the mutuels in those races are dismal. Profit often exists in areas overlooked by the crowd.

Again, all mutuels have been corrected for outliers to 1.5 times the mean of the interquartile range.

[imofe]

Yes. The top drivers are overbet. But look at the results for the under 10% drivers. What do you think the results would be if we looked across all tracks at drivers that are 5% or less, or horses that are 0 for their last 50 racing in some claiming race, or ones that have not made any meaningful move in many races? My point is that while we need to avoid commonly overbet situations, we also need to stay away from horses that have little chance of winning even though the public is right with us on this one.

Now we can get somewhere. We eliminate those horses that have little chance of winning. Out of the horses we have left, we have to avoid overbet situations like the ones you have mentioned in several posts or we will end up with a lot of winners but no profit.

[traynor]

Yes. The top drivers are overbet. But look at the results for the under 10% drivers. What do you think the results would be if we looked across all tracks at drivers that are 5% or less, or horses that are 0 for their last 50 racing in some claiming race, or ones that have not made any meaningful move in many races? My point is that while we need to avoid commonly overbet situations, we also need to stay away from horses that have little chance of winning even though the public is right with us on this one.

Now we can get somewhere. We eliminate those horses that have little chance of winning. Out of the horses we have left, we have to avoid overbet situations like the ones you have mentioned in several posts or we will end up with a lot of winners but no profit.

Exactly. If you look at situations that the betting public overlooks, ignores, or fails to understand--then apply filters to determine the good bets from the bad bets (fairly easy to do, but time consuming), you come up with a whole lot of "counter-intuitive" bets that most bettors don't even know exist.

I agree completely that this is a worthwhile area to pursue. However, as I said, it takes quite a bit of additional coding, which all takes time. From what I am seeing, and as you mention above, I think there are a lot of overlooked profit opportunities in areas that the betting public generally ignores. To paraphrase Gordon Gecko in Wall Street, "Profit is good."

The attributes I posted are not the be all and end all. They are simply pointers (which is what statistics really are, rather than carved-in-stone certainty) that can be used as the basis for further exploration.