DB Query: Weights in Handicap races

[ceejay]

I read in a recent DRF Simo Weekly Davidowitz's idea that weights for horses are meaningless because even a 10 pound swing on a 1200 lb horse plus 115 lb rider represents less than 0.8% swing. I agree intuitively but was wondering if any of the DB guys out there can run a test of the idea?

I was thinking that the way to test would be to calculate win%, and win, place, show ROI's of High-weight, second high-weight, and third high-weight limited to Handicap Races.

Thanks in advance!

[billk]

I have found that assigned weight represents a horses ability. At some tracks they better carry more then 115 which means normally getting rid of the bug boys and a much higher win%. As far as I can see at these tracks only certain jocks win often and the rest make up the exotics. Top weight don't seem as important to me as is to little weight meaning a inferior horse in this race and they are trying to make him fit.

[JimG]

If it's meaningless, why do so many trainers try to get weight breaks? Why do the Raggies use it and people buy at $25 a pop? Just food for thought.



Jim

[Tom]

Originally posted by JimG
If it's meaningless, why do so many trainers try to get weight breaks? Why do the Raggies use it and people buy at $25 a pop? Just food for thought.



Jim

Good point, If weight really is meaningless to a horse's performance (within reason) but the trainer swears by it, might he not tell the jock not to push the horse today? I know of trainers at FL that would scratch a horse if it had to carry 122 pounds. We also used to have a rider that weighed something like 125 pounds and got caught paying off the clerk of scales. (FL is the only track I know of that has the scale positioned so that the public cannot see it). This guy was winning.
Tom

[PaceAdvantage]

This seems like an obvious application for the DB guys to run. I for one am anxious to see the results of this query. Larry, Que, are you out there???

I've always considered weight a fairly meaningless angle, but that's most likely because I never researched its impact on a race's outcome......


==PA

[Dick Schmidt]

Looked at weight in a small database (it seemed big at the time) of about 6,000 races. We found that the higher the weight, the more wins, which shows you that the good horses can carry the lead and still win.

A note on heavy jocks. A jock much make a certain weight, but can weigh as much over that as he wants so long as the trainers will put him up. Pinkay at one time was coming in at about 128, though he rode officially at 119 (at Santa Anita you can see the scales). No one said a word since he was winning and not breaking any rules.

Dick

[Aussieplayer]

Weight handicapping is as big as the Beyers down here! Our version of Andy Beyer is Don Scott who wrote about class/weight handicapping in Winning More and other books. Don't know if you guys could get a hold of it as it's very rare.

It's still a class table, but everything is expressed in kilograms (or pounds for you!).

It's the same in the UK too, though there probably almost as big on speed handicapping now. A lot of people are using speed here as well now.

In fact, Don Scott absolutely refuted the use of times, and used to love to quote one of your old punters Pittsburgh Phil in this regard.

It really is "class" handicapping WITH numbers at it's best.

If I wanted to handicap class I would use this method (class/weight figures) - as I could never get my head around a subjective use of class.

Developing a class table a la Scott is interesting - not sure if you guys could do it, as I'd need to know more about how your racing works.

Cheers
Aussieplayer

[Dick Schmidt]

Aussieplayer,

Don Scott sounds intriguing. Right now everyone in the US seems to be number crazed and computer driven. Not that this is a bad thing, but it is really dominating the tote boards. I've been a computer player for years, but I think that right now there is a big opportunity for a new look at handicapping.

I used to correspond to a chap who was using US style pace handicapping in Aus. He hand times all the races from his TV. It wasn't perfect, but he didn't have any competition and made a lot of money.

I use an offshore account at times and they cover Australian racing. Don't know a thing about it, but I sure am impressed by the payoffs. We never see so many double digit winners over here. Are the fields really big? I've been told there are lots of very sophisticated players down under and I was surprised by the payoffs I saw.

Good on ya,

Dick

[Larry Hamilton]

Damned if I can see it guys. First I ran weights and found that the various weighted horses won from 10-14%. Since this is approximately random and there were no visible ROI niches exploitable in the data, we can assume that weight doesn't seem to work here. So, I included two more factors--age and sex. With these factors, the distrubution remained close to random with all ROI and win% niches attribuatble to small samples. Sample size 40,000+ races.

Now, if I had a "number" to attribute to class, we could run weight and class, but I don't, does anyone (one that is acceptable to all)?

SInce weight is a subset of class, at least in the eyes of the guys who assign the weight, and some trainers believe in weight, I would guess that you can find a use for weights only when combined with SOME trainers. Look at the question like this: Does Scott Lake win with horses carring 113 lbs as frequently as horses that carry 126?

[ceejay]

Originally posted by Aussieplayer
Weight handicapping is as big as the Beyers down here! Our version of Andy Beyer is Don Scott who wrote about class/weight handicapping in Winning More and other books. Don't know if you guys could get a hold of it as it's very rare.

It's still a class table, but everything is expressed in kilograms (or pounds for you!).



Aussieplayer,

How much weight differential is commonly assigned down there?

[BIG HIT]

Hi guys the way i understand weight it was not supposed to weight to make him that much slower. But to slow or make him a bit off balance.Which made him have trouble changeing leads on the turns into the str.Read it some where. Which make a little more sense?

[Rick Ransom]

In one of my old spot plays involving horses in good form, if I played low weight horses or horses dropping in weight I about doubled my ROI. This won't work universally though. I think it was because I had a lot of horses stepping up in class.

[Tim]

High weight, low wieght, live weight, dead weight, apprentice allowance, over weight, whatever... I've walked around the issues concerning "weight" in horseracing for over 30 years and have never fiqured it out.

Larry's right. The weight on a horse doesn't appear to have a measurable effect on his race.

An individual horse? Trainer intent? Maybe, but thats not something you can quantify with a database.

Tim

[NoDayJob]

You might want to refer to the scale of weights; both American and English. -NDJ

[andicap]

One place the experts seem to agree on is that each horse has a limit, a breaking point. Obviously Forego's was pretty damn high (137).
Of course its all tied to class. A horse might be able to beat Gr 3 horses carryiing 128 but not Gr. 1 horses.

I've never used weight much so I'll agree with Larry here but the Raggies did a lot of research of their own which always made me wonder.

andicap

[Tim]

The following table is the raw result of a quick and dirty query on 73,535 races with 602,406 starters.
I eliminated the horses that carried less that 100 pounds (3) or more than 130 pounds (mostly steeplechasers) from the query. Because of the small sample size and the impact of steeplechaser starters I would treat the 124 pound and over entries with a grain of salt.

My point is, if you slice and dice this data set by class, age, field size, odds, condition book author, etc.... you will probably get the same flavor of results.

Definitions:

Wght: The actual weight a horse carried as extracted from Bris Import Charts
Strs: The number of horses carrying that weight
Wins: The number of winners carrying that weight
Win%: Wins / Strs
% of total: Number of starters divided by 602,406

Wght Strs Wins Win% % of total
100 7 0 .0 .0
101 17 0 .0 .0
102 37 3 8.1 .0
103 93 4 4.3 .0
104 176 14 8.0 .0
105 545 37 6.8 .1
106 1,472 98 6.7 .2
107 2,399 192 8.0 .4
108 5,210 457 8.8 .9
109 6,984 657 9.4 1.2
110 10,151 1,098 10.8 1.7
111 12,163 1,316 10.8 2.0
112 21,482 2,460 11.5 3.6
113 31,164 3,419 11.0 5.2
114 49,691 5,655 11.4 8.2
115 57,069 6,943 12.2 9.5
116 79,299 9,749 12.3 13.2
117 102,178 12,903 12.6 17.0
118 69,125 8,564 12.4 11.5
119 55,559 7,335 13.2 9.2
120 39,408 4,918 12.5 6.5
121 19,644 2,651 13.5 3.3
122 31,900 4,154 13.0 5.3
123 4,827 639 13.2 .8
124 892 143 16.0 .1
125 51 20 39.2 .0
126 801 94 11.7 .1
127 16 6 37.5 .0
128 11 3 27.3 .0
129 6 0 .0 .0
130 29 3 10.3 .0
Tot. 602,406 73,535 12.2 100.0


andicap,

While I don't know the answer, my instints say that the "raggies" are applying a subjective weight factor similar to the "projection handicapping techniques" that we see in speed and pace numbers. I would love to know that answer to that one.


Tim

[NoDayJob]

Tim,

It would be interesting to apply the scale of weights to the various age/sex of the different races; then see what the results are. Something you database people might want to do when you have the time. -NDJ

[ceejay]

Warning: Don't read this unless you are comfortable with numbers and statistics!

Thank you, Tim for posting. I see a distinct positive correlation between increasing weight and increasing win percentage. Specifically, the predictive equation (reduced major axis regression method) using weights of 110 to 123 pounds is:
Win% = 0.221 * Weight (lbs) - 13.57

The relationship has a correlation coefficient (r) of 0.93, which is significant at the 99.9% level of confidence.

Of course, correlation does not necessitate causation!

So, I interpret weight as being important in handicapping, just not as the track handicapper expects. The difference in win percentage between 110 and 123 pounds is only 3% so I wouldn't use the relationship to predict odds and wager.
Rather, I interpret the data as meaning that added weight does not impair a horses ability to win. Weight may simply be a reflection and function of recent form.

[Aussieplayer]

Dick,

Yes, whilst the class/weight rating style is popular here, it could provide a good edge over there. That was why I used speed down here - as no one was really onto it. Trouble is (now), others are on the bandwagon - plus it can be quite a headache - so I'm not sure how long I'll be using speed handicapping for here.

Don Scott's ideas can be computerised by the way. (To finish the story he died in the mid-nineties, a millionaire punter).

Your friend would have had to hand time the races for pace ratings, as we only have the final times recorded. We also have the times internally for when the first horse hits - but unless you know how far in lengths back the others were, what's the use? But of course (as you can see from my signature, lol) that is why your friend had/has an edge. After all, who'se going to bother compiling raw data? Your friend did, and reaped the benefits.

Interestingly (sorry I'm jumping about a bit) Scott used a kind of projection method that reminds me of what you guys talk about with the Beyers. He called it, "Q" factor.

Geez and I complain about our payoffs, lol :)
Seriously, I guess we do have quite a few $10+ horses ($20+ to you as we're in $1 units).
The best way to answer about field size is that metropolitan racing (city tracks) have an average field size of about 11.5, whilst non-metro. is probably a bit smaller at around 10.
We really complain about anything less than 8, and most prefer at least 10 for playing trifectas.
Field sizes of 12, 14 and 16 are not at all uncommon. And the Melbourne cup is of course the maximum allowed at 24 runners over 3200m (2 miles).
We have quite a few races at 2000m+ (1m1/4+) as well.

I've heard others from the states talk about the sophisticated players in Oz. Whether it's true or not, I guess it's perhaps due to the either true or fabled "computer gangs" that have supposedly made a fortune here and in Hong Kong.
We have three "totes" (betting pools), one for New South Wales (Sydney), Queensland, and another for the rest. I have been told from pro punters that most big players here are arbitrage players, with direct access into those 3 totes, and who maybe make 1% on turnover from very big bets. (This is the same thing that you guys had an outcry about with the Dakota guy - it's legal here, the called the tote high rollers!).

We also have bookies that you can bet with at the track, on the phone or on the net with. Which is all great as seeking the best price on your horse can really make a difference to your profit line.

Ceejay:
What do you mean? In the class table, or by the handicapper to winners??

Anyway, enough from me.
Always good talking to you guys :-)

Aussieplayer

[Dick Schmidt]

Aussieplayer,

Thanks for taking the time to reply. I did some searching online for the Don Scott book and you were right - it is impossible to find. I looked in every bookstore I could find both here and in Australia. If you know of any way to get a copy, I'd love to read it. I've got a small collection of racing books, some of which were so hard to find I had to photocopy them. Never can tell where you'll come up with a winning idea.

Thanks again,

Dick

[hdcper]

Tim,

Thanks for the posting of your test of how weight effects win percentage. Would you mind re-doing the data and including the average mutuel of the winners in each weight category? Would like to see if high weight horses are generally the favorites and the increased win percentage is explainable.

Thanks,

hdcper

[ceejay]

Aussieplayer:

To rephrase my question:
What is the common difference between high weight and low weight in a single race down there? Here in the States, we commonly see less than 10-12 pounds between high- and low-weight in a given race.

[karlskorner]

Ceejay;

Al Urlich had an interesting theory years ago about weight (lead) carried in the saddle and that carried by the Jockey. His thoughts were that the saddle weight was "dead" weight, while the Jockey could distribute his weight over the shoulders.

Karl

[Aussieplayer]

Ceejay,

thought that's what you meant, but wanted to clarify.
At 2.2 pounds per kilogram (2.2046), it actually works out to around the same as you at 10-12 pounds for our races. There are exceptions where you could see anything from 9-17 pounds, but the 10-12 pound range is the norm.

Hope this helps
Aussieplayer

[andicap]

I remember when Forego ran with 136 pounds someone saying that he would do better over longer distances because it would take more time to build up momentum. Kind of like a freight train. The weight inhibits the horse's start, but once he starts rolling the momentum -- HELPED by all the that -- carries him forward like a slingshot.

andicap

[Tim]

hdcper,

Because of an active schedule between now and Sunday (including another futile stab at the Bris contest on Friday) I probably won't get to that query until Sunday evening.

Tim

[hdcper]

Thanks Tim, appreciate you running the query and understand the time frame.

Thanks,

hdcper

[hdcper]

Hi Tim,

Just wondering how your time is looking with regard to the additional query info?

Thanks,

hdcper

[Tim]

Hdcper,

The following table contains the query results you asked about. The statistics, 73,824 winners and 603,725 starters are different from the original query because;
(1) I eliminated the starters over 124 pounds because the samples are to small to be useful and the mix of steeplechasers.
(2) My database grows by about 300 races or 2,500 running lines a week.

Definitions:

Weight: The actual weight a horse carried
Strs: The number of horses carrying that weight
Wins: The number of winners carrying that weight
W%: Wins / Strs
Avg: The average odds of all starters at that weight
Lose: The average odds of all losing starters at that weight
Win: The average odds of all winning starters at that weight
L/W: Average losing odds / average winning odds at that weight



Weight Strs Wins W% Avg Lose Win L/W
100 7 0 0.0 23.01 23.01 0.00 NA
101 17 0 0.0 24.85 24.85 0.00 NA
102 37 3 8.1 26.86 28.22 11.45 2.5
103 93 4 4.3 30.51 31.54 7.50 4.2
104 176 14 8.0 28.22 29.85 9.40 3.2
105 543 37 6.8 24.49 25.60 9.19 2.8
106 1,472 98 6.7 26.03 27.32 7.85 3.5
107 2,400 193 8.0 23.47 24.73 9.01 2.7
108 5,218 458 8.8 22.27 23.71 7.29 3.3
109 6,988 657 9.4 21.20 22.67 7.03 3.2
110 10,138 1,099 10.8 20.62 22.32 6.65 3.4
111 12,197 1,320 10.8 20.23 21.91 6.38 3.4
112 21,506 2,468 11.5 18.65 20.26 6.24 3.2
113 31,213 3,424 11.0 17.70 19.12 6.22 3.1
114 49,706 5,653 11.4 16.74 18.14 5.87 3.1
115 57,219 6,958 12.2 15.96 17.39 5.62 3.1
116 79,643 9,786 12.3 14.90 16.23 5.41 3.0
117 102,377 12,934 12.6 14.36 15.69 5.16 3.0
118 69,616 8,629 12.4 16.19 17.75 5.13 3.5
119 55,788 7,363 13.2 15.11 16.71 4.58 3.6
120 39,597 4,944 12.5 17.47 19.26 4.94 3.9
121 19,744 2,666 13.5 15.54 17.23 4.67 3.7
122 32,037 4,170 13.0 16.54 18.34 4.56 4.0
123 4,910 648 13.2 16.62 18.52 4.12 4.5
124 925 146 15.8 15.38 17.60 3.50 5.0

Interesting stuff, hope it helps,

Tim

[ceejay]

Thanks again, Tim, for the additional data. Yes, it is interesting. What circuits or tracks are in your DB?

Although I concluded above that a direct relationship between weight and win % exists, the ROI data (if I'm understanding the table correctly) shows that that low weight tends to correlate with higher ROI's.

ROI (2.00 breakeven) = -0.0197 * Weight (lb) + 3.839, using reduced major axis regression.

The correlation coefficient (r) is (-)0.91, almost as strong as the win% r and still statistically significant (99.9% level of confidence) based on a T-test.

I interpret this as meaning that the stronger, higher-weight horses are overbet.

A PowerPoint presentation with the regression lines is attached.

[hdcper]

Thanks Tim for all your hard work. I thought that what is reflected from this data, is what one would expect. Although higher weighted horses win more races percentage wise, the public over bets them sufficiently to offset their edge in ability.

The below chart reflects the weight categories, the win percentage for each weight and the overall ROI where 2.00 is breakeven. Basically this was calculated by multiplying the win percentage for each weight times (the average odds times 2 plus $2 for the actual wager or in other words the actual average win price).


Weight Win % Avg Odds ROI
102 8.1 11.45 2.0169
103 4.3 7.5 0.731
104 8 9.4 1.664
105 6.8 9.19 1.38584
106 6.7 7.85 1.1859
107 8 9.01 1.6016
108 8.8 7.29 1.45904
109 9.4 7.03 1.50964
110 10.8 6.65 1.6524
111 10.8 6.38 1.59408
112 11.5 6.24 1.6652
113 11 6.22 1.5884
114 11.4 5.87 1.56636
115 12.2 5.62 1.61528
116 12.3 5.41 1.57686
117 12.6 5.16 1.55232
118 12.4 5.13 1.52024
119 13.2 4.58 1.47312
120 12.5 4.94 1.485
121 13.5 4.67 1.5309
122 13 4.56 1.4456
123 13.2 4.12 1.35168
124 15.8 3.5 1.422

Conclusion:

High weight horses do win more races, but don't result in a better bottom line.

Thanks again Tim,

hdcper

Edit Note: not certain how to make my chart in line with the columns, but they appeared fine in my post prior to submitting it. Sorry everyone!

[hdcper]

Hey Ceejay,

Guess you were posting your analysis the same time I was. I must admit a picture is worth a thousand words and makes it very easy for others to see the relationship.

Great post,

hdcper

[Tim]

ceejay,

Circuits in the dataset,

NYRA = Aqu, Bel, Sar
NJ = Mth, Med
S Cal = Sa, Hol, Dmr
Fla = GP, Hia, CRC
Ky = CD, Kee, Tp, Elp
Fg & Op to help round out winter racing

The last column L/W shows the relationship between average winning and losing odds. The higher the number, the higher the public recognition of the measured factor.(in this case weight) The table tells me that when the weight is at 118 pounds(3.5 or greater) or over the public at large is pretty good at identifying its importance as a winning or losing factor. That number by itself won't help you much, but I find it a good indicator of wether to drill down and search for concepts that have value. While I didn't consider ROI's when generating the query, your conclusion is correct, the high weights are overbet.

What I took out of the table is the question, can positive or negative factors be isolated in horses with a weight between 110 and 116 pounds. The win percentage is pretty close to normal, the average winning odds don't look bad and about half the population fits the bill.

Tim

[Rick Ransom]

Since low odds is also associated with a higher ROI than high odds, I wonder if a low weight AND low odds horse might have an even better ROI. Anyone interested in checking it?

[Rick Ransom]

In these weight studies, do you guys include the apprentice allowance or not. Also, are you using the assigned weight or the corrected weight?

[Tim]

Rick,

All the weight stats I post are based on the actual weight carried. My data source is charts, not pp's. The weights are after overweight and apprentice allowances.

Tim

[Rick Ransom]

Thanks Tim. That's what I thought. You wouldn't have any way to tell whether a jockey is over the assigned weight then. I assume that the weights we see in the past performances are also corrected weights.

[Tuffmug]

Maybe you guys are looking at this wrong? I think weights need to be considered RELATIVE to weights carried by other horses in THAT PARTICULAR RACE and RELATIVE to what that horse carried previously

Consider this, assuming 1 length =8ft, there are 825 lengths in 1 1/4 miles and a very good horse can run that in 120 seconds. Now let's use Larry's numbers where he assumes that 10 lbs of weight added to a 1200 lb horse is a very small .8% difference.

If speed is related to weight then a horse .8 % heavier will run 1 1/4 that much slower:

120 second X 1.008 = 120.96 seconds or approximately 5 lengths slower than if he were 10 pounds lighter.


I don't know about you but I've seen an awful lot of races won by less than a length. A small difference in weight CAN make a big difference but you have to weight it relative to what each horse has done before at different weights

[Larry]

Let me restate this to make sure I understand. You are saying that the question is the weight carried delta is more relavent that the total weigt carried. For instance, the query posed could be, "When the favorite carries 3 pounds more than the closest competitor, how well does he do; or when horses odds are 10:1 and is carry 3 pounds more than the favorite, how well does each do?" Let me know if this is a correct restatement of your point, I am sure Can handle that.

[Rick Ransom]

Does a horse's speed rating decrease with an increase in weight from one race to the next, or is it obscured by the increased weight indicating good form?

[Aussieplayer]

Tuffmug,

This is exactly what Don Scott did. He dealt with what he called "relative" weight. A class/weight ratings was made up of the class rating (from a class table expresed in weight), and as a big believer in weight he also accounted for the "relative weights" or the "weight over the minimum."

Cheers
Aussieplayer