Dr. Z books - anyone have a spare copy?

[majestyx]

I've been unsuccessfully seeking the two Dr. Z books, BEAT THE RACETRACK and BETTING AT THE RACETRACK, and have had zero luck. Amazon, B&N, even the GBC in Vegas - none of them have them because they are out of print. If anyone has a spare copy of either that they'd be willing to sell, please let me know.

[GR1@HTR]

What is Dr. Z's real name? or name he used when writing the book? Let me know and I'll try to find it...

[Dave Schwartz]

William Ziemba

[Butch]

The Google search engine turned up about 70 hits.

The man's real name is William Ziemba.

Ziemba also wrote a third book about horse race mutuel betting, called: "The Efficiency of Racetrack Betting Markets", published in 1994. He's also written about lotteries, baccarat, sports, and other subjects. You can find a bibliography of his publications at: http://www.interchange.ubc.ca/ziemba/horses.html

You can buy the Dr. Z books at http://www.hypernormal.com/html/drz_place_show_browser.html , although that fact is buried in the middle of the page, which touts computer software that uses his system and the online toteboards. The seller wants $105 for the two books, postpaid.

If you assume that Dr Z's approach of "market inefficiencies" is valid (I actually believe that it's contrary to the peculiarities of the parimutuel market, and if anybody wants to know why, I'll be happy to explain it here), there are two basic problems with the approach today.

1) The place and show pools are much smaller, relatively, than they were when he wrote his book in 1984, the reason being the proliferation of exotic betting.

2) The online toteboards show the place and show pools, and these lead to players monitoring them and sending in bets on the alleged overlaid pools at the last minute. Horses who were overlaid in the place and/or show pools suddenly find themselves big underlays as everybody gets the same idea at the same time (similar to the horse that the public thinks should be even money going off at 8-5 a minute before post. Everybody pounds on it in the last minute, and it ends up at 4-5). The fact that there's now software out (re: above) formulated after the Dr. Z theory that reads the online totes, and this adds to the problem.

Butch

[majestyx]

Thanks for the quick replies. I do have the EFFICIENCIES OF RACETRACK BETTING MARKETS, a rather highly mathematical discourse which is pretty tough to follow without a degree in statitics/higher math. The numerous papers in this volume, which to me seems more like a college textbook than a handicapping tool, refer to the books I am seeking so I would like to get a hold of them. The $105 the guy is asking seems a bit nutty (as in outrageously overpriced), but if this is the last resort, I may have to make the plunge.

There is a somewhat simplified version of the Dr. Z system in the textbook (since I don't have the Dr. Z books, I don't know how complicated it is) which I've been experimenting with and it does seem to hold merit in the fact that there are still indeed discrepencies in the place and show pools, moreso the latter, as compared to the win odds. Since it appears that much research has shown that the win odds are a very good reflection of the horses real chances of winning, other than the favorite-longshot bias, it is considered that the win odds are an accurate reflection of winning probablities. Granted, this "system" isn't one that makes huge returns (about 8-12% ROI), not to mention that it takes a LOT of time to research because live odds need to be watched since this information doesn't seem to be available anywhere after the races have been run, but it is interesting from a purely "systematic" approach to playing the horses as compared to handicapping. I've been using the 2-minutes-to-post (MTP) cutoff to do my research which, as we all know, is MUCH different than using the closing pools. This is due to the fact that usually (at least in my studies) only around one-fourth of the money on average is being accounted for at the two-minute mark as compared to when the betting windows close (e.g. $12000 in the win pool at the 2MTP mark while there may be $48000 at closing). I've picked the 2MTP mark in order for it to be considered workable in real time so that a bet can be placed.

On a different topic, this discussion site is great! I didn't realize there was so much info out there and so many knowledgeable folks who follow the sport of kings.

[GR1@HTR]

sorry dude...can't find it.

[Rick Ransom]

If everyone's going to start playing Dr. Z maybe I'll do the opposite. I think there are a bunch of people that are doing a rough approximation of this by looking at online toteboard pool percentages. Since the information is so readily available, I think it must have very little value.

[Butch]

I can give you Dr. Z's system in a nutshell here if you like, and I might as well state why I don't feel it works in real life, too.

Z's system is based on the premise that the tote odds accurately reflect the horses' relative chances of winning, which is generally true over a large number of races, although (as Z notes) favorites tend to win a little more than their share (have a better than track take ROI) and longshots win considerably less than their share (a worse than track take ROI). The system boils down to this: Use each horse whose tote odds is 8-1 or less (I personally would go no higher than 9-2). Divide the total amount bet on the horse to win by the total win pool. You come up with a percentage. For example if there is 100,000 in the win pool and $20,000 is bet on the horse, you would come up with .20 or 20% Z would use the number of 20. Then do the same thing with the money bet in the place and show pools on this horse. Then divide the win number by the place and show numbers respectively. Example: The horse above has $10,000 of the $50,000 place pool and $3,000 of $20,000 in the show pool. The appropriate percentages are W: 20, P: 20, and S: 15. Dividing each produces a ratio of 1.0 (20/20) for Win to Place, and 1.33 (20/15) for Win to Show. The idea is to have a number larger than 1.0, and to overcome the track take (of 17%), one needs a number larger than 1.205 (100/83). A number smaller than 1 means that the horse has more money, relatively, bet on it to place and/or show than to win. This is true of almost all longshots. A number higher than 1.0 means that the horse has less money, relatively, bet to place/show than it does to win. This is true of most favorites and short-priced horses.

The reasons that I don't feel Z's system works are many, but I'll isolate the three biggest ones.

For one, Z hints that the inefficiency may often be due to 'smart money'. Supposedly, insider money is bet to win only. In real life, this simply isn't the case any more, although it might have been back in the late 70's when Z formulated this theory (and somewhat in the mid 80's when he wrote the books) . Most insiders (by this I mean owners, trainers, stables in general) bet equally to win and place, and the stables that go for the big payouts (call it a fixed race, ringer, drugged, etc., if you like) usually hide their money in the exotics, which weren't as plentiful when Z was researching as they are today.

For another, the online web totes are available to bettors all over the world. How many times have you seen a horse that you thought should have been about even money going off at 9-5 with one minute to post. You phone your dialup account, bet $20 to win on it----then sigh as you see it get off at 4-5 by mid-race. The problem, of course, is that everybody else saw it getting off at 9-5, too, and they all pounced on it. With place and show pools, which aren't nearly as big, it doesn't take much money to move them from overlay to underlay status at most tracks. And there lies the rub. If you're betting, say, $50 to show, your $50 itself might cause the inefficiency to disappear. If five other guys---this being from all over the country---are doing the same thing, you have a nice fat underlay.

The third reason is tied into the end of the previous one. Straight pools are much smaller now than they were 15 to 25 years ago (and before). This is due to the fact that racetrack handle at many (not all) tracks is down from that period and also because there are more exotic wagers now than there were then. Naturally, some of the money that would be bet in the straight pools gets diverted to the exotics, and it takes less money in the place or show pools to move a horse from overlay to underlay status by Z's criteria.

You can also add in the fact that place and show payoffs are more likely to result in larger breakage percentages, especially on the lower priced horses.

If you must use this method, I have a suggestion: The last race at each track tends to hold up well to Dr. Z's system, especially . The reason for this is that many players are looking for a way to get even at that point, and very few are interested in betting a 7-5 shot for the $3.00 that it might pay to show. This was a lot stronger in tbhe 80's and earlier, before simulcasting, and before the web totes of the mid 90's, but it still has some merit today.

A popular system for a couple of years, from about 1995 to mid 1998 was actually a form of inverted Dr Z system. It looked for horses with considerably MORE bet to place or show on them, relatively, than to win, this being read early in the tote activity, the earlier the better. The idea was that the 'inside' money was betting to win and place, and stables put their money in earlier than most bettors do. The system was to bet to win, which is sort of saying that the place pool was accurate and one was looking for inefficiency in the win pool! This system actually worked well for a while, until somebody published it through a popular online reseller of systems. The reseller said that about 150 copies of the system were sold, and horses bet in this way dropped considerably in the odds in the last minute as bettors using this system hopped aboard. The result was underlaid horses, and if there were any insiders of note, they changed their betting habits, because such betdowns became much rarer.

Butch

[majestyx]

Wow, Butch, thanks for the great information and insight. I think there was an additional part of the system that involved money management using the Kelly Criterion, meaning that the more likely the bet is to pay off, the higher percentage of one's bankroll would be invested. It uses a very involved formula that was supposedly simplified on a card or sheet that came with the book.

The phenomenon you referred to in your first paragraph is the favorite-longshot bias.

It's amazing that only 150 copies of a system had that huge an impact on payoffs. I hope whoever came up with it got a pretty penny for each copy to help offset the ruined returns at the track.

[NoDayJob]

Dr. "Z" made his money selling books and programs, certainly not betting the nags. There's nothing wrong with that. However, I have yet to see any Stat or Math whiz making a living at the race track. There's a Professor of statistics that I met while in the S.F. area. He taught at Hayward State University. He wrote several books on gambling and a horse racing program that he actively sold. When I asked him how much he bet on the horses he said, "Bet on the horses, ha, that's for fools". Now you know why I'm so cynical. -NDJ

[Rick Ransom]

Another problem with "Dr. Z" is that the place and show probabilities for low odds horses are overestimated by a substantial amount. At least that's how it was in the original version; maybe they've corrected it by now. One of the books by Dick Mitchell mentioned this and said that Barry Meadow had studied it. Barry, can you clear this up for us?

[takeout]

Originally posted by NoDayJob
When I asked him how much he bet on the horses he said, "Bet on the horses, ha, that's for fools". Now you know why I'm so cynical. -NDJ

That put me in mind of that old bootlegger's quote: "I make this stuff to sell, not to drink."

[anotherdave]

Originally posted by NoDayJob
Dr. "Z" made his money selling books and programs, certainly not betting the nags. There's nothing wrong with that. However, I have yet to see any Stat or Math whiz making a living at the race track. There's a Professor of statistics that I met while in the S.F. area. He taught at Hayward State University. He wrote several books on gambling and a horse racing program that he actively sold. When I asked him how much he bet on the horses he said, "Bet on the horses, ha, that's for fools". Now you know why I'm so cynical. -NDJ

Hey, I happen to have a degree in Statistics. Now I may not be a whiz, but I'm up $3000 in 7 months so far this year. Let's see that means I could make over $5000 in a year if I went full-time! Hmmm. Who needs a job with dough like that flowing in!

[Butch]

It might work if the track it's being played at is large enough to handle your show bet without influencing the odds (examples: SoCal, NYRA, or any medium-large track on weekends). The problem lies in the fact that it doesn't take much money to turn an overlay into an underlay in the place/show pools, and since you know there are probably a hundred guys out there in Cyberland using Z's system with web totes and phone or internet betting, it's a shaky proposition at best.

One way that it might work, though, is through an offshore book (i.e. one that doesn't commingle into the track pools) that takes place/show bets without a win bet on top. There, of course, your money wouldn't affect the track prices.

Butch

[bdhsheets]

majestyx:

I sent you a private message regarding the books, let me know.

Regards,

bdhsheets

[NoDayJob]

Originally posted by anotherdave


Hey, I happen to have a degree in Statistics. Now I may not be a whiz, but I'm up $3000 in 7 months so far this year. Let's see that means I could make over $5000 in a year if I went full-time! Hmmm. Who needs a job with dough like that flowing in!


Just increase the size of your bets by a factor of 75 and you can quit your day job, maybe. -NDJ

[BMeadow]

Originally posted by Rick Ransom
Another problem with "Dr. Z" is that the place and show probabilities for low odds horses are overestimated by a substantial amount. At least that's how it was in the original version; maybe they've corrected it by now. One of the books by Dick Mitchell mentioned this and said that Barry Meadow had studied it. Barry, can you clear this up for us?

In Money Secrets At The Racetrack, I described a study I did to check whether the Harville formulas, the basis of the Dr. Z system, were accurate. I checked 1000 odds-on horses and compared their predicted place and show percentages under the Harville algorithms with their actual place and show percentages. In every case, the Dr. Z system substantially overestimated how often these horses would hit the board. For instance, the Dr. Z method estimated that horses who went off at 0.80 - 0.99 to 1 would hit the board 90.6% of the time; they actually finished 1-2-3 only 83.4%.

The basic premise--some horses are underbet to place and/or to show--is of course true. Many of us study the pools to find these underbet horses, and then hope they're not pounded in the last few seconds of betting.

Dr. Z was a mathematician and statistician, not a gambler. He was a smart guy, but his assumptions simply weren't true. Still, he made a contribution to handicapping by getting players to think of the right way to look at place and show--as two more possible avenues of profit.

[Dave Schwartz]

Barry,

It is good to have an opportunity to agree with you for a change. <G>

My contention would be that, even if the formulas were correct back then, that racing has changed since, as evidenced by the worsening of the loss on favorites.


Regards,
Dave Schwartz

[hurrikane]

I find a lot of out of print old race books at the public library...surprising the old crap they have there.

[Lefty]

AnotherDave, please read NDJ's post very carefully and
the give that money back. You shouldn't have it.
Experts also say bumblebees shouldn't be able to fly but
the dumb loittle bees just keep on doing it.

[Rick Ransom]

anotherdave,

You might enjoy reading "Efficiency of Racetrack Betting Markets", edited by Hausch, Lo, and Ziemba (yes, Dr.Z). It was originally published in 1994, but republished by Academic Press and available at Gamblers Book Club in Las Vegas. It's expensive and has lots of math but contains some interesting technical papers. One of them is about a Hong Kong betting syndicate using a multinomial logit model to bet horses that showed a nice profit over a five year period. My experience is that ordinary run-of-the-mill multiple linear regression models don't work too well in horse racing. What are your thoughts on this?

[Rick Ransom]

Barry, Dave

The book I mentioned in the last post discusses this very problem. In order to get the probabilities of finishing second and third right, they had to adjust them by downgrading high win probability horses and upgrading low win probability horses. They justified this by saying that there is increased randomness in the second and third place finishes. The whole point is that if the facts don't fit the theory, you should change the theory.

[anotherdave]

Originally posted by Rick Ransom
anotherdave,

You might enjoy reading "Efficiency of Racetrack Betting Markets", edited by Hausch, Lo, and Ziemba (yes, Dr.Z). It was originally published in 1994, but republished by Academic Press and available at Gamblers Book Club in Las Vegas. It's expensive and has lots of math but contains some interesting technical papers. One of them is about a Hong Kong betting syndicate using a multinomial logit model to bet horses that showed a nice profit over a five year period. My experience is that ordinary run-of-the-mill multiple linear regression models don't work too well in horse racing. What are your thoughts on this?

Thanks for the tip. I have read it. I didn't buy it, I got it through the library. A few of the articles were even references in my thesis.

I have tried multiple regression and found it to be unsatisfactory on its own. I use the odds line on HowTheyRunFast as a guideline and make adjustments using my handicapping knowledge. My main focus is on looking for underbet horses.



The Harville formulae do overestimate the place and show probabilities, in my opinion. A statistician with no knowledge of horse racing doesn't understand. I have seen too many horses who will either get the lead and win or finish down the track. A front runner who might have a 20% chance of winning, might have a 21% chance of being in the first three.

[Rick Ransom]

anotherdave,

There are a few uses for multiple regression. One is reverse engineering somebody else's power rating or speed rating or whatever. It's kind of a nasty trick, but sometimes you learn something from it. Another thing I've used it for is making sure my ratings add significant information above and beyond the odds. I predict finish position using odds (actually log odds), number of horses, and my ratings (normalized to the average rating). If the coefficient for my rating is significant, then I feel that I have a pretty good prediction. If it isn't, say at a particular track, I know there's trouble ahead.

There seems to be no value in using this scheme to develop the rating though. Normal distribution assumptions don't apply and the "win or die" syndrome you mentioned causes a lot of noise. You get significant coefficients, they just don't work. It winds up being pretty good at predicting odds or morning line odds though. Apparently the logit model works, but you need a lot of data and it would be better if you only played one track.

A little bit of knowledge is a dangerous thing as far as statistics goes. Fortunately, I read a lot about financial and economic forecasting before I jumped in. Their problems sound very familiar to ours as horseplayers.

[Rick Ransom]

By the way, while I think of it, I have a quick and dirty way of deciding whether I'm doing well for the day when I'm at the track. I just compare the finish position of each of my selections to the odds rank of that horse. In other words, if my horse was 4th in the odds, but finished 2nd, I'll say I'm doing a good job even though I didn't win the race. If you do this for a few races you'll get a good idea of how you'll probably do the rest of the day. If your choices are finishing worse than their odds, it may be time to go home. I use this a lot when I'm not familiar with a track and only have a few races to go on.

[majestyx]

In the Efficiencies... book there is a discussion of two new models for evaluating the chances of place/show probabilities - the Henery and Stern formulas which are even more sophisticated and tougher to implement due to having to use integrations to arrive at the results. The Henery model was used successfully in Hong Kong and the USA, while the Stern model worked better in Japan, showing that the models aren't necessarily universal. The Hong Kong test was shown to provide a profit in Hong Kong quinella bets. Unfortunately, these papers are only discussions of results, not showing HOW the "system" was implemented nor the actual results which I'm sure would fill a volume even larger than the book itself.

For anyone who has the book, there is a paper by Ritter that is interesting in the fact that he presents a simple betting rule which showed a profit in post time odds, but, as has been discussed earlier in this thread, it was not shown to work in a real-time basis such as placing the bet based on 1.5 minutes to post time. The sampling of the real-life races was quite small - only 10 races, so perhaps there is some merit in the system presented after all. However, the races were from 1979 and race betting has changed significantly since this time.

[Rick Ransom]

majestyx,

I'll have to look at those papers again. The one that showed negative results at 1 1/2 minutes I remember though. With the interstate simulcast money pouring in there at the last second now, I just don't see how any of these toteboard systems could work. My local track used to have separate pools up until a couple of years ago and there used to be a guy who looked for inefficiencies in the win pool compared with the track odds.

He would always be found at the machine at one minute to post and since I would rather bet on a machine than with a person (call me weird, but I've never made a mistake punching buttons and a lot of clerks have), we would frequently encounter one another. We called him Mister Toteboard. I don't know if he made any money or not but I can guarantee you that he bet a lot. In fact, he was so preoccupied with the odds that he left his voucher in the machine more than once with hundreds of dollars on it.

Anyway, my point is that if there is a situation to be exploited there are plenty of people already trying to do it. Look at the OJC pools sometime. You'll think you have a big time overlay and when the race is off, it'll disappear.