Trading and Betting the Horses

OddTrader · Feb 27, 2016

http://lindaraschke.net/trading-and-betting-the-horses-2/ June 13, 2014

Trading and Betting the Horses

While some people question how easy it is to make a consistent living TRADING, how about making a consistent, comfortable living betting on horse races? Here is a story about Ernest Dahlman – someone who has experienced amazing success at betting on horses for the past 35 years. Some of the principles that have contributed to E.D.’s success are the very same applied by top, professional traders. The following includes excerpts from a June 3, 2001 article titled THE WIZARD OF ODDS, published by the New York Times and written by William Grimes. We recommend that before you read our story you read the original article in its entirety. It is necessary to register first on the NY times website. They will have the article freely available only until June 10.
More...

OddTrader · Feb 27, 2016

http://www.nytimes.com/2001/06/03/magazine/03DAHLMAN.html?pagewanted=all

The Wizard of Odds
By WILLIAM GRIMES
Published: June 3, 2001

Like many people, Ernie Dahlman shows up at the office around 8:45, switches on his computer and spends the rest of the day making telephone calls. His office is nondescript, and so is he: blandly handsome, of average height, with an open, friendly expression and rimless glasses that give him a mildly intellectual look. He could be a travel agent, or a stockbroker, or perhaps a real-estate agent.

Actually he is one of the biggest horseplayers in the United States. Just how big is anyone's guess. In a busy year, Dahlman might bet as much as $18 million. Mostly he's concerned with races at tracks in New York and California, and he doesn't shy away from the big races. (He took a bath at the Kentucky Derby when Point Given failed, but he figures the horse, which rebounded to win the Preakness, looks good for the Belmont.) Dahlman bets so much money, in fact, that he has to avoid smaller tracks and certain kinds of bets because, in essence, he would be betting against himself.

-->

The Suncoast Hotel and Casino, on the outskirts of Las Vegas, provides Dahlman with an office, and at this moment that's where he is, bearing down hard on the fifth race at Aqueduct. A bank of four television monitors show the action from both coasts, and his desktop computer flashes the probable payoffs on exacta bets, Dahlman's bread and butter. To cash, he must pick the first- and second-place horses in the race, in the exact order (hence the name). It's a high-risk, high-reward bet compared with win-place-or-show wagering. It pays for Dahlman's silver Corvette and his big house with a swimming pool in a gated community near the casino.

"I'm going to put a big box on the five, and do a little bit on the four, because it's a claim by Scott Lake," he says, quickly adding a column of figures with a pencil. The five is Top Bunk, a horse in sharp form that is stepping up in class from slightly softer competition and is the main threat to the horse Dahlman is keen on, Borntoberegal, which, as it happens, he owns a piece of.

A box is a kind of hedge. It's two bets in one (and twice as expensive as a straight exacta), and it lets Dahlman have things both ways. If his horse gets beat by Top Bunk, he still wins, although his net profit will be much less, of course, than if he took the risk of putting all his money on Borntoberegal in the win spot.

And Dahlman is not through. Scott Lake, a very successful trainer from Maryland, has been running horses in New York, at this point without much success. Hemisphere Dancer, which he recently "claimed," or bought out of a claiming race, in which all the horses are for sale, does not look like much. But horses in Lake's hands have been known to experience a religious conversion, and Dahlman does not want to get burned when Lake's horses start firing. (Sure enough, Lake's horses have been winning at a phenomenal rate in recent months.) After placing a $2,400 exacta, Borntoberegal on top of Top Bunk, and a $2,000 exacta the other way around, he constructs a series of defensive bets: a $600 exacta, Borntoberegal over the four horse, Hemisphere Dancer; a $600 exacta, Top Bunk over Hemisphere Dancer; and two $200 exactas, Hemisphere Dancer over those two.

Then he watches the race. Borntoberegal breaks alertly from the gate, sets reasonable fractions and pours it on in the stretch. Top Bunk chases in second. Hemisphere Dancer flattens out in the stretch and finishes sixth. The exacta pays $12, and Dahlman collects $14,000 on a $6,000 bet. It nets him roughly $8,000.

If Dahlman does not look like a gambler, he does not bet like a gambler either. At least not like any gambler I've seen in action, most of them, admittedly, losers. In the two days I watched him bet, he never put serious money on a horse with long odds. Not once did I get the vicarious thrill of seeing him hit a 20-to-1 shot or collect on a $500, $200 or even $50 exacta. He is a plodder. And he talks like one too.

"Winter is is my favorite time of year," he said cheerfully. "It's more predictable. The horses tend to be older, and you know what they're going to do. And out in Northern California it rains, and you get a lot of grass races switched to the main track, resulting in complete mismatches."

Most horseplayers love a contentious 12-horse race with the promise of three-figure exactas and monster trifectas. Not Dahlman. "Anyone who knows anything about gambling will tell you I'm not a great gambler," he says. "What I'm good at is arithmetic. I can add and subtract."

That's the truth, although not the whole truth. Dalhman is in the business of spotting opportunities and minimizing risk. He resembles nothing so much as an arbitrager or commodities trader, only his market is the racetrack and his commodity is Thoroughbreds. His workday is a series of 20-minute market analyses punctuated by 90-second races that yield, when the penciled calculations and racing luck harmonize, a steady profit.

At this point in his career -- he's 58 and has been betting seriously since his teens -- he has amassed the kind of bankroll that allows him to bet very conservatively and still live very well. He compares himself to the early and late Muhammad Ali. In his younger days, Dahlman spent 20 hours a day studying and playing the races, shooting for the moon and hitting phenomenal payoffs. Now, he says, he favors the rope-a-dope approach.

very horseplayer can tell tales of woe. Some of them make you laugh. Read the comment lines in The Daily Racing Form next to the running line for a horse's past races, and you'll get an idea of the strange things that can happen when animals with a brain the size of a prune run at 40 miles per hour around an oval track. My favorites include "stomped duffel bag near line," "savaged foe" and "scattered geese on backstretch." I've seen horses round the far turn and head straight for a parking lot. I held a ticket on a 60-to-1 horse at Belmont that was ahead by seven lengths in midstretch when suddenly, for no reason, it ran right into the rail and sent its jockey sailing into the infield.

The point is, there's no such thing as a sure thing. Favorites win only a third of the time, an immutable racing statistic. Even successful bettors tear up more tickets than they cash. The trick is to cash enough tickets, at the right odds, to offset the losses and turn a profit. That's where addition and subtraction come in.

Of course, the addition and subtraction mean nothing without good information and a refined ability to size up a race and assign probabilities to various outcomes -- the arcane exercise known as handicapping. Dahlman thrives on information. He excels at analyzing a race based on that information, enriched by a lifetime of watching races. It's an art with just enough science to it to make it possible for a very tiny percentage of bettors to take money away from the herd of less intelligent, or less disciplined, bettors.

"I call him the logic doctor," says Chip Taylor, a buddy of Dahlman's and a songwriter whose credits include "Wild Thing" and "Angel of the Morning." Taylor dropped out of the music scene for years to bet on the horses, often with Dahlman. "I was good at analyzing pace and bias" -- the tendency of a racetrack to develop fast or slow areas -- but Ernie picked up on all sorts of things, like when a horse changed from one trainer to another, whether that was a plus or minus, and the answer was not always what you might think."

Dahlman puts it a little differently. "I'm a MOTO player," he says. "Master of the Obvious. It's just that more things are obvious to me than to most people."

Like many other professional horseplayers, Dahlman relies in part on the work of others. The Daily Racing Form gives him a compressed description of the last dozen races run by each horse entered in a race. That merits a quick glance. The more serious numbers come from "the sheets."

Published by a professional horseplayer and Marxist named Len Ragozin, they go a long way toward solving an age-old handicapping problem. By analyzing and quantifying a variety of factors, ranging from track resiliency to wind direction to the distance actually traveled by each horse, they assign a speed rating to a horse's performance.

The Ragozin sheets are shipped to Dahlman every day, overnight, and one of his rare visits to the casino floor is his daily walk past the sports-betting operation on the main floor to a little office that holds his packet of sheets. The raw speed numbers Dahlman accepts as highly accurate. Other assumptions in the sheets he rejects, like Ragozin's opinions about which horses are likely to regress, or bounce, after turning in a stellar performance.

Dahlman also pays close attention to pace, or the time for each quarter mile of a race, which he jots down in one of two fat, spiral-bound notebooks. The other is devoted to his overriding preoccupation, horseshoes. Years ago, Dahlman began noticing something funny about horses equipped with mud calks, cleats that some trainers use for extra traction when rain turns dirt into mud. Dahlman noted that even when rain failed to materialize, a lot of horses seemed to improve several lengths when wearing mud calks for the first time.

He began keeping detailed records, and he now considers it his biggest edge. It's the reason he loves Golden Gate Fields, near San Francisco. It rains a lot there, so plenty of mediocre-seeming horses are switching to mud calks for the first time and then sneaking into exactas at good prices. A second reason for loving Golden Gate is that the track posts very detailed shoe information before each race. Not all mud shoes are created equal, in Dahlman's view. Mud nails, which turn shoes into a kind of hobnailed boot, make no difference, in his opinion. But jar calks, which have a kind of high heel, do. The New York tracks do not provide detailed shoe information, so 10 minutes before every race, the phone in Dahlman's office rings. It's a man named Gene (Dahlman professes not to know his last name), a sharp-eyed informant in New York who stands by the paddock with a pair of binoculars and relays shoe information to Dahlman.

"No one," says Chip Taylor, "has made as much money off shoes as Ernie has."

So why doesn't everyone do the same thing? Because in the handicapping game, not everyone believes the same things. Len Ragozin swears by the bounce. Many handicappers look feverishly for signs of a developing bias. Steven Crist, the editor and publisher of The Daily Racing Form and a big bettor, says that biases are greatly overrated. And so on and so on. Every successful horseplayer has his idiosyncrasies. Dahlman's is horseshoes. And if he's beating the game, who's going to tell him he's wrong?

ahlman grew up in Patchogue, Long Island, where his father, Ernest Sr., owned a delicatessen and spent his free nights betting the trotters at Roosevelt Raceway. Ernest Jr. tagged along. In time he became fascinated by the challenge of finding patterns in the seemingly random circulation of horses around the track. Poring over old programs in his basement, he set about cracking the code.
More...

OddTrader · Feb 27, 2016

http://www.nytimes.com/2001/06/03/magazine/03DAHLMAN.html?pagewanted=all

By the end of high school, Dahlman was hooked. He made a halfhearted attempt at college, enrolling in Alfred University with the object of becoming a ceramics engineer. "It wasn't until I was 40 that I finally found out what a ceramics engineer actually is," he says. He shambled along, then transferred to Wagner College on Staten Island. He dropped out of the economics program when he realized he would eventually have to write a term paper.

His education at the harness tracks was going great, however. In 1964, the summer after he dropped out of Wagner, Dahlman created a stir when he hit a complicated bet at Yonkers Raceway called the twin double -- two linked daily doubles, or four races -- and earned a record payout for the track of $171,084.60. The local press went wild over the whiz kid with the glasses.

Dahlman went right back to the racetrack and hit the twin double the next night. "Twin Double Winner Does It Again!" shrieked the headline in The Daily News. This time, the national press jumped on the story. Time magazine featured Dahlman in its college-life section, headlining the story "Success on the Oval Campus." The New York Post hired him as its harness handicapper for $50 a week. The world lost a ceramics engineer.

"I didn't want to be a horseplayer," Dahlman says. "I know my mother was upset. But I kept making money at it. It's a ridiculous choice of occupation that's worked -- so far."

In the early 1980's, with harness racing in steep decline, Dahlman shifted his action to the Thoroughbred tracks, where he had to learn a brand-new game, one with many more variables. "It's the difference between checkers and chess," he says. Sometimes he won, sometimes he lost. But mostly he won. Even more surprisingly, in a profession populated by misfits, losers and borderline sociopaths, he managed to live a normal suburban life. He married twice and reared six children, none of them gamblers.

He also diversified his portfolio, so to speak. Dahlman and Eugene Hauman, a childhood friend, started buying and running trotters in the 1960's and 1970's. In the early 1990's, they turned to Thoroughbreds, and in 1993 the two men became the leading owners on the New York Racing Association circuit, winning 55 races. Today, the two men own dozens of horses in partnership with Barry K. Schwartz, the chairman and chief executive officer of Calvin Klein Inc. Dahlman's daughter Jennifer Gurney manages Schwartz's 800-acre horse farm in Westchester. He pays taxes like any self-employed businessman, declaring income (winning bets) and writing off losses (losing bets). He bets somewhere between $10 million and $18 million a year. His actual income depends on racing luck. He has had losing years. But on average, he says, he earns 3 or 4 percent on his investment. In a good year, that would give him a pretax income of about $700,000. That does not include income from the horses he owns.

Dahlman lived on Long Island until the mid-90's, when New York State racing authorities dealt him (and other big horseplayers) a stunning blow. They raised the takeout on exacta bets from 17 percent to 20 percent. Takeout is the amount that a track skims off the top of a betting pool to finance its operations. It's the hurdle that bettors must clear before they can earn a profit. The 3 percent increase represented at least $300,000 a year to Dahlman, who did his betting at a teletheater in Happauge. Dahlman did a little adding and subtracting, then took his money and moved to Las Vegas, where the casinos were famous for striking agreeable deals with big bettors.

Dahlman settled in at the Sands. After the Sands was demolished, he switched to the Suncoast, a casino that caters to locals. His arrival was, to put it mildly, unheralded. One day, when the casino was offering free pastries to customers who signed up for a handicapping contest, Dahlman drifted over and asked if he might have a doughnut, even though he wasn't interested in the contest. No dice. Ernie Dahlman, as high a roller as walks into the casino, could not score a 50-cent freebie.

s it happens, Dahlman was mired in a slump during the two days I spent with him at the Suncoast. Consequently, he was betting light, about $40,000 a day. He lost about $7,000 the first day I watched him, and the second was no better.

There were some tough beats. Rocking Ruby, in the ninth at Aqueduct the second day, got stuck behind tiring horses, had to switch paths and by less than a length missed cracking the exacta at odds of 5 to 1. Another of Dahlman's picks was disqualified from a winning exacta after bumping a rival just steps from the finish line.

Most bettors would be chewing the carpet or throwing heavy objects at the television screen. Dahlman reacts to victory and defeat with Spock-like equanimity. Occasionally he will worry out loud. "I'd like to see the five horse closer up," he might say. Or, "Uh-oh, I'm gonna get beat." After the disqualification, he clenched a fist, glared at the screen and said softly, "Gosh DARN it." Then it was on to the next race.

"If I keep my losses at $7,000, I can sleep easily," Dahlman says. "I can get that back in one race."

The playing field is constantly changing, though. An unidentified man betting through an off-track location in North Dakota uses a computer program that searches betting pools for anomalies, then places hundreds of bets on as many as seven horses in a race in a matter of seconds. The "professor," as Dahlman calls him, places huge bets -- including $215,000 on Monarchos to win the Florida Derby -- and as far as anyone can tell, makes money (as he did on that race with Monarchos). "This guy is good," Dahlman says. "Fortunately he does not seem to bet at the same tracks I do."

A few days after leaving Las Vegas, I called Dahlman to see if he was faring any better. "I'm on fire," he said. He sounded almost apologetic about hitting the Pick Six at Aqueduct, a bet in which players must call the winners of six consecutive races. Dahlman put together a $1,000 ticket and, because the pot had swollen after no one hit the jackpot two days in a row, earned $26,000. More important, his day-in, day-out exactas were coming in. Life was back to normal.

His life, anyway. After watching Dahlman in action, I decided to apply his methods. Like many small-time bettors, I sneer at short prices, scoff at exactas paying less than $50 and never bet on a favorite. But I've been having doubts. Why not try to do it the Ernie way?

I plunked down $35 at the one Manhattan newsstand that sells the Ragozin sheets and set about deciphering the numbers for Aqueduct. The next day, I showed up at the Winner's Circle, a teletheater in Manhattan's garment district, carrying the astronomical sum of $1,000 provided by my naive employers, five times the maximum I would bet with my own money. Just carrying the cash made me nervous.

What followed was a date that will live in infamy. To test the waters,

I organized a bet in the second race focusing on a horse, Supernal, that had scored a good sheets number in its last race, when it had raced very wide and rallied mildly for third place. I bet a $50 box using Supernal and Romantic Novel, a horse that had finished second in its last three races. As a defensive measure, I bet a $20 exacta of Miss Mickey, a horse recently transferred to the trainer James Jerkens -- and, I failed to note, wearing mud calks for the first time -- over Supernal. I also did a $10 box, Miss Mickey and Romantic Novel. Miss Mickey, the favorite at 6 to 5, broke quickly and never looked back. Supernal, at odds of about 5 to 1, closed with painful slowness to get the place spot, and Romantic Novel, as the chart for the race put it, "posed no threat." The exacta paid $18. My $20 saver gave me $180, a $30 profit on a bet of $150. I could live with that.

It was downhill from there. Near misses, wild long shots and strange tactical decisions by a few jockeys played havoc with my conservative betting strategy. Wicklow Warrior, which looked as if it could crush the competition in the fifth race, was rushed to the front and contested a hot pace, for no reason that I could see. It tired and finished next to last, taking my $220 with it. I picked a nice-priced winner in the next race, Worthy Crane, but failed to match it up with the right horse in the No. 2 slot. The exacta paid $89 for a $2 bet. I collected zilch on $250 in bets.

I discovered several things very quickly. First, adding and subtracting two minutes before a race is no small skill. I could not do it fast enough to construct a logical hedge. Second, big bets rattle me. When several hundred dollars go down the drain, I start flailing and stabbing, desperate to pull even. By the ninth and final race, I was down to about $10. The money had disappeared faster than butter hitting a hot frying pan. Third, the Dahlman method is not a method, like counting cards at blackjack. It's a lifetime of racing knowledge. And I'm too old to learn.

There was some consolation. I called Dahlman to describe the debacle. He cheerfully told me that he had lost $30,000 the same day. He could afford to laugh. He's well ahead for the year. "The first two races today," he said, "I lost $12,500, and I thought: What am I doing? Nothing is going right; why am I here?"

Then he hit a $213 exacta at Keeneland in Kentucky that erased the deficit in the blink of an eye. "The winner was 18 to 1," he said. "When that happened, I said to myself: 'Yeah, I'm in the right place. I belong here."'

William Grimes is the chief restaurant critic of The Times.
More...

OddTrader · Feb 27, 2016

http://lindaraschke.net/trading-and-betting-the-horses-2/

This Wizard of Odds, “Ernest Dahlman, may be the world’s most successful horse bettor. The reason he’s so good is that he doesn’t gamble.” But, according the article, “in a busy year, Dahlman might bet as much as $18 million.” Let’s look at some of the main principles that contribute to Dahlman’s success and see how they are transferable to trading.

Dahlman narrows his playing field. He specializes primarily “with races at tracks in New York and California” where he perceives himself to have a larger edge. This way, he can become familiar with the subtle nuances such as local trainers (in whose hands a horse can experience a “religious conversion”) and track conditions. A trader will do best to concentrate on a select “stable” of markets and learn their individual personalities, than to jump into markets or individual stocks that have not been thoroughly researched.
Dahlman specializes in a very specific type of bet – the exacta, his “bread and butter.” A trader will do best if he specializes in just one style or pattern or trading methodology. Know the risk/reward characteristics of your individual technique. Dahlman chooses what would at first glance appear to be a high risk technique, but one which offers good rewards.
Dahlman pays more money to minimize his risk. He hedges his bet by creating a box (“twice as expensive as a straight exacta” but which “lets Dahlman have things both ways”). Successful top traders are far more interested in strategies by which to minimize risk than those that go for the big gains. Stops, protective options, or spread strategies are just some of the ways to minimize risk.
Be a “plodder,” a nickel and dimer. “On average, (Dahlman) says, he earns 3 or 4 percent on his investment.” He never bets serious money on long odds. Stick with the high probability setups. Don’t try to hit home runs. “The point is, there’s no such thing as a sure thing. Favorites win only a third of the time, an immutable racing statistic. Even successful bettors tear up more tickets than they cash. The trick is to cash enough tickets, at the right odds, to offset the losses and turn a profit. That’s where addition and subtraction come in.”
Know the seasonalities of your game. “‘Winter is my favorite time of year,’ he said cheerfully. ‘It’s more predictable. The horses tend to be older, and you know what they’re going to do. And out in Northern California it rains, and you get a lot of grass races switched to the main track, resulting in complete mismatches.'” If you are a grain trader, there are distinct times of year when volatility increases. If you trade options, there has historically been a strong tendency for volatility to contract going into the summer – good if you are a short premium player, bad if your game favors a momentum style. Do you do best at the beginning of the year or the end of the year? Know the cycles in both your game and yourself.
Do your daily homework. “Most horseplayers love a contentious 12-horse race with the promise of three-figure exactas and monster trifectas. Not Dahlman. ‘Anyone who knows anything about gambling will tell you I’m not a great gambler,’ he says. ‘What I’m good at is arithmetic. I can add and subtract.'” “Like many other professional horseplayers, Dahlman relies in part on the work of others. The Daily Racing Form gives him a compressed description of the last dozen races run by each horse entered in a race. That merits a quick glance. The more serious numbers come from ‘the sheet'” … which analyze and quantify… “a variety of factors, ranging from track resiliency to wind direction to the distance actually traveled by each horse.” He considers his detailed record keeping to be his biggest edge. What type of detailed record keeping do you maintain with your own trading?
Specialize, Specialize, Specialize! Who would have ever thought of specializing in Dahlman’s “overriding preoccupation, horseshoes? Years ago, Dahlman began noticing something funny about horses equipped with mud calks, cleats that some trainers use for extra traction when rain turns dirt into mud. Dahlman noted that even when rain failed to materialize, a lot of horses seemed to improve several lengths when wearing mud calks for the first time” It’s also another “reason he loves Golden Gate Fields, near San Francisco. It rains a lot there, so plenty of mediocre-seeming horses are switching to mud calks for the first time and then sneaking into exactas at good prices. A second reason for loving Golden Gate is that the track posts very detailed shoe information before each race.” Find your niche!
Believe in your own game and don’t listen to anyone else! Although “Dahlman accepts as highly accurate … the raw speed numbers,” he rejects “other assumptions in the sheets.” Others dismiss Dahlman’s horseshoe theory. It does not matter who thinks your style is right or wrong … all that matters is that you believe in it yourself and follow it consistently.
Keep your losses at the sleeping level. In the best quote from the article, “If I keep my losses at $7,000, I can sleep easily,” Dahlman says. “I can get that back in one race.” Your sleeping level may be different from anyone else’s.

Here is a person who has managed to keep a balanced life and raise 6 children while still maintaining his passion for the horse racing. As the game changed, from trotters to thoroughbreds, he was able to adapt with the times. We have seen many games come and go in the markets, from equity options arbitrage to the disappearance of SOES traders. We have seen markets change from a momentum environment to a trading-range game. A trader must specialize in one thing yet be ready to recognize when it is time to learn a brand new game. Recognize that Dahlman built up a lifetime of racing knowledge. He relies on that experience to interpret the incredible amount of information he gathers. For newer traders just learning to make their way in this business, recognize that, ultimately, experience is the best teacher. Every day that you trade, you gain experience. Keep the discipline, keep the faith. To repeat one last quote: “It’s an art with just enough science to make it possible for a very tiny percentage of bettors to take money away from the herd of less disciplined bettors.”

Trading? or betting the horses?

Linda Raschke
LBRGroup, Inc.
More...

OddTrader · Feb 27, 2016

http://www.ms.unimelb.edu.au/documents/thesis/AlexThesis.pdf
Profitable Strategies in Horse Race Betting Markets
More...

https://www.stat.berkeley.edu/~aldous/157/Papers/ali.pdf
Probability models on horse-race outcomes
More...

Predicting Horse Race Winners Using Advanced Statistical Methods
More...

http://www.nessis.org/nessis07/Victor_Lo.pdf
Probability and Optimization Models for Racing
More...

OddTrader · Feb 27, 2016

http://bleacherreport.com/articles/27781-cracking-the-horse-racing-code
Cracking The Horse Racing Code

Yes, I have been away for quite some time now. I have been saying that I had something worthwhile up my sleeve, but it has not been ready to put into words.....until now.

For months I have been researching horse racing. It has not been research to find great horses, jockeys and trainers, although that was an inevitable side effect along the way, or to find a better feel and respect for the tradition, another glorious side effect, but to predict horse races more accurately than the morning line odds, and find the probabilities of each horse coming in first. In other words, I approached horse racing with a gambler's eye, a microscope, a Daily Racing Form, and a calculator, to see just how close I could come to predicting the outcome of a horse race on a regular basis.

I completely understand that succeeding in this endeavor could ruin horse racing as it exists. If the tracks cannot make money, then they will no longer hold races. The sport could collapse. Really though, it doesn't have much further to fall. Name the jockey that rode the winner of last year's Kentucky Derby. Name the Horse that won last year's Kentucky Derby. Precisely. I'm sure Street Sense doesn't care that you do not remember him, but Calvin Borel should be upset. Seriously, only the Triple Crown gets people talking, and only a terrible accident like Barbaro can get people to remember a horse anymore. On the other hand, if the betting public shows an interest in the sport......making money has a tendency to do that to people, and then the sport of horse racing will have a revival. Did the MIT blackjack team destroy cards? Absolutely not. In fact, the time following the MIT blackjack team saw resurgence in the game. Horse racing can have that resurgence too, and it won't take a Triple Crown winner to do it. People just need to think they are smarter than the game and believe they can win. The money can be made up to a degree in television ratings and advertising. If people think they have an edge over the house, they will pay very close attention. As soon as the system starts getting beaten, the people in charge of the track will change it. The house is always designed to win. Horse racing will be fine, especially after finding a new casual audience of supporters. They get their rake no matter what. All it takes is a formula that will produce winners with consistency. A formula that can pick horses better than just looking at the morning line odds would make people feel like they have a huge advantage. Now I have my task.

The base of my research was academic journals. Curiously enough, there are more than a few scholars who have been enthralled enough by the sport and the math involved to have published there own works on the subject. In “Searching for Positive Returns at the Track: A Multinominal Logit Model for Handicapping Horse Races”, by Ruth N. Bolton and Randall G. Chapman, I found the information that would carry me through this project. Bolton and Chapman set out to find which variables would be most important to consider when evaluating the horses in a race, and predicting the outcome. There research suggested that "average amount of money earned per race in the current year" and "average speed rating over the last four races" were the two most important factors. "Lifetime win percentage" was also considered a significant variable, but not so much as the first two. The shocker to me was that jockeys, post position, and weight were deemed inconsequential for the most part. The best jockeys were often put on the best horses, and their correlation nullified much of their value.

Armed with this new information, it was time to make an equation of my own. I have the variables deemed most important by Bolton and Chapman, and needed to weigh them. In the case of betting on a Maiden race, where none of the competing horses has ever won a race in its life, I decided to change the "lifetime win percentage" to "lifetime in the money percentage", or how many times the horse has finished in the top three in the number of times the horse has competed. (Please note: All of the statistics that I use in this formula can be found in he Daily Racing Form, and you can use this the next time you go to the track.) Of my three important variables, "money earned per race in the current year", and "average speed rating over the last four races" were the most important, while "lifetime win percentage" was significant. To weigh them based on ten, the first two, which I will label "$/race" and "AVSPDRT" were given the value of four, while the last, "LifeWin%" is given the value of 2. 4+4+2=10, and all is right with the world. Each horse is then rated by these past performances in comparison to the other horses in the race. Given, in a three horse race, that horse #1 has an AVSPDRT of 64, horse #2 has an AVSPDRT of 61, and horse #3 has an AVSPDRT of 58. Horse #1 earns 3 points for having the highest AVSPDRT, while horse #2 would earn 2 points and horse #3 would earn 1 point. (If it were a four horse race, the top horse would earn 4 points, a five horse race, 5 points, etc.) This point-based ranking would be done for $/race and LifeWin%.

Ok, are we all still together on this one? It's ok to go back and go over that again. It took me a few tries myself. I'd hate for someone to fall astray, get a bad ranking system, and then come back and yell at me because they lost a lot of money. We're good? Moving on then.

Let us suggest that the equations were performed for all the horses in an upcoming five horse race. After each horse was given points based on its ranking in those variables, they stand as such:

Horse Number__pts from $/race__pts from AVSPDRT___pts from LifeWin%
1________________1______________5_____________1
2________________4______________3_____________5
3________________3______________1_____________2
4________________5______________2_____________4
5________________2______________4_____________2

This is based on an actual race from the Aqueduct on April 24th. The #3 and #5 horses both have a 2 in the last column because they have the same LifeWin%. Now, remember how we weighted the values from before? I wasn't just screwing around with nonsense. Multiply their points in each category by the values they were given earlier. Pts from %/race is multiplied by 4. Pts. from AVSPDRT are multiplied by 4. Pts. From LifeWin% is multiplied by 2. Let us look at how they stack up now.

Horse Number__________________New Point Total
1_________________________________26
2_________________________________38
3_________________________________20
4_________________________________36
5_________________________________28

Horse #2 is the favorite, with horse #4 being the second favorite, and the rest of the horses straggling behind. This is your own independent way to calculate the horses on your own. The best part is that now we can take it one step further. With the information you have in that last table, you can calculate what percentage of the total points each horse has, and have a rough estimate of what percentage each horse has of winning the race.

In the case of horse #1, it has 26 of a possible 148 points, 17.5%. Should we have it in a table? I say yes.

Horse Number________________% chance of winning
1______________________________17.5%
2______________________________25.6%
3______________________________13.5%
4______________________________24.3%
5______________________________18.9%

In the case of a small race like this, every horse has a significant chance of winning.

Important Note: Morning line odds can also be put into percentages like this. The equation is even easy to remember. Add 1, and then divide 100 by that number. (Example #1 3-1 odds: 3+1=4 100/4=25% chance example#2 (this one is tricky) 5-2 odds: since it’s a two, you add that to 5, then divide the new number by 2........5+2=7 7/2=3.5 100/3.5= 28.57) After all the odds have been converted to percentages, you must add them all together. They will always come out to more than 100%, because the track takes into account their own rake, plus about a one point margin of error per horse in the race. In order to get the percentage back to 100, divide 100 by the sum of the morning line odds percentages. (Example: if the sum of the morning line odds percentages is 125, then 100/125 =.8. Multiply all the morning line odds percentages by .8 to find their true value. (Example: 25% becomes 20%)

Now you can compare the track percentages to our own, to see which horses you feel differently on. If you have a horse that you feel has a 35% chance to win, and the track is giving it a 25% chance to win, you will make more money betting on it correctly than you would if you were betting correctly on a horse that the track felt was better than your evaluation.

Finally, when betting exotically, such as exactas, trifectas and superfectas, take into account each horse's "lifetime in the money percentage". A horse may not win many races, but often finish in the top 3. That horse, even if it ranks lowly on your list of winners, may need to be put in your exotic bet, and forgetting those horses could cost you. Also, if after all this analysis, you are still deadlocked on which horse is the favorite, and then go to the jockeys and trainers. In a close race, a better jockey could make the difference. Even if they had been an afterthought until now, you would be wise not to forget them when you need just one more variable.

Now that I made you read the entire thing without telling you if it works, allow me to say.........kind of. I have only tested it over the course of nine races, and much more data needs to be collected. It is basically a lock that I will be running this strategy everyday that Saratoga is open, and a definitive answer should be ready by the beginning of September. Until then, I'll let you know that over the course of nine races at the Aqueduct, my strategy yielded more horses finishing in the money three times, the morning line odds yielded more horses finishing in the money four times, and we had the same number of horses finish in the money two times. My total number of horses finishing in the money over the course of the day was 14 of a possible 27, (one was a technicality, because I picked the 1A horse and the 1 horse placed, but since they are tied together, I would have still gotten the money), and the morning line odds had 14 1/2 of a possible 27 horses finish in the money, (they get a half point because in the second race, three horses scratched, leaving four to run. The morning line odds gave their bottom two horses the same odds. Both finished in the bottom two, but since one had to be third, they were right on a technicality.) So with a technicality on both sides, the track beat my number of in the money selections by 1/2. Only time and more races can really determine which system is really better, and since I have time to tweak mine before Saratoga, I am feeling pretty good about coming out on top.

If this all works out, and people get wind of my system and start using it, racing could find the interest it has been losing slowly since Affirmed last won the Triple Crown in 1978. If I were you, I'd back this horse to win.
More...

OddTrader · Feb 27, 2016

http://cs229.stanford.edu/proj2007/Kempston-HowToWinAtTheTrack.pdf

How to Win at the Track

Cary Kempston
cdjk@cs.stanford.edu
Friday, December 14, 2007

1 Introduction
Gambling on horse races is done according to a pari-mutuel betting system. All of the money is
pooled, the track removes its take, and whatever is left is divided among the winning bettors. A
consequence of this is that the final odds are not known at the time of betting, since later bettors
could influence the odds. Usually, however, estimated odds are posted based on the money wagered
so far[4]. Another consequence of pari-mutuel betting is that all bets have a negative expected value
if one uses the “market” odds posted by the track. This negative expected value is a consequence of
the track take. Thus, to develop any successful gambling model of horse racing, one must develop
better estimates of the horses win probabilities than the posted odds[2].
The other necessary component of a horse race betting system involves figuring out exactly how
to bet. This sounds like an easy task, but there are many different possible types of bets. One can
wager on a horse finishing first (win), first or second (place), first, second, or third (show), as well as
many other bets involving multiple horses (quinella, trifecta, exacta, superfecta) and multiple races
(daily double, pick 3, pick 6) [1]. Determining the optimal strategy in the face of these complex
betting choices is beyond the scope of this project, so I will only consider simple, single-horse and
single-race bets.
These two components would comprise a complete horse race betting system. This project,
however, is concerned only with the first part of this system: developing a method to predict the
winners of horse races.

2 Data
My data comes from a one month TrackMaster subscription[5]. This allowed me to download data
about every horse race from twenty-three tracks from September 20, 2007 until November 26, 2007,
comprising 58565 different races and 6827 finishing horses. I only considered thoroughbred horse
races, since that is the most common type of horse racing in my data. I also only considered horses
that finished, since my model has no notion of not starting or not finishing a race.
2.1 Features
My data consisted of information about each horse in each race. The features I selected out of this
data are given in Figure 1.
1
Feature Explanation
Previous Wins by Horse Percentage of races won by this horse
Number of Races by Horse
Weight Carried Weight of jockey and saddle
Horse Age
Horse Sex One of: Colt, Filly, Gelding, Mare, Stallion
Medication Given Bute and/or Lasix
Previous Wins by Jockey Percentage of races won by this jockey
Number of Races by Jockey
Previous Wins by Trainer Percentage of races won by this trainer
Number of Races by Trainer
Post Position How far from the inside of the track the horse starts
Final Odds Final odds for this horse to win given by the track
Track Conditions One of: Freezing, Fast, Good, Heavy, Muddy, Snow,
Slow, Sloppy
Figure 1: Features for machine learning algorithms
All features that have values outside of [0
,
1] are normalized to be within that range. The past
performance history is only for the time period that my data spans, as I was unable to find a
source of win data by horse that was both comprehensive and inexpensive. The values in the horse
sex category are Colt (young male), Filly (young female), Mare (older female), Gelding (castrated
male), or Stallion (older male). That feature, and the Track Conditions feature were encoded as
binary variables - i.e. each sample has a feature that is 1 if the horse is a filly, and 0 if not, for each
of the five possible values of horse sex and eight possible values for track conditions.
As target output value, I have both the finishing position of the horse, and the lengths behind
the leader that the horse finished.

3 Approaches
This project models horse races. There are some number of entrants in a race, and they cross the
finish line in a ranked order. There are a number of machine learning approaches that can model
races, including ranking algorithms (i.e. ordinal regression), classifiers, and regression. All of these
approaches assume independence of each horse’s chance of winning; the other horses in a race are
not involved in any probability calculations.
3.1 Ranking
Initially horse racing seems like a natural place to use a ranking algorithm or some sort of ordinal
regression, which, given a training sample, tries to learn it’s ordered rank. In this case, the rank
would be the finishing position of a particular horse. Using an ordinal regression classifier would
then involve giving it the feature vectors of each horse in a race, and having it predict the finishing
place for each horse. There is a problem with this approach, however, because of the independence
assumption. Ordinal regression would output that a particular horse is a “fourth place” horse in
general, ignoring the other horses in the race. This could result in either multiple predicted first
2
place finishers, or no predicted first place finishers. I discovered no easy way around this problem,
so I did not implement any ranking algorithms.
3.2 Binary Classification
Binary classification is a more promising approach. This approach learns a classifier on the data
that predicts whether a horse will win the race or not. This algorithm still faces the same problem
as trying to rank finishers, in that multiple or no winners could be predicted, but most classification
algorithms give us some measure of the confidence of the prediction (such as the distance from the
separating hyperplane in an SVM). After running a classifier on every horse in a race, we can
predict the winner by picking the horse that the classifier has the highest confidence of winning.
This approach is shown schematically in Figure 2.
Figure 2: Using individual binary classifiers to predict one winner.
3.3 Regression
The final approach is using regression to predict how far behind the leader a horse will finish. After
predicting this for each horse in a race, it is an easy matter to sort the horses by the distance behind
the leader and pick the smallest one as the winner. The schematic for this looks very similar to the
classification approach, but with “regression” replacing each box labeled “classifier,” and sorting
by the output of the regression instead of confidence.
3.4 Error Rate
Since the data is on a per-horse level, but we are really interested on a per-race level, some work
is required to determine the error rate. Instead of counting every win/loss misclassification as
an error, we should instead go through the data for each race, predict the winner based on the
classification approach described in Section 3.2, and count the prediction as correct if the picked
horse won, regardless of any other horses classified as winners.
Furthermore, it is possible that our approach may not be good at predicting the winner of a
race, but is still good at predicting “fast” horses. Since there are win, place, and show bets, we
3
Finishes in top
n
Probability horse wins
1 12.6%
2 22.7%
3 33.6%
Figure 3: Success rate of na ̈ıve prediction strategy.
can calculate the error rate based on the horse coming in first, first or second, and first, second, or
third.
To determine whether these approaches are successful, consider a na ̈ıve prediction strategy that
involves predicting the winning horse by picking the horse with the most favorable odds. The
results for this strategy are given in Figure 3.
3.5 Method
All learning (classification and regression) was done with support vector machines, using both
linear and gaussian kernels. I used
SVM
light
[3] for all SVM calculations. Cross validation holding
back 30% of the data was done to generate the test success rates. In experimenting with different
parameters, I found that using
C
= 10 and
σ
2
= 0
.
1 (for Gaussian Kernels) to give reasonable
values for the classification problems. I used values
C
= 1 and
σ
2
= 10 for the regression problems,
although it turns out that regression is not a good approach for this problem, and all values of the
parameters are equally bad.

4 Results
Initial results are given in Figure 4. These results are not very good. For the regression tests,
the test set success rate is higher than the training set success rate, while for classification the the
training set success rate is significantly higher than the test set success rate, suggesting overfitting.
n
= 1
n
= 2
n
= 3
Kernel Type
Train
Test
Train
Test
Train
Test
Regression Linear
3.0
12.4
14.0
22.1
27.6
32.8
Gaussian
2.9
12.3
12.9
21.2
27.1
33.6
Classification Linear
59.7
13.6
67.1
24.0
73.6
36.8
Gaussian
60.0
13.9
67.4
24.2
73.5
36.6
Figure 4: Training and test set success rates (%) using all features (
n
is the place the predicted
horse can finish and still count as a success).
The performance of the regression tests is not very good, most likely because the output variable
(i.e., the finish distance behind the leader) is not consistent across races. For example, a horse could
come in second in two races, but finish right behind the leader in one and a long distance behind
the leader in another. These outcomes should be treated the same, but they are treated very
differently by a regression algorithm. Regression is therefore taking some element of “competition”
4
into account (since finishing immediately behind the leader suggests a more competitive race), but
not in any meaningful or consistent way.
The success rates of the classification experiments suggests overfitting. The most likely cause
of this is that I only have win/loss history for each horse for the 67 days that my data spans. This
data includes results for 29409 horses, of which approximately 40% only raced once. This resulted
in a large number of my training features about the win/loss record of the horse to be either 0%
or 100%, which is not good data for probability calculations.
To correct for this likely overfitting, I repeated my calculations with all the win/loss data
removed. These results are given in Figure 5. Removing the previous win/loss history corrects the
overfitting for the classification tests, but the problems in the regression tests remain.
n
= 1
n
= 2
n
= 3
Kernel Type
Train
Test
Train
Test
Train
Test
Regression Linear
5.3
12.5
10.3
21.9
18.5
34.4
Gaussian
5.0
12.0
10.9
21.4
20.2
32.6
Classification Linear
17.1
13.0
29.2
22.6
43.2
34.3
Gaussian
13.7
13.2
23.8
23.0
36.2
34.5
Figure 5: Training and test set success rates (%), excluding all previous win features (
n
is the place
the predicted horse can finish and still count as a success).

5 Conclusion
My results show that machine learning techniques can be used to better predict the outcomes
of horse races than the odds. I was able to predict first, second, or third place finishes .9%
more often than the na ̈ıve strategy of picking the horse with the best odds. Although this is a
small improvement, the large size of the horse racing industry needs to be taken into account in
interpreting the results.
References
[1] Glossary of horse racing terms.
http://www.drf.com/help/help_glossary.html
.
[2] Ruth N. Bolton and Randall G. Chapman. Searching for positive returns at the track: A
multinomial logit model for handicapping horse races.
Management Science
, 32(8):1040–1060,
1986.
[3] T. Joachims.
Making large-Scale SVM Learning Practical. Advances in Kernel Methods - Sup-
port Vector Learning
. B. Scholkopf, C. Burges and A. Smola (ed.). MIT Press, 1999.
[4] Pari-mutuel betting.
http://en.wikipedia.org/wiki/Parimutuel_betting
.
[5] Trackmaster.
http://www.trackmaster.com
.
5
More...

OddTrader · Feb 27, 2016

dtrader98 said:
http://www.elitetrader.com/et/index.php?threads/betting-on-horse-races-vs-trading.133436/
betting on horse races vs. trading

It's funny to watch all these gambling analogies to trading (poker, blackjack, etc...) surface from time to time. I was recently reading a book on how modern horse racing handicappers succeed.

The funny thing is, these guys go through the same types of debates that some of the trading forums have. Some say handicapping (trading strategy) is more important than money mgmt. etc. But all pretty much agree, both are important.

Let's face it. When it comes down to it, trading is gambling. The more you understand that relationship, the better off you are IMO. I wish I had studied these types of books/strategies long ago, rather than wasting time on the proliferation of cherry picking TA visual arts books.
Not that pattern recognition TA is completely devoid of merit, but the subjective approach that the far majority of authors seem to advocate is sorely lacking IMO.
More...

dtrader98 said:
excellent example of the type of thinking I was promulgating. Thanks for mentioning.

Here's a byte from his 'scenarios for global risk management and global investment strategies'.
More...

OddTrader · Feb 27, 2016

etfarb said:
http://www.elitetrader.com/et/index...se-betting-and-trading-have-in-common.266164/
Can somone explain to me what horse betting and trading have in common?

I get it there both forms of gambling but can someone give me some direct examples of how the two are similar
More...

Unequivocaltim · Feb 28, 2016

I was a co-owner of a handful of runners.
You have no advantage over the others as they also incorporate their plot for the race.
The advantage to trading is, all the numbers and sentiment are presented to you before you make a trade.
T-breds are as unpredictable as your next day of life.

Na zdrowie,
Tim