Our Innovators in Racing Series is brought to you by our partners at LTN Global. LTN Global is a Technology and Production Company that is helping racetracks create and distribute content at both high quality and good value. They offer distribution services to get tracks seen in more online and offline spaces than ever before. Visit LTNGlobal.com to learn more.
I first became aware of “computer players” in horse racing in the early 2000s, when working on a book project called Six Secrets of Successful Bettors. I even was able to interview of the OG computer players for the project, Alan Woods. In the intervening years, computer groups have become responsible for more and more of the pools in U.S. racing and I suppose that’s just the nature of markets – when something works, more and more players are going to gravitate towards the methodology that proves effective.
There have certainly been plenty of rank-and-file horseplayers who rail against the computer groups – as well as smart efforts to limit their pool access and ability to make wagers faster than so-called retail players – but I’ve always had some sympathy for the point of view that I first heard espoused by Marshall Gramm. I recall his saying that what we need to do as an industry is not so much eliminate the computer players, but to give everyone the same tools so more people can operate as they do.
When we first conceived the idea of our Innovators series, I immediately thought of the Dark Horse app, currently available in Canada for HPI Bet customers. I spoke with Chris Lush, the Senior Vice President of Technology, Wagering and Broadcast at Woodbine Entertainment Group about the app. He described the idea of the app as a way to broaden racing’s appeal.
“Part of our problem in racing is that our existing demographic is getting older and it’s harder to attract that younger player,” he told me on a recent podcast. “So we started to focus on a more easy and intuitive app and to make it available on mobile platforms.”
His research told him that the younger audience was attracted to data – something our sport has in abundance. And the very nature of our wagering also affords us an opportunity to help players use and refine that data.
“The interesting thing about parimutuel wagering is that we’re the host, whereas in a fixed odds environment you’re betting against the house,” he summarized, “but here it’s more like daily fantasy sports where it’s player against player.”
This is the critical distinction. It’s very much in horse racing’s best interests to directly help our newer players win, in a similar fashion to the aforementioned computer groups. “[These computer groups] are using advanced algorithms to try and arbitrage the parimutuel liquidity in the pools,” he laid out. “So we thought if we can use data sets, algorithms, and even AI to build scenarios and strategies for a new customer.”
Not surprisingly, it was complicated process. “When we started running our algorithm, we were actually using AI to find 105 different weighing points that determine the outcome of a race. We threw these points into a model and told the model to learn how to choose a winner.”
The model was given over 10,000 historical races to analyze. One note I found interesting was that part of the Dark Horse mission was to find specific races where a bet might be found, rather than just an answer to every question, as it were.
“Instead of just giving you a synopsis of each race, we give you a list of races we assess as being bet-able,” he said.
Another point I thought was great was that the app isn’t meant to stop on the handicapping side of things. It’s not merely a “black box” spitting out each runner’s presumptive chance of winning. It’s meant to help you wager more intelligently.
“[The model] goes ahead and predicts what it thinks will be the top horses and conversely, we used another algorithm to bundle bets on these top horses that were predicted,” Lush continued. “The data from the live liquidity in the pools as well as true and projected odds are used to generate a mix of bets designed to give a positive result with 70-80% more efficiency in wagering.”
The strategies are designed to generate a positive payout and they go way beyond “caveman tickets,” incorporating ideas like dutching combinations and playing around with bet sizing, both things JK and I have spent plenty of time talking about on the shows.
“At Woodbine for example there is a dollar minimum for a bet type but a 20-cent minimum bet increment,” he continued. “In that case we build a exacta list, each bet will produce different values assigning probability values to each.”
The player has control over how much he or she wants to outsource the handicapping or the betting.
“The system asks for several horses,” he said. “After you select your horses, we build a bet strategy for you and occasionally the AI engine will throw in an extra horse if it is needed. We have customers who win bets they never thought they had the horse!”
I also love that like the computer groups themselves, the model refines itself with real-time tote information.
“It skims the pools analyzing all the tote information coming in to best make its decision,” Lush told me.
His background came working on the technology side in the finance world.
“When I was at Toronto stock exchange, there were a lot of computer trading companies who built very sophisticated algorithms and they were arbitraging the markets with direct fiber optic links in to the exchanges and they were looking for signals that the bid was going to move up while the ask was staying flat. They would put bids in right before the market would bid and arbitrage it and finds that difference.”
A wiseguy friend told me years ago that these days an aspiring horseplayer should read Michael Lewis’ Flash Boys in addition to the classic books of Andy Beyer. Lush’s next words made me think back to that conversation.
“What you see is these computer wagering teams coming in and they have these set values of what they think a horse is worth, and right before a race goes off, if they see that variance, they arbitrage that,” he explained. “The algorithm says this horse, for instance, is 2-1 and at post time if that horse is 3-1 then it automatically puts a bet in faster than any human could do so. That’s done knowing what that final odds will be, and making the bet.”
Lush believes humans can still find a way to compete with the machines, both in finance and racing. “There is still room for customers and room for players in the game,” he assured me. “Right now, liquidity in most of our pools runs anywhere from 15-20% computerized wagering companies and the rest is pure retail. These computer wagering teams don’t bet on every race and, secondly even when you have the right bet there is only so much you can bet before you start taking money away from yourself.”
I’ll note that I firmly believe the percentage of computer money in the typical North American pool is much higher than that, like maybe 2x even, but with his access, you’d think Lush must be accurate when it comes to the Woodbine numbers.
I ended with the obvious question: will we see this tech south of the border any time soon?
“We have had discussions with American companies to bring this in [to the states], but that’s not our first goal,” he told me. ““The dark horse app is about leveling the playing field.”
To listen to my conversation with Chris, check out this podcast.
