Predicteform Intro (4/16/14)

Paul Bessire, General Manager, PredictionMachine.com
PredictionMachine.com General Manager Paul Bessire weighs in with his thoughts on the launch of Predicteform.com and the differences between handicapping horse racing and sports.

I have incredibly fond memories of trips to Arlington Park just outside of Chicago with my father soon after it reopened in the late 1980s. Barely in elementary school at the time, I loved watching the horses run, seeking out concessions and cheering for whomever my dad told me to - all while my dad would try to teach me how to read a racing form. In a slightly less glamorous way, I remember the same visions from trips to an Off Track Betting parlor in Rockford, IL. Not too surprisingly, given what I am now doing with my life, my dad was always looking for value, a way to think outside the box and win against the competition (you should have heard our nightly conversations about our Cub Scout Softball lineups when I was in second grade). Thoroughbred horse racing was and now is again an avenue for that intrigue and challenge.

My dad got hooked on racing during his days stationed next to Golden Gate Park in the Army. That continued after he became friends with, Frank, a retired teacher (my dad was also a teacher) from my hometown (Janesville, WI) who loved to go to the track and needed a ride on the weekends to Arlington and back. Not too long after I remember meeting Frank, he became ill and passed away. My dad lost someone he could talk to about horse racing all day and the OTB (now closed) was not necessarily my mom's favorite place for him to take his three sons so we stopped following the races and my dad filled that void by coaching us in just about every sport growing up - and later competing against us in fantasy sports leagues.

When I first got into this business a decade ago and more so since starting PredictionMachine.com, many, many people have asked me about analyzing horse racing. My dad, now a retired teacher in his own right, still playing fantasy sports with his sons and their friends, is back into racing - following TVG on his DirecTV package (part of the reason my parents have satellite TV) and incessantly checking Twinspires.com - has been an ardent voice among that group. Though I have been interested in the sport and live in Southwest Ohio, just outside of several top tracks, I continually resisted. Building, tweaking and improving six engines for six sports in what I believed to be more exploitable markets is tough enough, not to mention maintaining essentially 600 rosters of teams in the sports that we currently cover. That has certainly been enough work to keep me swamped during my career. And, having worked on a project with TVG earlier in my career and understanding the horse racing market to the degree that I did, I knew that there were many challenges facing such a project that extended well above and beyond sports wagering (more on that later).

However, the team at PredictionMachine.com has been growing (so it's not all or even mostly on my shoulders) in a good way, technology is improving in general and we have done what we can to keep up with it to the point that automating much of what we do on PredictionMachine.com has eliminated much of the maintenance (though no pick still ever hits the site without human review). And as that was happening, though through unfortunate circumstances, a tremendous opportunity came to the PredictionMachine.com team to apply our skills to analyzing and handicapping horse racing.

Sports betting and horse racing are very different markets. That's probably obvious to most, but here are some of the key ways that is so:
  • Fixed Price vs. Parimuteul Wagering - Among wagering options, the most noticeable difference between the two options is that sports books are held to the price (spread, total, money-line, etc.) of the ticket purchased by the individual for an event, while horse players see their wagers pooled together with the house taking a fixed vig and the rest proportionately divided among those with horses finishing in the money. Odds for horse racing are not fixed and the likely payout is actually not known until after the race. It is interesting to note that, in most European countries, wagering on horses is done at a fixed price. For these reasons, we are not able to actually make picks with Predicteform.com because exact value is not as clear, but we can be very good at identifying horses likely to improve or regress in the next race (which is certainly valuable).
  • Legality - Sports betting is essentially, currently only legal in the U.S. within the state of Nevada. Horse racing can be wagered on legally at any track in the country, at various off-track wagering venues or online through sites like TwinSpires.com.
  • Classes vs. Leagues - In sports, there are specific leagues that are well-known with published schedules. Though other levels feed into the top level, wagering is usually only available at the top one or two level (and levels do not mix). In horse racing, schedules are rarely finalized until the week of the race and changes can still occur after that. Horses tend to race in groups by age, class and gender but overlap can exist and class is not really well defined.
  • Accessibility of Data - This may be the biggest issue from my standpoint. Sports data is widely available to anyone with an internet connection. It goes back more than 125 years and is robust with many of us keeping information in check. In most sports, there are fewer than 2,000 players to track at any wagerable level over the course of the season. In horse racing there are upwards of 300,000 horses each year. The horses that race in the U.S. could have raced anywhere previously against anyone. Finding that information and validating it is not only incredibly difficult, it comes with potentially substantial costs.
  • Technology - This surprised me. Though one would think that a sport for which online wagering is allowed and that has been spurred on by gambling for decades would have some of the more advanced analytics and technology to track information on everything happening in any race. That is not the case in horse racing as I would rank it behind the four sports we cover - football, basketball, baseball and hockey - in technological innovation and analytics. In many ways, these last two notes are why we also have been hesitant to grow globally into soccer markets - so much information to compare and data validity concerns without uniform tracking technology or process.
  • Interaction - All of the sports we cover leverage simulation to gain an edge because that is the best way to account for interactions between competing players and teams. We don't dive into individual sports like golf or most Olympic sports because this interaction does not exist (my joke used to be that if we simulated golf, Tiger would just be the favorite every time). Horse racing may have the interactions between the jockey, horse and trainer, but it's still not an interactive sport (thus simulation may not be the best approach).
With that in mind, here are some ways in which sports wagering and horse racing are similar - most of which only became evident to me while delving deep into Cary Fotias math behind what is now Predicteform.com:
  • Weather and Surface - This may not apply to hockey and basketball, but understanding the impacts of weather, field conditions and dimensions is very important to gaining an appropriate edge in football and baseball, just as it is in horse racing.
  • Motivation - At its core, horse racing is a sport in which the objective by those playing it is to win. Motivation can be a concern in horse racing, since the possibility exists every horse in every race may not be competing to win. Like in sports though, it is possible to research and understand these situations, to remove bias and appropriately handle data moving forward. Also, for the most part, participants in every sport are trying their best to the point that motivation concerns are often outliers.
  • Relativity - Though data issues may be apparent, how horses fare relative to each other can put context around performance, which allows analysts to account for data biases. The same thing can and is done in sports (I've basically spent half of the last decade analyzing the relative context of team and player performance). This was key to realizing that both the approach created for Predicteform.com is sound and analyzing horse racing can be very similar to what we do with sports.
  • Exploitable Markets - Just like sports fans tend to wager with their hearts and their eyes, horse players develop exploitable biases on which an objective approach can capitalize.
  • Biology - Similar to players, horses experience cycles of biological growth and improvement (we call these Form Cycle Patterns at Predicteform.com and they are crucial to getting the most out of our analysis).
  • Objective Analysis - Ultimately, the combination of enough trust-worthy data, technology, objectivity, an understanding of conventional wisdom, a willingness to look outside the box and trust a strategy and passion can lead to very powerful ways to gain an advantage over any market in any industry.
After renowned and respected horse racing handicapper Cary Fotias passed away last summer, we were presented with a unique opportunity that, in the end was too logical to pass. We could take a site called Equiform.com, for which Cary would spend around six hours every day calculating Pace Figures and Form Cycle Patterns for just a handful of tracks and utilize our mathematical and technological skills to open the process up to frequently updating, incredibly thorough and tremendously beneficial objective information that can be used by anyone headed to the track or playing at home.

I personally, along with our lead developer (who admittedly had far more to do with translating Equiform.com into Predicteform.com than I did), combed through the code to vet the math behind the process. As a sports analyst for the last decade there are a few things I knew to look for. Normalization? Check (by track, surface, distance, class and more). Removing strength-of-schedule bias? Check. Scaling to pace? Check. Understanding and projecting regression? Big check. Ways to handle missing data? Check. Fitting individual data to an expected progression curve? Check. It is all there and then some.

A couple of tweaks have been necessary, but solely due to the ways in which Cary (and his team, which often included, Bruce Kittle, also a retired teacher and an amazing ally in this process) was previously making periodic changes manually. For instance, starting with races taking place on April 9, Par Times, the average expected time in seconds that it takes the leading horse in a race to reach each tracked distance (two furlong, four furlong, six furlong and finish) broken down by track, surface, distance and class, will be dynamically calculated with each race. Cary had been updating these manually. We found 98%+ of them to be within 0.5 seconds of the most recently calculated values. Par Times are used to evaluate daily variance due to any other factors, which can then be used to account for any bias in a horse's time. Unless technology or data changes to a point that warrants modification, though, this approach makes sense and looks rock solid to me. I do not see us stepping in to alter it more.

With the support and trust of Brisnet, a parent company of TwinSpires and one of the world leaders in horse racing data, who also believe in the Equiform/Cary Fotias approach to Pace Figure calculation and Form Cycle Pattern analysis, our data concerns have been alleviated as well. This is an amazing opportunity, not just for me personally as I reconnect with images of my youth while aiding in one of my father's greatest passions (he calls Predicteform the "Holy Grail" of horse racing handicapping analysis - account for that bias as you wish; for all I can tell, he is right), but as we get a chance to use the principles of PredictionMachine.com to find value in another venue altogether. The more I have thought about and looked into this challenge, the more I have realized that the ways in which these areas are similar greatly outweigh the ways in which they are different and that we are a perfect fit for Cary's approach. I just wish I could have thanked him for that realization in person...

If any of this is remotely interesting to you, I recommend checking out Predicteform.com (we have a new Tutorial Video as well). Send us a note over there with any feedback or questions.

Comments

Search Blogs

Keyword:

Archive

Ad

The Predictalator

The Predictalator is the most advanced sports forecasting software available today because it has the ability to account for all of the relevant statistical interactions of the players (playing or not playing/injured), coaches, officials, fans (home field advantage) and weather in each game.

04/14/2014 Highlight: Using the ResultsFinder for the week of April 7th - 13th, one could find that all highlighted, MLB "normal" or better picks went 26-19 (58% ML, O/U, RL). This includes "normal" over/under picks going 10-4 (71% O/U). All-time (since the start of the 2012 season), all normal over/under MLB plays are 270-212 (56% O/U).

As the regular season concludes, NBA plays are heating up. All normal plays for the week in the NBA went 7-4 (64% ATS & O/U). Total picks continue to be strong, going 4-2 (67% O/U) with "normal" picks last week. Such "normal" plays are 43-32 (57% O/U) since the beginning of the year in the NBA.

The Predictalator plays every game 50,000 times before it's actually played. This provides us the ability to assign probabilities to the likelihood of just about any outcome occurring in any contest including straight-up, against-the-spread and over/under winners of each game.

Ad
Ad