This Saturday at Santa Anita could prove something of a watershed moment in North American horse racing-the carding of two handicaps built upon a new national ratings system.<\/p>\n
One race, a $35,000 handicap over 6 furlongs for dirt horses rated between 70 and 79, has drawn seven entries. The next race on the card, a $65,000 handicap over 6 1\/2 furlongs on the hillside turf course for fillies and mares rated between 80 and 95, has drawn six entries.<\/p>\n
The impetus for launching a ratings system in the U.S. has been well documented. Aside from helping racing secretaries card a suitable set of races for their backstretch inventory, such a system gives the connections of a claiming class of horse options to run without fear of losing it.<\/p>\n
\n
\n
<\/p>\n
One thing it’s not designed to do, those familiar with the system stress, is replace these claiming races, which has long proven the backbone of the U.S. product.<\/p>\n
While North American is one of the last major jurisdictions to use such a ratings system, it appears to be the first to use an algorithm to generate the numbers (the others use human handicappers).<\/p>\n
\u201cIt’s novel to my understanding,\u201d said Chance Moquett, senior manager of strategic solutions at Equibase, the company responsible for building the algorithm. \u201cI believe it’s the only algorithmic rating of a population of racehorses in the world right now.\u201d<\/p>\n
For any such system to work, however, the industry needs to have trust in it.<\/p>\n
\u201cThere’s no right or wrong way. But the right way is one that works and people have confidence in,\u201d long-time track executive Rick Hammerle told Steve Byk’s radio show last Thursday. \u201cYou’re only as good as the number.\u201d<\/p>\n
Hammerle was involved in the building of a previous iteration of the ratings system which fell by the wayside earlier this year, but not with the version that has just launched.<\/p>\n
According to Moquett, Saturday’s ratings-based races are part of a soft launch of a system meant to be refined and improved over time and with experience.<\/p>\n
\u201cIt’s not necessarily a polished thing at this point in time. We wanted to make sure that we are continuing to work on it, and that people understand that,\u201d said Moquett.<\/p>\n
\u201cFeedback is important,\u201d Moquett added. \u201cFeedback in any form is good feedback.\u201d<\/p>\n
\u00a0<\/p>\n
The Building of the System<\/strong><\/p>\n Since February, around 40 to 50 industry stakeholders within four different \u201ccohort\u201d groups have been working on the system, said Moquett, estimating around \u201c50 formal meetings\u201d as it came together.<\/p>\n These four groups comprise the worlds of data science, competition (including trainers and track executives), handicapping and wagering, and the international community.<\/p>\n In April, the racing secretaries from 12 different tracks or jurisdictions sent to Equibase a batch of around 20 horses for the team to rate.<\/p>\n (The 12 participating racing offices were at Keeneland, Churchill Downs, Gulfstream Park, Maryland, Oaklawn Park, Sam Houston, Prairie Meadows, NYRA, Canterbury, Assiniboia Downs, Del Mar and Santa Anita)<\/p>\n \u201cThen we would give back the ratings and ask, ‘how does this look? Does this look right?’ We worked with each one of those racing offices individually,\u201d said Moquett.<\/p>\n That process took a couple months. And then, when Equibase \u201cfelt like all the regional racing secretaries were largely pleased with how their selected bands seemed to seem to be rated,\u201d they went back to test five years-worth of data on an individual yearly basis, said Moquett.<\/p>\n \u201cThat means we took every horse and ran it through a ratings history process for each race,\u201d said Moquett. \u201cWe basically re-ran every race for every horse for the calendar year 2020 forward, caught out any outliers.\u201d<\/p>\n What were some of those outliers?<\/p>\n \u201cWe did correlation analysis between purse earnings versus what their figure was at any point in time, to make sure the horses that generally were earning more money were in line [ratings wise]. And we identified that our turf horses seemed to be rated a little bit higher at first,\u201d said Moquett.<\/p>\n More broadly, \u201cI think another thing that we found, our rating was a little more volatile to surface switch and performance, and it had a hard time differentiating good horses from the elite level,\u201d said Moquett.<\/p>\n The modified result is a multi-factor algorithm that gives horses a numerical rating (from 0-150) based on performance metrics, race class adjustments, surface and distance factors, track-specific variability, and recent form.<\/p>\n For a horse to be rated, they must have at least two North American starts. It’s updated daily to reflect ongoing race results, so, even if a horse remains in its stall, its rating might fluctuate as its former rivals go out to compete.<\/p>\n The formula is based on three main buckets, with the \u201cperformance score\u201d providing the core of the ratings.<\/p>\n \u201cThis has a litany of other factors peppered into it, but in essence, it’s how fast did a horse run and how many lengths did it win or get beaten by,\u201d said Moquett.<\/p>\n The second is the \u201crace strengths adjustments\u201d component, which looks at things like the class of the race and field strength. The \u201ctrack and distance adjustments\u201d factor in information like track variance and bias, as well as post position.<\/p>\n (A breakdown of the formula can be found here in an FAQ of the new system)<\/p>\n The computer algorithm won’t be open-sourced-Moquett pointed to the proprietary nature of the system. That said, \u201cwe welcome people to come under the hood, see what’s going on,\u201d he added.<\/p>\n