Pit Rho is trying to give NASCAR its “Moneyball” moment. Its plan seems to be working.
Illustration: Tim Marrs
Race strategy’s always been a gambler’s game, if you think about it. There’s calculation and undeniable risk, bluffing and calling, fear and stakes. It’s not about just your move, but about predicting others’. So it’s only natural that the game-changing software that reinvented NASCAR pit strategy started out as a way for Josh Browne to place smarter bets.
“I met a group of guys that all had PhDs from MIT. And they wanted to do predictive analytics for motorsports,” Browne said. As a former chief engineer at Red Bull’s short-lived NASCAR team and an academic himself, he was the perfect man for the project.
Welcome to The Professionals, a Road & Track column where we talk to the people behind some of the automotive industry’s most fascinating jobs.
“We spent about a year or so building a predictive model to do gambling because we recognized that you could pretty much beat the system in Vegas on some of the sports books betting on NASCAR because their lines were sub-optimal in some cases,” Browne told Road & Track. “But what that very quickly turned into was a General Motors funded project.”
Josh Browne, co-founder of Pit Rho/Rho AI.
Handout: Josh Browne
The auto giant wanted to raise the stakes. If Browne and his team could predict race winners, surely they could help create them, too. So after dubbing the new project “Pit Rho”—so named for the Greek letter that represents the strength of a relationship between two variables in statistics—Browne began work with Richard Childress Racing. The idea was to see if, using the predictive modeling, you could make smart pit strategy calls that no human would see on their own. NASCAR, it seemed, was ripe for its Moneyball moment, where data analysis fundamentally changes the sport for good.
Years into the project, it’s hard to say that Pit Rho—no longer the company name but the product of parent company Rho AI—has had quite the earth-shattering impact on the sport as Billy Beane’s work for the Oakland Athletics had on baseball. There’s been little coverage of it and few have acknowledged its role in any major successes. Yet look closely and you can see Browne’s theory being validated on the track. Take, for instance, Ryan Newman’s race win in Phoenix back in 2017. After 127 races without a win himself—and 112 races without Richard Childress Racing scoring any wins—Newman makes an alternate strategy call to the field, stays out, and comes from seventh to win the race.
Ryan Newman leading the pack
Sean Gardner/Getty
That winning strategy was suggested by Pit Rho, Browne says. While media for the time praised the bold call of Newman’s crew chief, RCR Racing’s own site notes that the decision to stay out was based on the data and analytics in front of them. Though it’s worded not to reveal that they’re using a fundamentally different form of data than other teams, the statement confirms what Browne told R&T. That race validated the model. Since then, Rho AI has significantly expanded its relationship with GM, ending up in far more NASCAR boxes. Its deal with GM is exclusive, but that doesn’t mean Ford and Toyota are sitting still on strategy.
“We know that everyone’s friends in the industry, so we know the Ford and Toyota people and we know that they’re working on similar things. Especially because there are people that leave GM teams and go to other teams and explain what Pit Rho does,” Browne says. “But to date, perhaps when this article goes up they will have caught up, but to date they haven’t caught up.”
Ryan Newman celebrating his victory at the Camping World 500 in 2017.
Chris Trotman/Getty Images
Yet the Pit Rho team isn’t going to squander its early lead. Though Browne, 50, has moved to a part-time role as he juggles the company, his role as an adjunct lecturer at Columbia, and his other company that is working on reducing emissions in fuel and chemical production supply chains, the team has grown. Andrew Maness, 34, came on board in 2014 and is now the company’s Technical Director.
Maness, who studied Mathematics and Statistics at Kansas State University before a stint at the Federal Reserve, was a natural fit. He had also been captivated by the intersection of racing and data, founding the fan blog NASCARnomics to share interesting data he found. Eventually, it grew big enough to draw ire from the NASCAR suits who weren’t thrilled with Maness sharing data about the series’ dwindling TV ratings. Still, the work got him noticed by the Pit Rho team, who brought him on as an analyst to make sure the model’s projections were meshing with the real world.
Andrew Maness
“As a data analyst I paid attention to details, making sure we’re projecting pitstop times correctly, projecting driver times on to and off of pit road, making sure that all of the little nuts and bolts that went into our predictive and recommending models were accurate,” Maness says. “I worked on just a lot of those tiny things that can add up to a lot. Maybe teams don’t give a lot of credit for that, but if it’s wrong it’s incredibly wrong and you take a lot of blame for it.”
That kind of painstaking focus on the small stuff is a large part of what makes Pit Rho so good. Because, while Browne and Maness were reluctant to share details, they noted that it’s not just about the ultra-precise data you feed into the system, but also the complicated ways they interact. Ways that, they say, few humans would think about.
“If we have an advantage, it’s the interconnectedness between the intuitive and non-intuitive factors that actually matter every lap,” Brown says. “Some of it is pretty complex.”
Handout: Pit Rho
When a car is running fifth, for instance, it makes a big difference whether it’s in fifth at Fontana or at Martinsville. The gap between the cars is obviously important, as is the fact that the four cars in front of you matter. What’s less intuitive, Browne says, is how much the cars behind matter. Because it’s not just about what line they take and their general aggression, but also how likely they are to cause a caution when they’re in a bad position or when they’re running this strategy. Sometimes the obvious strategy will still be the right one, but looking at all of these factors gives you a lot more room to make bold calls.
“If you’ve identified something that you think is going to load the dice in your favor, you can act on that,” Browne says.
Newman’s celebratory donuts.
Brian Lawdermilk/Getty Images
That can’t happen every race. This is, perhaps, why this data-driven approach hasn’t completely disrupted the sport or given Chevy teams an insurmountable lead over Ford and Toyota teams using more traditional strategies. But when the answer is buried beneath layers of complex, intertwined factors, Pit Rho seems to be the best tool for identifying it. And now that the company is sending analysts down earlier and earlier to the track, collecting more and more data, it hopes to keep that lead alive for even longer. After all, it has at least a five-year head start on any competing efforts.
Asked about long-term plans, Browne won’t say what’s next. But, he says, he’d expect it to work in Indycar. Going after a new sport, partway through a revolution you kickstarted in another, is quite the risk. But Josh Browne knows a thing or two about placing bets.
Keyword: The AI Startup That’s Disrupting NASCAR Race Strategy