A new micro-betting company finds itself in the sights of the baseball analytics community over its live probabilities.
nVenue is the company behind the live at-bat odds shown on Friday Night Baseball on Apple TV.
The company said its machine-learning model uses around 110 inputs to accurately determine the outcome of an at-bat after each pitch.
nVenue is also looking to sell those probabilities as a micro-betting product to operators and providers. It aims to be integrated into US sportsbook partners by next baseball season, CEO Kelly Pracht told LSR.
What are the odds?
However, the odds themselves have come under scrutiny since being shown on Apple TV.
FanGraphs writer Ben Clemens noticed some counterintuitive changes in the odds and built his own simple model to test them. As one example, nVenue showed Texas Rangers second baseman Marcus Semien increasing his chances of a hit despite falling behind 0-2. Semien has hit better this year when behind in the count but that appeared to be merely noise.
To test, Clemens built a model based solely on the count, i.e. a player in an 0-2 count has a 47% chance of a strikeout. Clemens then used that model to bet against the nVenue model and promptly demolished it.
Overfit model?
As Clemens put it:
“In general, it seems to me the on-screen odds suffer from over-fitting. Even if you’re not a statistical sort, you can implicitly understand this from watching a few baseball broadcasts over the years. If the screen displays that a hitter is 5-for-7 on Fridays against opposite-handed pitching in the sixth inning or later, you rightly say ‘yeah, that doesn’t sound like it matters.’ A poorly calibrated model might not, though.
“These odds are worse than not seeing odds on-screen, at least as far as I’m concerned. I wish more people thought about baseball probabilistically, but having clearly inaccurate odds – Marcus Semien isn’t more likely to reach base on 0-2 than he is on 0-0, no matter what the screen says – could result in people trusting odds less, not more.”
Ready for bettors?
Of course, inaccurate odds are a bigger problem if nVenue also plans to take bets against them.
When asked about the FanGraphs article, Pracht disputed the findings.
“We respect their view but they’re using 2,000 data points compared to our millions of insights,” Pracht said. “They did mention our work got better over time, which is indicative of machine learning. The model gets better over time. That’s a natural data expectation.”
Micro-betting plans for nVenue
Pracht said the company was in talks with a “number of betting operators, platform providers and data providers.”
“We are one of a handful of companies that can do micro-betting at the moment and at scale,” Pracht said.
Pracht described nVenue as the Pepsi to SimpleBet’s Coca-Cola.
“Hats off to SimpleBet for being first mover, it’s a difficult thing to do,” Pracht said. “They proved out the concept of micro-betting and that it can be done. We are thankful.”
Huge addressable market for micro-betting
SimpleBet accounted for a 30% share of live college football betting handle at DraftKings last bowl season, per Morgan Stanley. It also accounted for 25% of live Super Bowl handle at DraftKings.
Pracht said the micro-betting market could generate $160 billion in bets per year by 2025, thus supporting more than one provider.
One point of differentiation is the ability to surface relevant bets rather than simply the same list for each at-bat, Pracht said.
Next steps
nVenue is aiming to have its first operator integrations in time for the 2023 baseball season, Pracht said.
The company is also expanding its product to the core four sports.
