Data is playing an increasingly important role in the overall betting experience, and with that comes a growing demand for more in-depth insights. This is where SportsDataIO’s BAKER Predictive Engine comes in.
Speaking ahead of this week’s SBC Summit North America, Dustin Sullivan, President of SportsDataIO, explains how the BAKER engine simulates every game for a league’s season down to the play-by-play level before discussing how the predictive engine can have wider uses for media, sportsbooks and other industries.
SBC: For those that might not know, what is SportsDataIO’s BAKER engine?
DS: The BAKER Predictive Engine allows customers to ask pretty much any question about a future sporting event, from a play, quarter, half and game, to the probability of a winning streak, playoff picture, championship and even career. Questions are asked using BAKER’s robust query interface with results delivered via an API and within our widgets.
The engine simulates every game for a league’s season down to the play-by-play level, resulting in predictions that range from micro to macro. And the outputs are seamlessly tied to our other scores, stats and odds IDs.
SBC: How does this predictive engine provide deeper, more micro insights into gameplay? And how can these insights be used for futures markets?
DS: Because BAKER is simulating down to the play level, it’s able to look more granular than other traditional models currently available to license. It utilizes dozens of machine learning models to drive the behavior of each play. This allows us to fine-tune BAKER to account for many different factors and provide insights that would be impossible to come up with otherwise. For example, one of BAKER’s models focuses exclusively on if a Quarter Back’s pass will be complete or not.
Improvements to that model can impact predictions for the first to score prop, whereas traditional models wouldn’t be able to pick up that interaction because it’s too noisy at a game-level. The same is true for futures markets. Since BAKER can simulate one game’s play by play, we have it trained to simulate the entire season’s slate of games to the play level. The result is that these targeted, well-trained models are combined in a robust way to look at scenarios throughout the regular and post-season well before the first game is ever played.
SBC: What benefits can the Baker engine bring to media companies?
DS: BAKER can help media companies with not only predictions but also trends utilizing SportsDataIO’s historical database. Those outputs can supplement written and video content, power premium tools, and help with conversions through programmatic and other campaigns. We’ve found that it’s useful for media brands because their audiences want to hear a story around a certain game or odds price in addition to seeing the smart computer predictions.
Also, depending on their audience, media brands can use either pre-calculated best bets, or present more advanced data such as histograms of player predictions so those audiences can make their own decisions for what action to take.
We’ve found that consumers are very particular about how they bet, either following expert advice or only trusting their own research, and BAKER can provide both persuasive and factual information, depending on the voice needed for the story being written.
SBC: And how can sportsbooks and trading teams use these insights to create a more well-rounded offering?
DS: For every game simulation that BAKER runs, it summarizes every possible statistic at the game, team, player and play level. It also has a full distribution of results that account for correlations. There are median (50th percentile) projections available for every statistic, which are usually more valuable to sportsbooks than traditional average (mean) projections. The outputs also include exact game score predictions, and correlated probabilities needed for same game parlays.
Most other models ignore or sort of guess at correlations between players within a game. BAKER is 100% internally-consistent and fully accounts for correlations. For example, a team’s passing yards always add up to the individual players’ receiving yards, and the probability of a bad game from a quarterback will be reflected equally in his receivers’ projections.
Lastly, BAKER’s simulations are lightning-fast, running almost 30,000 plays per second, ensuring the outputs are always current as new information becomes available.
SBC: Outside of the use for media and operators, where else do you see this product fitting?
DS: We see sportsbooks and media brands as just the tip of the iceberg. For example, BAKER can help ride-sharing companies with resource planning around probabilities of teams with large traveling fan bases playing in various bowl game locations, help brick-and-mortar retailers with merchandise planning to support potential win streaks and playoff runs, and help ticket resellers with demand planning, just to name a few.
SBC: What can we expect from SportsDataIO for the remainder of 2022?
DS: For the majority of our 15 years in business, we were strictly focused on leveraging our API as a data delivery company. But in the last couple of years, we have expanded in two areas: our BAKER predictive engine, which is finally ready for its debut, and our display widgets, which helps both operators and media brands convert new users and better engage with their existing audiences.
Now with our three divisions, Data, Display & Predictive, we’re able to mix-and-match components of each to deliver bespoke solutions across a broad spectrum of verticals and use-cases.
For the remainder of 2022 and throughout next year we will be intently focused on expanding the capabilities of those three divisions and working directly with our customers to identify new and unique ways to weave the divisions together in order to create more value.