The BAKER Prediction Engine is an advanced sports prediction platform. It combines the ability to simulate every game down to the play level 10,000 times with a robust JSON query API. The result is that it can answer practically any question about a future sporting event or outcome. This guild will break down everything you need to know to get started with BAKER.
- Artificial Intelligence
Simulation Based model overview:
To deliver results quickly and accurately, BAKER uses the latest technologies from these areas:
- Artificial Intelligence
- Data Science
- Data Engineering
- High Performance Computing
- Database Design
Many other prediction systems provide results by using one or two of these areas while falling short in others.
BAKER is different because it pushes the envelope in each of these technologies to provide fast, accurate and innovative data that you can build into your applications.
The result is that it can crunch the numbers to simulate up to 26,000 plays per second to give you the most up-to-date and reliable projections possible.
Below are the definitions of three general types of predictive engines and how their capabilities compare to each other.
Types of Prediction Engines
- Play-by-Play Simulations: Every single play is predicted at the player level. This is how the BAKER NFL, NBA and MLB models are built. See Example Play-By-Play Simulation.
- Game Outcome Simulations: A model has generated an expected average outcome for a game or player statistic. The simulations then use some form of random number generator to create a distribution of outcomes from the one predicted average value.
- Machine Learning (No Simulations): A data-driven model using some form of regression or machine learning that predicts the average outcome for a game, team, or player.
- ‘Fantasy Projections & Points’ allows the user to generate a list of fantasy projections for upcoming games. All key fantasy stats are represented on a game-by-game basis, and projections can be viewed for individual players or entire teams.
- ‘Best Bets’ generate a list of best bets for upcoming games. These can be either player props or game line bets, and you can set a threshold of win probability, or an individual sportsbook, to filter down results.
- ‘Prediction’ returns traditional outputs such as projected stats, score predictions and true odds, but also answers bespoke queries that accept parameter combinations containing any category (game, team or player) for any statistic (points, spread, yards, assists, etc.) over any time period (full game, first half, first inning, etc.) that are then available for further filtering by thousands of different scenarios.
- Advanced query API enables customers to have full access to build custom question sets based on their desired “ask anything” criteria. Stack and layer multiple AND operations to populate betting and fantasy outcome probability.
The simulation engine is the core of BAKER's predictions.
NBA and MLB Simulations are run the morning of that day’s games and run periodically throughout the day until each game starts.
NFL Simulations are started the Monday morning of the upcoming game week. NFL simulations are refreshed multiple times per day until games occur each week.
All simulations are pregame simulations and stored in a database for instant querying via the API.
BAKER's prediction database is designed to provide instant querying of any type of future-looking sports question.
This can empower you to create innovative data-driven applications, widgets and content for your end users that would otherwise be impossible to build with any other prediction system.
Below are examples of the types of questions BAKER can answer about future sporting events.
Sports Betting Questions
BAKER can instantly calculate the probability of any of these scenarios occurring:
- Dalvin Cook to run for 10 TDs & the Vikings miss the playoffs.
- C. Kupp, T. Higgins & K. Allen are all first to score a TD in their games this week.
- Steph Curry hit 5+ three pointers & the Warriors to lose.
- Knicks vs. Celtics to score more than 245.5 total points.
It can also predict the most likely median outcome, which is useful for determining what the projected betting line should be:
- What is the median outcome for David Njoku's receiving yards this week?
- What is the median outcome for Gerrit Cole strike outs this year?
Fantasy and DFS Questions
BAKER can instantly answer these questions to help fantasy and DFS players:
- What are the projected season-long fantasy projections for every player?
- Should I start Amari Cooper or Terri McLaurin this week in non-PPR scoring?
- Which running back has the highest projected fantasy points for the rest of the season?
- Does Darren Waller or T.J. Hockenson have a higher 95th percentile projection this week?
In addition to the new and unique questions listed above, BAKER also has pre-built API endpoints that answer typical questions for sports fans:
- What are the most profitable bets to make this week?
- What are the fantasy projections for each player?
- How has this player done in a similar game situation previously?
In addition to simulating future events, BAKER has custom curated trends that analyze historical data in dozens of dimensions to find unique facts about upcoming games.
Trends are tied to sports betting lines and can even provide interesting trends on both sides of a betting line to allow sports bettors to do their own research.
Here are example trends that BAKER can generate:
- The Phillies have covered the spread 6 out of 7 times when playing on the road against the NL West this season.
- The Dodgers have won 15 out of 20 times when playing at home this season.
- The White Sox have not covered the spread 7 out of 7 times when playing at home against the AL East this season.
- Bryce Harper has gone under the betting line for hits 3 out of 3 times when playing on the road against the NL West this season.
- Aaron Judge has gone over the betting line for hits 15 out of 24 times this season.
- José Abreu has gone under the betting line for total bases 10 out of 14 times when playing at home this season.
While the full details of the simulation engine are proprietary, some of the input factors for the simulation engine are listed below. Each of the inputs are sourced from SportsDataIO’s League, Play by Play, Player, and Live Odds feeds.
Input Types - Game Situation, Game score differential, Time on clock, Last play, Player Historical Performance, Player Splits, Team Historical Tendencies (Defensive and offensive tendencies).