The baseball analytics world is filled with analytics systems*, all built to try and predict how good each player would be by season. The value from getting this right is huge and stretches from fans to set the right expectations, to fantasy GMs to draft the right teams, the real-life GMs who have to negotiate contracts, manage rosters, and make major decisions that affect players lives and livelihoods.
My colleagues at DataRobot and I decided we could do better than what was already out there, so we spent the better part of March building the first AI-driven player projection system, the DataRobot Intelligent Value Estimator for MLB 2020 (D.R.I.V.E. MLB 2020).
Unlike the other systems, which rely on regressions, regression curves, similarity matching, and other basic techniques to determine future performance; D.R.I.V.E. is the only system that harnesses the enterprise-grade machine learning tools to build better projections. Read the full blog post here to learn more about our methodology, and explore the results dashboard below.
Obviously, these WAR totals won’t be matched since they presume a 162 game season. However, the relative ranking and performance of all players still holds up.
Given these results, we can predict the following about the 2020 season:
2020 Playoffs Matchups:
| American League | National League |
| Division Series | Division Series |
| (1) Astros vs. Wild Card (2) Yankees vs. (3) Twins | (1) Dodgers vs. Wild Card (2) Braves vs. (3) Reds |
| Wild Card Game | Wild Card Game |
| (5) Angels at (4) Indians | Nationals / Mets / Padres (3-way tie) |
2020 Major Individual Awards:
| American League | National League |
| MVP | MVP |
| Mike Trout Alex Bregman Francisco Lindor | Mookie Betts Christian Yelich Cody Bellinger |
| Cy Young | Cy Young |
| Gerrit Cole Justin Verlander Lance Lynn | Jacob Degrom Max Scherzer Stephen Strasburg |
* PECOTA, Steamer, PoRKChoPS**, ZiPS, Marcel to name the best known
** I made that one up
Start the discussion