The team over at Baseball Prospectus led by Jeff Long came to us at the end of last year with collaboration request. The concept was to use Ayasdi’s Machine Intelligence software to look at baseball in the same way that Muthu Alagappan looked at basketball.
The result of this collaboration is really cool and underscores how domain experts can leverage our software – without a stats background. It is a superb use of the software, and based on the comments to the article, is just the start of our partnership together as the roto community is already asking for different looks, looks over time and more granularity.
We won’t recreate the analysis here but will share a couple of screenshots that describe the segmentation.
Jeff and the team at BP broke down the overall population of 311 major leaguers into the following groups:
- Speedy Hit Tools
- Balanced Skill Sets
- Hit Tools
- Power Bats
- On Base Specialists
- Outliers 1
- Outliers 2
- One of a Kinds
Here is an annotated image that shows where each group is represented on the map:
Again, check out the full article here as it describes topology as well as it describes player segments.