There seems to be a debate as to whether the next GM for the Mets should have a background in analytics or scouting.I can see how analytics can be really helpful in allowing a major league team plan how to pitch to players on other teams or where to position players in the field.I do not see how computers or mathematical formulas can predict which 18 year old will be able to hit major league pitching at age 23. I think that drafting really needs an eye for talent more than a computer.
I met with a head of MLB analytics and asked him what graduate level courses one should take if they want to work for an MLB team. He said "machine learning and image processing". That may be indicative of this type of approach.
I also have personal experience in seeing machine learning algorithms outperform humans in areas where the capability of human hearing or eyesight was assumed to be critical. I have learned not to underestimate the capability of machine learning.
It seems unlikely that machines could determine the ability of an MLB prospect to work at their craft but even brain imaging and psychological testing could come into play.
Is there a GM in baseball who doesn't scout players? Does someone draft solely by placing stats and attributes into a computer and selecting a player? I doubt it.
As I have read and heard, analytics help you understand what you are seeing. In some cases they validate the eye test, in other cases they warn you about it.
But I doubt anyone is making decisions in vacuum based solely on analytics.
The next Mets GM should understand and value analytics (and ideally have an analytics dept commensurate with the competition) and be an accomplished talent evaluator (or at least employ them) to spot the players they need to do some deeper analytics on.
but as Shecky said, he or she should (beyond stats or scouting) be a shrewd business person with a plan.
But I doubt anyone is making decisions in vacuum based solely on analytics.
I would like to introduce you to Matt Kelntak and Gabe Kapler.
I met with a head of MLB analytics and asked him what graduate level courses one should take if they want to work for an MLB team. He said "machine learning and image processing". That may be indicative of this type of approach.
I also have personal experience in seeing machine learning algorithms outperform humans in areas where the capability of human hearing or eyesight was assumed to be critical. I have learned not to underestimate the capability of machine learning.
It seems unlikely that machines could determine the ability of an MLB prospect to work at their craft but even brain imaging and psychological testing could come into play.
many of the terms you mention (machine learning, AI, data science) are all intermingled and used inappropriately. Surprising, you mention image processing and that is one of the hard problems and have a significant error rate to the point it slightly non useful. Most of these teams have small data (small number of players, and small models that can fit in a high powered laptop.
A few basic statistical 3c algorithms as exit routines from a target feature descriptive SQL query is about all you need for model prediction around a chosen player and some needed performance or other value. Not rocket science unless you are using spreadsheets and paper. The key point that is raised in this thread is combine experience with data facts instead of just gut instinct and human bets.