Thursday, March 14, 2013

Minor League Stats to MLB Career Projections

The Idea
I have been working on ranking minor leaguers with stats for 4 years now. It doesn't work. There are too many variables and even at the MLB level, developing one number to rank them all (WAR) doesn't work, so there is no way it will in the minors. What I have decided to do is look at projections instead. I don't use minor league equivalencies. I hate MLE's. I dislike the concept and the application even more. What I did was more comparison based.
I took the formulas I used in the past, applied percentages, found groups and found more exact comparisons, assigned MLB data from comps, regressed and viola. Projections for how similar minor leaguers performed in the majors and the chance that they even make it.
The percentages and groupings were released earlier this year for sale. Now I've developed this system. Right now, I only have a good system done for hitters. It's not great, it's good but could be improved upon. I've already altered it some and I regress TOO much right now. Essentially, these are "safe" projections. Players who are likely to succeed will likely produce better slash lines then what I project. From what I see, the most volatile stat is hits. The better players can outproduce their average projection by .020 to .030 points. In the full model I have created, I have a floor and average and a ceiling projection. This is only the average projections. I don't know if I will make the others available yet, because I am not confident in the results.

The Data

I chose a random year from the data I got from The Baseball Cube: 1996. No real reason for that choice other than it was long enough ago that most players would have much or all of their career played. I chose to use Low A ball for my group. Early enough in their career that it should be hard to predict but considering I was going to use 1996-98, they should have enough data to solidly predict their MLB skills.
After running these numbers, I liked the results but I wanted to see how much would change using A ball and lower levels. I used 1994-1996 in a separate calculation and the results were expected. Most players potential was decreased and a couple top players were overlooked. It's A ball. Hard to pick out all the best players. Mike Lowell was the biggest miss.

I am currently working on this projection for all the best current prospects. I currently have a list of 370 players I plan to do this for using their 2010-2012 data or as much is available.

Here is the Google Doc Link if you want to look at it or download it.

No comments:

Post a Comment