My big project for this off season is designing a model for projecting wOBA for hitters and opponent wOBA for pitchers. The model is essentially set, but the actual projections are still in progress. In the coming weeks I will post more about the projections as the project develops in the coming weeks, and hopefully I'll release my final wOBA projections by late March. But right now I'd like to focus on what I have come up with so far for the model.
When constructing this model, I considered the following factors both necessary and practical for use in a projection system:
1. Recent past performance (previous season's wOBA)
2. Average past performance (career wOBA)
3. "Luck" in recent past (previous season's BABIP compared with xBABIP)
4. Age
Of course, there are many other factors that one could take into account, but in the interest of practicality, I decided to try and keep this a simple as possible without comprimising reasonable accuracy. Ideally, I would regress wOBA over each individual player's entire career, personalizing each player's projection much more. However, using 2009 wOBA as a starting point, regressing towards career wOBA, and then adjusting for luck and age gets me well within the "reasonably accurate" range.
I'm very sorry, but I'm not going to release my exact formula...at least not yet. I will provide a bit of a teaser here, though. Below are three players' 2009 wOBA, Will Buckley projection, and Bill James projection.
I'll be posting again soon, hopefully, with some results. Thanks for reading!
