ZiPS stands for sZymborskI Projection System, the acronym chosen because the pitching portion heavily utilizes the DIPS system of Voros McCracken and the i being small in honor of everyone’s favorite television serial on the subject of highway patrols.
The batting portion is a heavily edited Brock5 system, with completely revamped aging factors based on historical 3-year aging trends, park factor support, and trends based on changes in component stats such as strikeouts.
A more expansive article on the subject will be available in the coming months.
Some things I would like to include in future versions:
- Support for defensive projections. For a true DIPS projection system for pitchers rather than a DIPS-neutral projection system, you need to be able to, at some level, predict defensive performance to get you closer to an expected $H for the team. You’ll never predict luck, but a pitching projection improves when defense can properly implemented.
- Improved playing time engine. The problem with this is that to predict playing time, you need to inject a healthy dose of subjectivity into the projection. This is at odds with the basis of a projection for baseball players, the main strength of which isn’t the genius of the results, but the objectivity and ability to put the past into proper context.
Also, you get into the territory at which predicting playing time can detract from the usefulness of the results. Say you have a minor-league player like Jose Leon of the Orioles. What is more useful? A projection that he’ll go 4-13 with 2 RBI for the season, or a full season line tempered with knowledge of his chances of making the team? I’ve opted for the latter choice, but nothing’s going to please all parties.
- Further research on the effects of position changes on offense. This might seem like a trivial matter but when the occasion pops up, it’s one of the times that a projection is more useful. We can guess what Pudge Rodriguez will do most years, play well and be injured, but what if he moves to 3rd or another position? It’d be nice to be able to make an educated guess at the results rather than “Catchers frequently hit better at easier positions.”
- More accurate projections for raw totals of W-L and RS-RA. It’d be great to be able to project a team’s W-L record through RS-RA projections but until you can predict the distribution of playing time, it’s hard to predict either. Won-loss projections are difficult, for instance, because instead of predicting the W-L record off of a league average offense in the team’s home park, you’d need to be able to predict the team’s offense. I can do that by making a subjective analysis of a team’s likely run total but that strips away a good bit of the objectivity of a projection.
This is certainly not an exhaustive list but just a few things I think to be important.
Posted: March 04, 2003 at 06:00 AM | 0 comment(s)
Login to Bookmark