02-23-2008, 04:33 PM
|
#3 (permalink)
|
|
Golden Poster
Join Date: Feb 2008
Posts: 647
|
Here's the description right from the horse's mouth:
Q. What are ZiPS?
A. ZiPS is a projection system written by Dan Szymborski to project performance in individual baseball players.
Q. What does ZiPS stand for?
A. sZymborski Projection System
Q. That's lame.
A. That's not a question.
Q. OK, do you realize how lame that is?
A. Yes. It originates from the Defense-Independent Pitching Statistics that Voros McCracken developed in 1999. Originally, I was just going to do projections for pitchers, but the realization came that doing one without the other was really dumb. Since it was the first attempt at a non-Voros projection system utilizing some of the discoveries in DIPS theory, I decided to name it in honor of DIPS.
Q. Why is the 'I' in ZiPS lowercase?
A. Because ZIPS looks like some bad shoe at Wal-Mart for little kids while ZiPS totally reminds me of CHiPs. Which is way cooler for a reason I can't put my finger on. I actually meant to call it ZiPs but I typoed it the first time I said the term and didn’t feel like changing it.
Q. So, how is DIPS applied to your pitching projections?
A. I try to predict future BABIP based on a number of factors - the knowledge, demonstrated by Voros, that there's a very strong regression to a mean for this number plus information about the pitcher's tendencies to pitch above his team, his natural BABIP tendencies based on information like handedness, whether or not the pitcher is a knuckleballer, and G/F ratio.
Q. How's the rest come about?
A. I use 4 years of weighted statistics (8,5,4,3) for players between the ages of 24 and 38 and 3 years for younger and older players. This goes for hitter with 3 years mainly being used for pitchers.
Q. How does the aging coming about?
A. My ZiPS program generates growth and decline curves based on player type. I don't try to find particularly similar players but instead large groups with similar characteristics, such as K rate for pitchers, Speed Score for batters, BABIP for batters, handedness, and a lot of other stuff. I do list most similar players at times, but that’s mainly for interest.
Q. Oh my God, my favorite team is projected to get 15000 games! Are you the biggest idiot ever?
A. I'm projecting equivalent production. A player with a projection of 230/270/400 is projected to hit that in the majors if he got that estimated playing time in the majors. A correct projection could be 230/270/400 in the majors or 260/320/450 in AAA or 290/360/500 in AA, depending on the league levels of scoring and the park factors.
Q. How do you project league offensive levels?
A. Weighted recent years. Anything else is just too damn difficult and it usually does the job. It's either that or give all statistics in relative percentages from league average, which is kinda boring. I like projecting a guy to hit 270/330/450, not +1%/-3%/+3%.
Q. How much do these projections cost?
A. They're free for now and for the foreseeable future.
Q. Can I sell your projections?
A. No.
Q. How accurate are the projections?
A. About as accurate (or inaccurate, as is probably the case with projections) as any other commercially available projection system. The 2006 results can be found at
BBTF's Newsblog Discussion :: 2006 Projection Results
Overall, I'd give PECOTA the gold for 2006, but I think ZiPS would deserve the silver, at least of the systems that Sean Smith had access to for 2006 (Ken Warren or Rotowire or someone could've waxed all of us - I don't know).
Q. How is playing time calculated?
A. Based solely on recent playing time. ZiPS, as a consequence, only projects injuries to the extent that the injuries were properly reflected by the previous playing time.
Q. Are platoon splits calculated?
A. ZiPS only projects platoon splits to the extent that the future platoon splits were properly reflected by the previous playing time.
Q. Is steroids included?
A. ZiPS only projects steroid play to the extent that past play reflects the same steroid use as future play.
Q. What about leadership? Paul Lo Duca's way too clutch to ever play badly!
A. ZiPS are computer projections and designed to do what computers do best. If one feels there's a truly compelling reason to ignore a projection, such as playing through an injury, or magic beans, or whatever, they should go with their gut and only use a computer projection as an initial objective evaluation, not a final prediction.
Q. What the hell? Where are save totals?
A. I don't project saves because it's way too much of an opportunity-based stat. RBIs might fluctuate based on hitting #3 or #6, for example, but nowhere near the degree that save totals would fluctuate between being the closer or the manager and setup.
Q. How do you decide what players to project?
A. First, I do the 40-man rosters. Then the non-roster invitees if it’s late in the winter. Then unsigned free agents. Then the guys that Sickels looks at if I feel they’re anywhere near being relevant. Then random players that don’t fit any of these categories, like Andy Abad. |
|
|