2010 Cleveland Indian ZiPS Projections
From BTF's Dan Symborski
over 2 years ago
TonyH
22 comments
0 recs |
Email
PrintComments
Saw that. Stupid. But Zips isn’t supposed to be smart.
by Jay on Jan 2, 2010 5:11 PM EST up reply actions
Well, it’s not quite zero, just rounded. ZiPS actually says 1-in-229.
Remember, though, those are the odds of him stealing 30 bases at the projected playing time. Not counting reached on error in either case, ZiPS has Sizemore standing on first base 168 times in his projection while that number is 229 for 2007 and 202 in 2008.
With 750 PA, the odds of 30 stolen bases increases to 24%
Well, it’s not quite zero, just rounded. ZiPS actually says 1-in-229.
Remember, though, those are the odds of him stealing 30 bases at the projected playing time. Not counting reached on error in either case, ZiPS has Sizemore standing on first base 168 times in his projection while that number is 229 for 2007 and 202 in 2008. With 750 PA, the odds of 30 stolen bases increases to 24%
--
Dan Szymborski
dan@baseballprimer.com
Fair enough. Although I understood it the first time :). And if someone wants to lay 228-to-1 I’m still game.
Seriously, though. I understand that expected playing time for any player is generally lower than one would expect, but 1 in 229 crosses the line from “surprising” to “obviously wrong” in my opinion. Though rate stats are generally the more important aspect of these projections anyway, so I’m willing to move on..
Maybe because it’s using Sizemore’s injured 2009 to partly project his 2010.
"I'm a baseball lifer. It's what I do." —Manny Acta
Sure, but how can any projection, however formulaic, look at 2007 and 2008 and think the chances of a repeat in 2010 are so low, regardless of 2009? I don’t get it. Again, it doesn’t really seem important in a vacuum, but it does give me pause about the underlying methodology. Is there a flaw that is affecting more important parts of these projections?
Only thing I can think of is that it first comes up with an expected playing time and sets that in stone, then calculates event probabilities based on that number of plate appearances. If this is the case, then the rate projections and expected total values may be fine, but the event probabilities would be dead wrong. Event probabilities would need to take into account that plate appearances itself has a range of values around its expected value. Not taking this into account would lead to a gross underestimate of the variability of stolen base totals around the expected value, thereby underestimating the probability of any deviation from the mean. Seems like a ridiculous error to make, though.
I can spit out the data for any number of plate appearance. Perhaps I am unclear. Those are the probabilities of such-and-such for the projected plate appearances, period.
The problem with taking into account a range of plate appearances is that we can’t really define, from a mathematical standpoint, the variability of Sizemore’s possible PA, simply because the playing time of a player has so much that a computer model can’t really deal with. In Sizemore’s case, you’d be asking a computer to guess the effects of “elbow surgery and abdominal surgery 7 months before the start of the season” which, at this point, looks to be the biggest and probably the only significant determinant of what Sizemore’s playing time will be.
--
Dan Szymborski
dan@baseballprimer.com
Why would you not count a projected ROE for the number of SB opportunities?
by Jay on Jan 2, 2010 5:12 PM EST up reply actions
so much that a computer model can’t really deal with
I’m calling BS on this. This would be much more accurate: there’s only so much my computer model can deal with.
Resident LGT results-oriented boob.
by mauichuck on Jan 3, 2010 7:45 PM EST up reply actions 1 recs
Jhonny’s 2nd comp is one Aaron Boone. How frighteningly appropriate.
As General Manager of this team, I demand to know when I'm getting a start.
I always thought Boone was better in Cincy so I initially was all for his comp. Then I looked at Aaron’s B-Ref page. Ouch. I looked at Jhonny’s, who gives his best comp as Juan Uribe for this season, Marcus Giles for his career.
Tribe fan trapped in Kansas
My question is, how do you deal with quantity? A starter who is projected to have an ERA of 5 can obviously reasonably end up between 3.5 and 6.5. Take five of those guys for the year, and you will probably end up with a ERA for those innings of around 5. But what happens if you have eight of those guys? Or 10? or 12? Sure, you’re protecting yourself against having to use the Cassels of the world in injury starts. But isn’t there a benefit beyond injury coverage? Aren’t you also allowing yourself an opportunity to eventually identify and settle on the best 5 or so? Can we make any predictions on this?
I’m aware that the answer might be “No, you (Shapiro, Acta) can’t really tell who the better pitchers are.” There’s just too much noise, not enough innings, and the differences between Carmona, Rondon, Huff, Carrasco are overwhelmed by the day-to-day, or month-to-month, fluctuations.
Asdrubal’s top 5 comps:
Jimmy Rollins
Robin Yount
Edgar Renteria
Marty Marion
Alan Trammell
Three of these guys have won MVP’s, and Trammell finished second once. One (after today, potentially two) HOFers.
That is… exciting.
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
Recent FanPosts
Featured Stories
In Association With
Big League Stew - MLB - Yahoo! Sports
- The Citi Field All-Star game has finally been made official for 2013
- Brett Lawrie receives four-game suspension, fined for ‘aggressive actions’ toward umpire
- PNC Park security guard loses finger in struggle with fan during Pirates game
- Create-a-Caption: Partying like it’s 1988
- The Ron Washington puppet is watching you while you sleep — and has possibly jinxed the Rangers
