THE BOOK--Playing The Percentages In Baseball

Ok, so current/former Reds on those lists:

Reese +10
Conine -12
Encarnacion -14
Dunn -18
Cantu -20
Pena -24
Griffey -31

Yikes.  And Reese wasn’t even on the team during the period of time sampled.
-j

Its amazing to me how well TotalZone, the retrosheet system, compares to UZR.  I just ran numbers from 2003-2007 (whole season) for a BTF thread because I was curious as to how players did on grass and turf.

My SS totals for those years, no regression, per full season:

Everett +25 (+24)
Counsell +17 (+17)
Bartlett +14 (+13)
Vizquel +13 (not on Uzr list)
Perez +11 (+12)
Uribe +3 (+10)
Barmes +11 (+10)
Eckstein +4 (+9)

Adams -16 (-10)
Young -13 (-11)

What that list tells me is that over a long-enough time, all the extra parameters that UZR is capturing all comes out in the wash.  There’s no extra skew/bias in the data.

Vizquel was +8.

adam everett is simply stupendous and i am sorry that the poor guy got stiffed out of Gold Gloves simply because he can’t hit for spit.

it is also sad to see more than a few astros names on the negative list. and i must say that i am stunned that carlos lee didn’t make the negative list.

Carlos Lee was -12, which just missed my cutoff.

The good news for Everett is that almost all the best fielding SS in MLB are in the NL.  So, he should have a better chance in the AL, presuming the Twins can stand to play him and Punto at the same time.

You’ll be hearing alot of b.s. about whether the Twins can “afford” to play two such slick-fielding, poor-hitting players.  They can have Everett, Punto, and Feliz, and I wouldn’t be worried.  A run is a run is a run.  Don’t matter if the score is 3-2 or 5-4. 

Some people think you get bonus wins if you win a game by scoring 5 or more runs.  At least, that’s what you’d infer based on their comments on the subject.

well tango

i see you have not been listening to astros fans who firmly believe that losing 9-8 is MUCH better than losing 2-1.

the twins also got mike lamb - i am curious - what is his score? i personally think he’s right down there with wigginton, but i’d love to see the numbers

thanx

UZR has him as average, but only 1 seasons worth of data.

Fans such as yourself concur with the horribleness of Lamb, Wiggington, and Carlos Lee:
http://www.tangotiger.net/scouting/scoutResults2007_HOU.html

Mike Lamb’s top comps:
http://www.tangotiger.net/scouting/sim2007_6390.html

All agree that he is a pure corner player (LF, RF, 1B, 3B). 

If I’ve got UZR, I’m regressing his performance toward the UZR of those 15 or 20 sorry group of guys.  I’m pretty sure mgl would concur.

That’s weird.  Several of those guys have fairly decent UZR numbers.  Not what I’d expect.  Maybe Lamb is a hidden gem?

The Twins signed Feliz?  I missed that, that will make Twins pitchers, especially groundballers, great targets in fantasy.

That would make Lamb the fulltime DH.

No no no!  I didn’t mean to imply that they did.  I was just using an example of had they done so, that Twins fans would be up-in-arms.

Re-reading.  The “no no no” was directed toward me, not anyone else.  Post is harsher than it should be.  Apologies for writing faster than I was thinking.

Over at BTF, Rally said this:

“The problem with zone rating is simple - balls hit 20 feet off the wall are counted towards a player’s opportunities, for a reason I can’t even begin to understand or explain”

The reason is straightforward:
1 - the scorers think of baseball as two-dimensional
2 - the scorers only note limited things about a ball: rather than think like a fan, they think like a techie; we want to know where the ball first hit the ground; how many hops it took to get to the fielder or get past him; where did the ball and player pass each other; where was the ball each time it hit something; how long did it take to get somewhere

I think I’ll add this to my wiseguy list.

would you please be good enough to provide the numbers for

Luke Scott
Kaz Matsui (as a second baseman)
Miguel Tejada

thanx

All those guys are average fielders. 

If you remove Kaz’s SS days, he becomes a +5 runs fielder.

(By the way, an average fielder, if you put him in the corner OF, will look above average compared to the guys already there.)

MGL and I had this exchange on UZR over at BTF (in a thread on fieldability of Wang’s GBs), but we’ll probably have a better discussion here:

GUY:  MGL, have you ever turned this around to see if UZR captures the full impact of hitters? That is, suppose you take a certain group of GBs defined by the UZR parameters—zones x, y, or z; LHH; avg GB/FB pitcher, etc.—and find the average out expectation for these GBs is 75%. If we then divided these GBs into groups defined by the BABIP of the hitters (or even better, the hitters’ BABIP on GBs), would we find that the out% on those hit by the .320+ BABIP hitters is the same as those hit by the under-.280 hitters? Or to the extent that isn’t true, could you then use hitter-BABIP as a parameter to create more accurate out expectations?

MGL:  Isn’t that circular reasoning (selective sampling)? Of course, if we did look at something like that, we would look at high and low BABIP batters from a different sample (like career or last few years prior to the UZR sample). In any case, I have never looked at this.  It is actually an interesting issue, as we would expect that guys with more power to hit harder ground balls. Of course, to some extent, that is captured in the speed designation (soft, medium, or hard) of the ground ball. Then again, a hard hit ground ball by a power hitter is probably hit harder (and thus has a lower out%) than a ground ball by a non-power hitter. That is a good suggestion though - to adjust for the power of the batter, using out-of-sample data for the power designation. That probably applies to fly balls or at least to line drives. Actually, the harder hit a fly ball is, the easier it is to catch (higher out %) I think. Interesting idea.

GUY:  MGL, you may be right that using out-of-sample BABIP for the hitters is better. But I’m not sure it’s necessary. If UZR were providing a “perfect” measure of fielding difficulty, shouldn’t the balls hit by a .320 BABIP hitter have a .68 out probability? But my guess is that a good hitter out-performs his own UZR estimate.

I was thinking that this would be more valuable on GBs, as hitters vary quite a bit on their GB hit rate. Is there a lot of variance on non-HR FBs and LDs as well?

For multi-year UZR scores this probably won’t change much, as hitters faced by fielders should even out. But it might improve accuracy (and narrow variance) in single-season estimates.

Actually, we know based on my article on THT how much the hitting evens out.  Just look at the out rate for the “other” shortstops for each SS. 

The range was 12.0% to 12.7%.  This suggests that there’s there’s some variability in the type of hitters being faced (over a long enough time period).  There’s still enough though that I’d want to (and did) make the consideration.

The pitcher range was 11.6% to 13.1%, about double the batter range.  The park range was 11.8% to 12.8%, putting it somewhere between the two.

So, in order, you want to know: who the pitcher is, what the park is, and who the batter is.  At the very least, tell us the type of each. So for hitters, you want to know: power or not, speedster or not, LHH or not.

BTW, here’s the link that Rally is talking about:
http://www.baseballthinkfactory.org/files/newsstand/discussion/rlyw_sg_the_wang_effect/

My guess is that if you have the detailed data that MGL does, adding information about the batter won’t help much.  If a guy is hitting .320 on his grounders, he’ll be hitting more of the hard grounders, more into the zones that are tougher to field, or both.  A medium hit groundball straight on to a shortstop is probably almost always going to be an out regardless of who hits it.

I could be wrong though, its probably worth looking into.  Another guess is that looking at BABIP from a separate sample, or hitter’s power, is not going to tell you much beyond what you know from having detailed data, but knowing batter speed might help a lot.

My data is only Retrosheet, plus every now and then I luck into some PBP database.

If something is not random, then it definitely has an impact.  Rather than figuring out how much impact it has (to decide whether to dismiss it or not), we might as well just calculate it (and at least report it as a separate number).

I would *highly* prefer that MGL would do something like this:
runs saved:
(Only including total balls in play, no zones)
+3 runs, by ground balls
+1 run, by line drive
-2 runs, by fly balls

+1 run, by zone for GB
-2 runs, by zone for LD
+1 run, by zone for FB

+4 runs, if we factor in how hard the ball was hit

+2 runs, if we factor in the pitchers
-1 runs, if we factor in the runners on base
+4 runs, if we factor in the batters
+2 runs, if we factor in the parks

+13 runs, final UZR

This way we know exactly how the guy does, and how everything affects him.  After a career, if we find that the “batters” line is all around zero, then we could be safe in ignoring it.

This approach is similar to what I do with catchers.  At first, it seems ridiculous to include balks for catchers.  But, everything is linked.  If a catcher happens to be behind the plate when all the balk calls are happening, compared to some other catcher (for the exact same pitcher), that *means* something.  (Unless of course this poor catcher happens to play during the short period of time when umpires when balk-calling crazy.)

But, if let’s say catchers don’t influence the pitchers’ balks, then we’ll end up with career totals around zero anyway!

It’s no extra effort to count them, so we might as well do so.

Rally:  You may well be right.  I would certainly expect that in an extreme case like Ichiro—whose hit% on GB is 100% points above average—we would find that the out% on his balls is lower than what UZR expects.  So I could imagine a SS or 3Bman on an AL west team geting unfairly dinged in any given year because he faced a lot of tough Ichiro GBs.  But there may be so few players like this that it doesn’t change much.  In any case, I would guess that it wouldn’t be a huge amount of work for MGL to check (easy for me to say, of course).