THE BOOK--Playing The Percentages In Baseball

Thanks for shining light on that series by Ken Arneson; I wasn’t aware of it, and it’s a great read.

As an AL West observer, I’ve long noted Zito’s reverse-platoon tendencies.  A few years ago, I came to the conclusion that it was his cut fastball (which Arneson labels the “jammer” in part three of his series) that was his equalizer against right-handed batters.  Arneson comes to a similar conclusion.

Your observation of his lower BABIP against RHB, as well as Ken’s conclusion that Zito allows easier-to-catch flyballs against RHB, support this contention from an anecdotal standpoint.  It seems that a fastball in on the hands of an RHB from a LHP would lead to an inordinate number of popups; in fact, his Infield Fly Per Fly Ball over his career is above 16%, which appears to be a very high figure; Matt Cain led baseball in that figure with a nearly identical figure, and Zito’s 13% (the lowest of his career) ranked 12th amongst major league qualifiers.

It would be great to have the pitch location data to verify this connection.  But it’s nice to see subjective and objective analyses complement each other.

Looking through my archives regarding Zito, it occurred to me that I have looked at other cut fastball pitchers who seem to exhibit reverse-platoon tendencies.

Mariano Rivera has the most acclaimed cutter of our generation; his career BABIP against:
LHB:  .253
RHB:  .285

His career OPS against:
LHB:  .520
RHB:  .602

Now, there’s a bit of a selection bias in that Rivera likely faces an inordinate number of bad LHB (opposing managers not wanting RHB to bat against him may pinch-hit inferior LHB), but that is a striking difference.

Another pitcher I long watched who featured a cutter was Jim Abbott.  His career BABIP against:
LHB:  .325
RHB:  .296

His career OPS against:
LHB:  .817
RHB:  .724

Rivera’s career Infield Fly Per Fly Ball is slightly higher than Zito’s.  Abbott’s is unknown, but it seems like someone with actual should be able to parse Retrosheet for such figures for players of his era.

FWIW, I took all pitchers and filtered for >= 600 Batted Balls, since 04 (courtesy of THT), I get 672 batted balls, and 182 pitchers, I get R = -.4 for the correlation between [GB% of Batted Balls] and IF/F.  The average IF/F is 11.54% from that sample

Barry Zito IF/F, via Fangraphs

2002 18.5 %
2003 17.6 %
2004 17.6 %
2005 16.9 %
2006 13.3 %

TOTAL 16.2 %

I’m not sure what his Home/Road splits were like.  Fangraphs has game logs, but doesn’t split out IF and OF in them.  It’d be interesting to see what his IF/F numbers looked like on the road.

There are some interesting differences in the leagues.  I get 12.00% IF/F for AL and 11.05% for NL from the above sample, and the correlation is way different (-.52 for NL, and -.26 for AL).  The correlation for LD/Air to GB% remains strong despite the league (.75 for NL, and .69 for AL).  I suspect pitcher vs. DH has a lot to do with those differences.  Of course I only had 94 pitchers in the AL and 88 in the NL sample.  I calculated the standard deviation of the z-scores seperately for each league, I get average 673 batted balls in the AL, and 671 in the NL, stdev(Z-Score) is 1.15 in the AL and 1.61 NL.  Not sure what that tells me exactly, but it’s interesting.

In Zito’s case, the “Jammer” is a four-seam fastball, not a cutter (two-seamer).  Zito does throw a two-seamer, but only rarely.  It’s a new pitch for him.

Hey guys, great stuff here. I was wondering the effect that The Coliseum had on Zito’s IFFb% or easy-to-catch fly balls. Anyway, I couldn’t find any game logs with IFFB’s so I just looked at fly balls. Here’s what I found:

2006
Home fly balls: 153
Road fly balls: 148
Per inning: 1.40 home 1.32 road
2005
Home fly balls: 137
Road fly balls: 112
Per inning: 1.16 home 1.01 road
2004
Home fly balls: 164
Road fly balls: 126
Per inning: 1.48 home 1.24 road

So, It appears he gets a lot more fly balls in Oakland. Do you think there is anything significant here?

Well, for some reason I think I was thinking that FB’s were fly ball outs, and obviously they are not...so maybe what I did there doesn’t make much sense. And the additional fly balls could be caused by the fact that Zito pitched worse at home, and faced more hitters…

But, I think there still may be something to the fact that Zito gets more fly balls at home, and perhaps more outs via the fly balls(and the large foul territory). Im not sure now, though?

I guess my question would be...How much are his high IFFB rates aided by that big foul territory, and by how much would that advantage disappear when he doesn’t pitch in Oakland?

Anyway, here’s what happened to Hudson and Mulder:
Hudson IFFB%(from fangraphs): 9.1 12.7 13.2...left oak… 6.3 10.3

Mulder IFFB%: 9.1 12.0 4.7...left oak...6.5 2.6

Fly calls caught in foul ground(on the infield) are considered iffb’s right? Anyway, I dont know if there is anything here....

Well, I just got a real-world example in why not to use Retrosheet batted ball data.  It shows 34% of his flyballs were pop-ups in 05 and 33% in 06.

MB, I tried getting his IF/F for road games from Fangraphs, but it doesn’t look like they seperate IF and OF on the game logs.

#2: I’ve seen quite a few managers pinch hit righties against Rivera. I’ve also seen several switch hitters bat righthanded against him. Opponents simply don’t play the lefty/righty game with Mo.

The analysis for Zito still needs to be deepened. I think Tango’s conclusion is likely to be overstated.

Some notes:
1) for the purpose in question, “BABIP” should include SF as part of the calculation. Baseball reference does not calculate it that way; counting SF shaves Zito’s “platoon difference” from 42 to 40 points.
2) as a starter, Zito cannot be compared to average for all LHP without adjusting for opponent batter quality. Over his career he has had 21.0% of his TBF to left handed batters, vs 28.1% LHB faced by all A.L. LHP over the same period. This should be an indication that generally only above average LHB hit against him and instead they are replaced in the opposing lineup by more marginal RHB.  Holding those LHB to an average performance should be a sign that he has an above average ability against them. Conversely, his superior performance against RHB may need to be discounted .
3) For what it’s worth, Zito’s G/F ration over his career is 1.02 vs LHB and 0.84 vs RHB. How much of that is different pitching approach and how much is differences in the quality of the opposing batter pool is unknown without more research.
4) Zito’s ERA may be worse at home, but he does appear to enjoy a substantial advantage in foulouts at home. I count 23 IF foulouts at home in 2005 vs 10 on the road and 4 OF foulouts vs 3. In 2006, it’s 21 at home vs 5 on the road for IF, and 2 vs 0 for OF. As a thought experiment in park adjustment, if those margins held throughout his career, that’s trading an extra 0x15 per year for a continued at bat with an extra strike. In some of those hypothetical at bats, a ball in play will not result; a few that did would conclude with a hit. Foulouts in his home park may have be worth 6-7 points in BABIP.
5) the retrosheet popups are accurate, reflecting balls caught by infielders, regardless of where. Perhaps fangraphs defines IFFB more restrictively, e.g. eliminating balls caught on the outfield grass?

(Hmmm… my first lost post on this site.  Someone’s gotta talk to the admin.  Luckily, I cut/paste.)

==========================

No question SF needs to be included.  And I’d argue to include RBOE, or at the least, remove it from the numerator AND denominator.

***

In The Book, Andy looks at platoon splits (wOBA, not BABIP), and Zito has the #3 reverse platoon split for a LP, starter.  Another leader was Moyer.  (I think the data was 2001-05).  Andy also showed the top starters vs LHB, using skill not performance, and Zito was #3 against LH and not on the list of top 10 (as opposed to Santana and RJ who were 1 and 2 against both LH and RH).

***

Insofar as DIPS is concerned, it cannot be stressed enough that as your sample size goes up, your correlation goes up.  Even if you were to regress say 80% toward the mean if you have say 500 BIP, this would make your 50% regression equation at 2000 BIP.  So, if you are now at say 3000 BIP, your regression toward the mean is now down to 40%.

If you have even more reason for a guy to have a different population mean than league average, then you use that information.  Blanket assumptions of regressing everyone to the same mean is something you start with, not end with.

Tango -

I see no indication in The Book that Andy adjusted for quality of opponent batter. It seems to me that he just applied regression to the mean to observed aggregate performance.

I do know that Randy Johnson is severe in the respect in facing so few lefties.  Zito may have faced an average number of lefties.  I don’t know though.

In Zito’s case, the “Jammer” is a four-seam fastball, not a cutter (two-seamer).  Zito does throw a two-seamer, but only rarely.  It’s a new pitch for him.

Ken, I’ll definitely defer to your expertise on it, but I’m not completely crazy—at least one major league scout calls it a cutter, too!

And, generally, the two-seamer is a sinker, not necessarily a cutter.  I do believe there is some variation from pitcher to pitcher in how they get the ball to “cut”, though.

Just going through some top lefties.  This is the % of LHH faced in their career:

RJ: 12%

Willis: 16%

Glavine: 20%
Pettite: 21%
Zito: 21%
Lilly: 21%
Santana: 22%
Moyer: 24%
Buerhle: 24%

(The league average for a lefty is 28%.  The league average for a righty pitcher is 48% of batters faced are lefties.)

So, I don’t think there’s anything special about Zito’s opposing hitter quality, relative to his peers.  It’d the the same as any star lefty… except the incomparable RJ.

RJ has a 23% normal split.  Lilly has a 4pt normal split.  Buerhle 18pt split.

Glavine has a 13pt reverse split. Pettite has a 16pt reverse split.  Santana has a 2pt reverse split.  Moyer has a 7pt reverse split. Willis 9pt reverse.

Zito has a 32pt reverse split.

(Joe is right that we would want to regress based on that.  I seem to remember Andy telling me he did that, but I’ll have to ask.)

On page 85-87 in The Book, I do show how you have to be very careful in looking at splits, since it’s possible to show zero split by increasing the number of bad hitters faced on one side.  It just “looks” like you have no split.

Definitely lots of research potential here, for lefty/righty splits, based on quality of opposing hitter, and looking at BABIP and K/BB percentages.

Ken Arneson comes back with more Barry Zito:
http://catfishstew.baseballtoaster.com/archives/724046.html

He shows the BABIP of RHH against Zito (Zito, a lefty, shows severe reverse platoon splits) as .245 this year.

http://www.baseball-reference.com/pi/psplit.cgi?n1=zitoba01

Zito’s career is .257 v RHH and .295 v LHH.

In 2007, he’s .245 and .... .368 (!!!).

As Ken points out, Zito now faces 25.2% of batters as LH, as the league likely figures out that Zito has reverse platoon splits (career 20.8%).

However, his K/BB ratio against RHH is horrible (46/39), while his career shows a 2/1 split.

It is basically impossible for a MLB pitcher to have a .123 reverse platoon split on BABIP. 

I don’t think that half of Barry Zito has turned into Jeff Weaver.

Looking at the list for splits against left-handed pitchers, does it not look like the better, and/or higher velocity pitchers face fewer left-handed hitters? This is my assumption, but wouldn’t you expect managers to switch out more left-handed hitters for right-handers against the better/harder throwing left-handed pitchers?

It really shouldn’t matter in terms of total expected change in offense whether you change out more or less left-handers against the better pitchers, but my guess is that’s what managers do.