THE BOOK--Playing The Percentages In Baseball

"We don’t hear anyone talking about how terrible Jeter and Werth are versus RHP”

Yeah, but his leadership and status as a True Yankee cancels it out.

Great stuff MGL. I’ll definitely be applying this in future projections of splits. It never occurred to me before, but it certainly makes sense.

So if the general population deviation is small, we are too assume that prehaps Ryan Howard platoon split is a bit on the flukey side?  And conversely, if the deviation was higher, we’d regress less?

Also, what happens if he performs at .719 OPS against lefties for the next say 5 or 10 years? With the big sample we’d regress far less?

When it comes to platoon splits how many pa’s does it take for the true talent to be right in the middle of league average platoon split and howrd’s career numbers?

Very true-- and the other hugely ignored fact about Howard’s platoon split is that it’s not worth platooning him.  In fact, ignoring the regression issue (which I agree is the correct method), his numbers against LHP are fairly typical numbers against same-handed pitching for 1Bs! (at least going into this year) Last year, I wrote the “The Ryan Howard Can’t Hit Lefties Myth”,
http://www.thegoodphight.com/2009/3/20/805264/the-ryan-howard-can-t-hit
which basically outlined this thinking.

Anyway, another thought I’ve had on the Howard issue is that pitchers have far deeper platoon splits than hitters, and managers almost always bring in deep-splits LOOGYs to get Howard out so the LHP Howard faces on average have deeper platoon splits.  You need to take into account that this is not experimental data, it’s outcomes of strategic interaction.

Great post, MGL.  Agreed thoroughly.

Two things:

1.  Where can one find wOBA splits?  I have only seen splits for OPS.

2.  Over his career since his first full season, here are his OPS totals vs LHP

2006: .923
2007: .826
2008: .746
2009: .653

The decline in his production against LHP has decreased rapidly every single year.  Isn’t it possible that LHPs have found and exploited a problem in his approach at the plate?  Or, do you chalk it up to each season being flukier than the last?

That said, great analysis.  As a Yankee fan, I was hoping for an answer in the affirmative.

Andy talks about it in The Book.  IIRC, he said you need 1000 PA against LHP for a LHH and 2000 PA against LHP for a RHH, to regress 50% toward the league split.  Something like that.  I’m sure someone can quote Andy directly.  Or, you can do a “Look Inside” for free at Amazon.

Great post, MGL.  Two thoughts:

1) Isn’t this a place where using minor league data could be very helpful?  That would give you over 2,000 additional PAs (and seems likely you could look at platoon ratio without having to worry about adjusting for league quality.)

2) It might be helpful to look at the pitch selection Howard sees from LHPs.  For example, if LHPs typically throw 48% FBs against LHHs, but only 30% against Howard (making up #s), we could infer pitchers have discovered he can’t hit breaking balls from LHPs.  That might make it a bit more likely that his poor performance is real.  (And if the % of breaking balls has increased over time, that would make it more likely the apparent trend noted in post 5 is not just noise.)

First, his platoon splits are not all that unusual. Jim Thome is a relatively similar player - with fairly similar platoon splits. Deep regression to the mean is warranted if the pattern is unlikely to be the result of underlying ability - but in this case that is a very reasonable explanation.

Second, he has been seeing fewer fastballs, and an increase in breaking pitches over the last few years. I don’t have the pitch selection for LHP - but I would assume they would follow the same pattern.

I wasn’t aware that people were suggesting Howard be platooned.  In favor of whom?  Dobbs and Stairs are also lefty hitters.  You could move Ibanez to first and play Ben Francisco, but if you want to play Ben it would make more sense to platoon Ibanez.

I’m sure if you ask Joe Girardi, he’d suggest platooning Howard with Miguel Cairo.

This may have been touched on- but my question regards pitchers.  Are platoon splits more salient for this group?

Yes, as discussed in The Book.  Check out the Amazon link, and click “Look Inside”.  You should be able to read it for free.

what caused the selection of “80% toward the league average?” # of PA we have for howard?

The 80% was just a guess.  But, it is based on the number of PA versus LHP.  Andy says 50% regression at 1000 PA, according to Tango, so to regress 80%, we would need 250 PA.  There were actually 750 PA in the sample, so the regression should have been only 57%, which would put his true OPS versus LHP at .779 rather than .805.

I like the idea of using minor league splits if you don’t have a lot of major league data.  You also have to know how much to weight each year for platoon splits, which I don’t really know, in case a player’s true splits do change over time. 

I guess with weighting, you get worse numbers for Howard.

I think it was Joe Sheehan who said something like, “And he should probably be platooned.” Not specifically with the Phillies as if they had someone to platoon him with.  Just in general, as in he should be a platoon player.

And sure, like anything else where we regress toward some mean if we don’t know anything else, if we look at pitch f/x data or even observe Howard (you can often see how come batters bail against same side pitchers and some don’t), we might be able to do better than just a “blind regression.”

Since people are going to quote me, let me quote Andy to make it official, p157:

Recalling the “regression” section of the Toolshed chapter, we find that a right-handed hitter’s platoon split is best estimated with a weighted average in which his platoon split is weighted by the number of lefties he has faced, and the league average is weighted by 2200.

In other words, a righty who has 2200 plate appearances against left-handed pitchers will be regressed exactly halfway toward the league-average. For a lefty, the number is about 1000 (due to the larger variation in platoon skills).
...
For pitchers, the break-even points come sooner (700 for righties and 450 for lefties) and thus are more easily attainable.

Does anyone know how to get baseball-reference.com to figure out the weighted performance of a hitter’s opposing pitchers?

By this I mean:

- for every one of a hitter’s PA’s, note the identity of the pitcher.

- for each of those pitchers, figure out their pitching stats for every PA except those against the hitter in question.

- create a weighted average of these pitchers’ performance, weighting by the number of PA’s each had against the hitter in question.

This would allow us to see how difficult the opposition was, in aggregate, for every hitter, so we could use that to perhaps adjust their stats somehow.

We’d be able to know if Ryan Howard’s numbers are bad vs. LHP in 2009 because of his own contribution to the at bats, or because he preferentially was facing tougher competition (e.g. facing a LHP specialist virtually every game because he’s feared, and still good enough against them to never be hit for…

I don’t think B-ref has any way to do this easily right now…

That would surely be a great addition, Greg.  We often want to know the quality of opposition.

On thing I do in my projections is I adjust all player stats by component, by the quality of their opposition.  Not for platoon ratios, but for overall stats.

I like this.  Just a question for clarification.  When you write: “.805 is higher than Werth and Jeter versus RHP. We don’t hear anyone talking about how terrible Jeter and Werth are versus RHP, which they face most of the time, do we?”

Have you regressed Jeter and Werth’s platoon splits as well, or are those figures raw?

Thanks.

Ok, so going by the above text from the book we’re creating a weighted average of Howard’s platoon split and the league’s average platoon split.  For the NL last year the platoon split was .088 of OPS for LHB.  Howard’s split was .316.  So, The Book is saying using a weighting factor of the number of PAs a LHB has against a RHP with the .316 in Howard’s case, and 1000 for the .088 from the league.  So that’s (2085*.316 + 1000*.088)/(2085+1000) = .242.  So then using your 62% of his ABs against RHP and a true OPS of .930, I’m getting a true platoon split against LHP of .780.  Now that’s for his career, not the last 4 years, just because it was easier for me.  But that still looks like a large platoon split to try and exploit.  If you can turn Ryan Howard into Carlos Ruiz, that’s pretty good.  Though I wouldn’t recommend getting a RHB to platoon with him.  It may help, but its probably not worth the roster spot.

The guy has a lot of PA’s agaisnt lefties and the numbers are really pretty bad for him. So his splits vs lefties might be a little low but I think he is still pretty bad agaisnt LH’s overall. And I think what they meant by “platoon” is if they happened to have somebody who could crush LH pitching then it is possible to get more production out of your first base position then just with Ryan Howard. Seems crazy but isnt the point of an offense to score as many runs as possible? What if the Phillies had a Matt Diaz who crushed LH pitching last year. If they could get a .950 ops out of Howard vs RHP and a .900 (just estimated numbers) ops out of Diaz vs LHP wouldnt that be better? Seems insane to platoon Howard with little known Diaz but it would result in more production. But the Phillies dont really have that type of person but I’m just saying it wasnt that crazy of an idea.

It seems like a bit of chicanery to weight Howard’s platoon splits and compare them with Jeter and Werth’s unweighted splits.