THE BOOK--Playing The Percentages In Baseball

I’m curious why you chose 4 years for Ibanez’ platoon splits.  I don’t think you’re cherry-picking or anything, but I think it’s kind of an interesting time frame.  If we look over a 4-year window, we would be looking at Ibanez with a .163 OPS platoon disadvantage at the end of 2007, instead of his .100 OPS platoon disadvantage today.

And in 2007, his 4-year performance vs. RHP was pretty much the same as his 4-year performance vs. RHP is today.  The main boost he gets is that he’s hit better against LHP the last couple of years.

From a quick google search, it looks like from 2000-6, LHB had a .084 OPS platoon disadvantage.  While OPS can lie, I don’t think it is here, and it seems like you’d have to weight the last couple of years pretty heavily to have him at a below-average platoon split (definitely to get a platoon split that is “a lot less” than average.) And while I understand weighing more recent performance heavily, it seems odd to me that Ibanez would suddenly learn how to hit left-handed pitching at this point in his career.  It wouldn’t be impossible, but it seems improbable, especially for someone who just two years ago you might have said was especially bad against left-handed pitchers.

No reason.  I just went to ESPN.com and looked at the 3-year (06-08) plus 09.  The correct thing to do would be to use all years but weight each subsequent year more heavily, which pretty much eliminates anything more than 4 years back anyway, or at least minimizes its value.  There is really nothing in baseball where it helps much to go beyond 4 years when trying to estimate true talent or construct a projection.

Last night in that chat, someone (I forget who) suggested Mariano Rivera in that situation.  Of course, you’re not allowed to bring in your closer to the 8th inning in a game you aren’t winning.  It’s in the rulebook.

Girardi did exactly that in game 3 and 5 against the Angels.  Yankees don’t think Rivera can pitch 2 innings every single day, so Girardi has to weigh the benefit of bringing in Rivera there against the benefit of having him for 2 innings tonight.

I can understand, in that situation, using your other relievers to try and get out of it.  I wouldn’t have used Coke though.  I prefer Pepsi.

In isolation, having Mo in that situation (hi lev) is fine.  However, if you pitch him to start the next inning, the LI will be low.  Overall, his LI for the game (say he faces 4 batters) will be under 1.  Not terrible, but not good.

Furthermore, you hamper yourself for today’s game.  He won’t be able to go 2 today.

Now, if you want to argue that you bring him in ONLY for 1 batter, that’s another story.  The problem is the warmup time.  A better bet would be to have a starter on his throwday ready for that, like Greg Maddux did once.

Reading newspapers and fan blogposts regarding Girardi is a good example of why it is tough to stray away from the norm as a coach/manager in sports.  They get skewered for doing anything differently if the results are negative.  But they don’t get much credit when the results are positive - the credit goes to the players.  There is little incentive for a manger/coach to deviate from the norm due to job safety ... and there is also selection bias, as the guys that show they do deviate from the norm probably get selected NOT to manage/coach a team. 

I’m not saying Girardi is making all the right moves.  I am saying he’s getting blamed for making any odd-looking move (such as using replacing a righty for another righty) based on results only.  When it goes right, he gets no credit...it was all Mo’s great pitching, or all Joba is back!  etc. etc.

I’d bring Rivera in for one batter.  He’s far and away the best pitcher against LHB.  It’s a high leverage situation.

Why hold him out on the hope that you might have two consecutive high leverage innings late in the game tonight?  What are the odds of that happening, anyway?

Rivera’s career numbers against LHB: .206/.256/.261 avg/obp/slg.  That’s in 2266 PA, so it probably doesn’t need to be regressed much.  You don’t have anyone else available in the bullpen, starter or otherwise, who can approach that, do you? 

Andy Pettitte against LHB is .271/.318/.393 in 2719 PA.  Give him a boost for relieving instead of starting, but he’s not as good as he was at his peak, either, so on balance he’s not nearly as good as Rivera.

Where would you estimate Coke’s true talent vs. one LHB as compared to Rivera or Pettitte?

I don’t want to bring Mo in for just one batter.  Say he gets out of the inning, you bring in a new pitcher in the 9th, and then the Yankees tie the game.  He’s gone before your extra innings.

If the score was 1-0 instead of 2-0, that makes it more likely that using Mo in the 8th maximizes his value.  At 2-0 they probably thought they’d have better opportunities to use him in game 2.

So MGL, especially with weighting recent performance more heavily, you would have concluded that Ibanez had a really large platoon split in 2007 and a smaller than average platoon split in 2009.  Almost to the point where it would be the difference between bringing in Coke or keeping Robertson in.  Doesn’t that seem a bit wacky? 

I know player talent levels can change, but if you weight the last couple of years of platoon splits heavily in your opinion of a player’s “true” platoon split, you’re essentially putting a large weight on 300-400 PA against left-handed pitchers.  To me, it seems like if you are going to heavily weight that few PA, the split is going to have to be significantly regressed.

And overall I’m more inclined to think that Ibanez wasn’t especially bad against LHP in 2007 or especially good against LHP in 2009, but rather had a close-to-typical left-handed platoon split all along.

He’s regressing his splits heavily.  He’s adding 1000 PA to the number of PA he faced against LHP.

Tango, that’s not what it looks like to me, and I still don’t see any way to really justify that Ibanez has a platoon split “a lot less” than average.

"So MGL, especially with weighting recent performance more heavily, you would have concluded that Ibanez had a really large platoon split in 2007 and a smaller than average platoon split in 2009.  Almost to the point where it would be the difference between bringing in Coke or keeping Robertson in.  Doesn’t that seem a bit wacky? “

That’s just the way it is.  That is why it is only an “estimate.” Sometimes, after the fact, our estimates turn out to be way off.  If a player hits .700 (OPS) in year 1 and we have a .750 projection for him and then he hits .900 in year 2 and we then have an .820 projection for him, did we make any “mistakes?” Is that “wacky?” No, that’s just the way it is and that’s all we can do.  We obviously made a “mistake” but there is nothing we can or could do about it, other than next time using a crystal ball or consulting a psychic or soothsayer.

I honestly don’t know how much to weight platoon splits from year to year.  To tell you the truth, I don’t think it is that much, because I don’t think that the true talent changes much over time.  In fact, I don’t weight them at all in the research that I do. I suspect that some weighting is correct, but it doesn’t really matter if it is 6/7/8, 1/1/1, or 3/4/5. It’s not that big a deal.

It might be all that you can do, but when your estimate is showing that much change over two years and you don’t think that true talent changes much over time, it seems that your confidence in the estimate of Ibanez’s platoon split is awfully low.  Low enough where you can’t really say definitively that it is different than average, and it seems like you should just stick to assuming he has an average platoon split for a LHB.  Low enough where it’s almost not even worth bothering to look up Ibanez’s platoon splits.

#13, no that’s not true.  If you don’t think that “real” (true) platoon splits do change much over the years, and I think that is a reasonbale assumption, that is all the MORE reason to pay no attention to the fact that a player’s sample (actual) platoon split has jumped around a lot over the years.

Your logic is completely wrong I am afraid.  Sorry, and with all due respect, but it is not even worth discussing.

Platoon splits WILL necessarily jump around a lot by sheer chance simply because the sample sizes are relatively small for one year of data (most batters don’t have more than 200 PA versus lefties).

In any case, if you are starting with the assumption that platoon talent doesn’t really change much over a player’s career, then this platoon ratio for the last 5 years:

1.00
1.40
1.10
1.50
.75 (reverse)

is EXACTLY the same as this:

1.15
1.15
1.15
1.15

If you don’t think that “real” (true) platoon splits do change much over the years, and I think that is a reasonbale assumption, that is all the MORE reason to pay no attention to the fact that a player’s sample (actual) platoon split has jumped around a lot over the years.

Where on earth did I say that we should pay more attention to year-to-year jumps in Ibanez’s splits?  I’m arguing the exact opposite of that.  We shouldn’t even be paying attention to jumps in Ibanez’s 4-year splits because our confidence in them is so low.  (Because the sample size is so low.)

You have stated that you don’t think a player’s true talent changes that much over time.  I agree with that.

You’re saying that with big jumps in year-to-year data, we should ignore the year-to-year jumps in his platoon split.  I agree with that.

I am simply saying that we should go even farther than that.  Seeing Ibanez’s 4-year platoon split jump so much from 2007 to 2009 makes me question even the validity of a 4-year platoon split.  At which point I have little confidence in any claim to Ibanez’s real platoon split until someone is actually taking 4-5 years of data, doing a regression, and quoting the regressed split as his true talent.

And the regression in this case would look Ibanez look like he had a more typical split against LHP in 2007 and in 2009.  And my thinking is that the regression against the data is going to be so large in general that it’s practically not worth looking up platoon splits for most players.

Your logic is completely wrong I am afraid.  Sorry, and with all due respect, but it is not even worth discussing.

There is really no need to patronize me.  You are free to state your opinion without apologizing for it, or fearing for anyone’s logic.