THE BOOK--Playing The Percentages In Baseball

Excellent work by Nick and also by Colin from what I know of his article (though I haven’t read the whole thing).

I love reading that kind of stuff, and it’s great work by these guys to get it out so fresh off the presses.

And there’s no way to hit that guy.  With that kind of command, you couldn’t hit Jamie Moyer’s stuff.  With Strasburg’s stuff...I honestly don’t know what I would recommend to a team.

Maybe tell them to “sit” on his offspeed stuff?  Yeah, you’re gonna look really silly on the fastball and I wouldn’t want to be the one standing in the box trying to do that, but does anybody have a better idea?  Short of just setting the bat on your shoulder and going back to the dugout?

You can add another good article to the list.  This one by J-Doug:
http://www.beyondtheboxscore.com/2010/6/9/1508668/hype-fails-to-live-up-to-stephen

Thanks for the nice words Mike.  I would say just sit on his curveball.  That was the only pitch he hung at all last night, but batters were so taken aback by it that they just watched it go in or flailed at it.

Then again, people have a ton of trouble sitting on Wainwright’s curveball and he throws it slower and his fastball isn’t nearly as good.

First of all, you can’t ever just “sit on a curveball” until the chance of a curveball coming is real high, since when you sit on off-speed, you cannot adjust to the fastball, but you can do the opposite (sit on fastball and still hit an off-speed).  Secondly, and most importantly, have you guys never heard of game theory?  If a batter sits on curveball even one percent more than they should, it would be correct to throw them nothing but fastballs.

“With that kind of command, you couldn’t hit Jamie Moyer’s stuff.”

Mike, that is silly. Greg Maddux had the best command I have ever seen. I doubt that Strasburg will ever have nearly the same command that he did.  And Maddux had 100 times the stuff that Moyer has.  In his heyday, Maddux had an 88-90 mph tailing fastball, a decent curve, and a great circle change.  Moyer with Maddux’ control would be a lot worse than Maddux, and Strasburg does not have nearly Maddux’ control, if that makes any sense.

you can’t ever just “sit on a curveball” until the chance of a curveball coming is real high, since when you sit on off-speed, you cannot adjust to the fastball

I said that.  But the Pirates were trying to do what you recommended, and look where it got them.  I know that’s how you approach most pitchers.  Not sure it works as well in this case.

Of course I know game theory.  I’m talking in relation to the approach against a typical pitcher.

My point is this: against the fastball they whiffed 17% of the time.  That is unheard of.  That’s 3x normal.  Perhaps the traditional approach isn’t working?

Btw, I probably should make it clear that post #2 contained a bit of hyperbole.  I thought it was obvious when I said “there’s no way to hit that guy”.  The Pirates did get 4 hits, including one home run.  So unless I’m blind or stupid, I’m not claiming there is literally no way to get a hit against Strasburg.

Tango had a good point in the other thread that part of what he was calling “location” included the knowledge of where to put the pitch to make it hard for that particular batter to hit, given how he was leaning on the last pitch, what he tendencies and weaknesses are, etc.

My understanding is that that is where Maddux excelled.  I doubt Strasburg is anywhere near him in that capacity.  For all I know he could be well below average in that, though presumably Pudge can help him along somewhat.

Greg Maddux had the best command I have ever seen. I doubt that Strasburg will ever have nearly the same command that he did.  And Maddux had 100 times the stuff that Moyer has.  In his heyday, Maddux had an 88-90 mph tailing fastball, a decent curve, and a great circle change.  Moyer with Maddux’ control would be a lot worse than Maddux, and Strasburg does not have nearly Maddux’ control, if that makes any sense.

I’m not following the part where you say that Moyer with Maddux’s control would be a lot worse than Maddux.  Moyer is/was worse than Maddux, although I don’t think I’d say a lot worse.  Maddux is Hall of Fame caliber.  Moyer of Hall of Good.  So are you saying that Moyer had/has better command than Maddux, or that Moyer stuff is better than we give it credit for, or what?

I’m saying that Strasburg did display among-the-league-best type of command last night.  By command, I mean the ability to repeat your delivery such that you can place the ball where you want it in both dimensions (along the arm path and side-to-side).  Strasburg’s groupings were very tight.  Some people were complaining about the location on his curveball, but if you compare to most other pitchers, that distribution of locations is actually really well controlled.

What I meant was this:

Maddux control + Maddux stuff = ERA of 63% league average.

Maddux control + Moyer’s stuff probably equals 75% of league average, since Moyer’s stuff is not as good as Maddux’ stuff (although it is apples and oranges).

Since Strasburg’s control is less than Maddux’ control (in my opinion), then:

Moyer’s stuff + Strasburg’s control has to be worse than 75% of league average.  That is a far cry from “you can’t hit him...” even with the hyperbole factored out (hyperbole-neutral statement).

As far as sitting on the curve, the bottom line is that a hitter cannot sit on a curve for any significant percentage of the time (or to any significant extent) otherwise Strasburg will simply throw all fastballs and he WILL be almost unhittable.  You can’t talk about what a hitter should do differently and assume that the pitcher won’t change his approach.  That is the essence of game-theory with respect to the hitting/pitching matchup.  That is why any suggestion that batters “sit on the curve ball”, no matter how you want to spin it, is wrong.  My guess is that batters should look for the curveball maybe 40% of the time at most (and it might be 20%) against Strasburg.

MGL, I’ll agree with all of post #8, except I’ll reserve judgment on this part:

Strasburg’s control is less than Maddux’ control

I dunno Mike.  I didn’t think Strasburg’s control was that great, certainly not approaching Maddux.  What makes you think it was so amazing?  Is this just an observation or did you plot out the catcher glove location or something?

If Strasburg’s control is better than Maddux’s control, then this would make him a better pitcher than Pedro.

So, uh, yeah, we better reserve judgement on the control part.  You’ll probably need at least 2000 batters faced before we can talk about his control being that good.

"Maybe tell them to “sit” on his offspeed stuff?”

I think you have to guess on him.  A fewer Pirates were able to hit the fastball the other way for a single.  Good luck trying to pull 99 MPH.  And the homerun was on a curveball.  With most pitchers you can look for one thing but still ready to adjust to something else.  Against Strasburg, just be a guess hitter.  Guess wrong and you’ll look foolish, but with his stuff the odds are you’ll look bad anyway.

Sitting on the 1B side of the field, I can usually follow the ball out of the pitcher’s hand.  With Strasburg, I could not see his fastball.  For most of my life I wished I had more ability and could be a player instead of a fan.  Last night I was happy to be a fan, and grateful I didn’t have to take a bat up there and face him.

Homerun was on a changeup at the knees, but middle pretty much in the center of the plate.  His changeup location was absolutely filthy last night, all of them either just below the strike zone or just outside to lefties and with incredible sink, and that was the only changeup in the strike zone.

I would say you pretty much have to guess with him.  You can hit his two-seamer as it doesn’t have great movement and isn’t too overwhelming at the lower end velocities (around 95 MPH).  You can hit is curveball because he hung that up a lot last night and it’s not a great enough pitch on it’s own (although it’s pretty freaking good).  You aren’t going to touch his fourseamer or his changeup without some good luck.

"Filthy”, “sick” and other terms like that: I’m not a fan.  Am I the only one?

I especially can’t stand it when they are used more than once, like “filthy, filthy pitch”.

I don’t know about Maddux.  I’m not really comfortable comparing to him.  I would have to go back and look at his data.

What I’m saying is that Strasburg, in this one start, was on par with the command of some of the best-command pitchers from the PITCHf/x era that I’ve looked at in detail:  Cliff Lee and Jamie Moyer.

My definition of command is outlined in the 2009 THT Annual.  I can possibly go dredge it up.  Basically, it’s what I said earlier:
“I mean the ability to repeat your delivery such that you can place the ball where you want it in both dimensions (along the arm path and side-to-side).”

I don’t think that’s necessarily what all people mean by command, so I’m specifying what I mean.  It’s something that can be measured in the existing PITCHf/x data.

Catcher glove data would be great.  Somebody’s collecting it, since MASN was displaying it last night.  Anecdotally, I was impressed a number of times when they replayed the PitchTrax with the catcher target how close Strasburg was in the side-to-side dimension.  But who knows if the sample they displayed was representative.  Probably not. 

I do remember seeing him miss the target badly once (a changeup that went inside when Pudge had set the target outside).  No MASN PitchTrax on that one. I wasn’t paying attention the whole game whether he was hitting the target or not, but from when I was paying attention he was doing quite well.

When I pull the data from the last game from Cliff Lee and this game from Strasburg, it does look like Lee’s command was a little better.  So maybe I was unduly influenced by what I saw.  I haven’t run numbers.  A command calculation is a bit of a pain to do.

I’d say he has good control in that he can throw strikes but it doesn’t seem like he’s masterful at placing the pitch anywhere inside the strike zone like Greg Maddux was. 

He basically just throws his curve right down the middle (if it’s going to be a strike at least) and he threw his fastball right down the middle to righties (and outside to lefties.)

This isn’t a knock - when you are that young and have that kind of movement it would be kind of ridiculous to expect you to have Madduxian control - but I think he has room to improve which is pretty scary

Nick/10,

“I didn’t think Strasburg’s control was that great, certainly not approaching Maddux.”

I agree with you on this one.  His changeup looked like it was in the right place overall, but watching the game in person, it seemed like he missed the glove with the fastball more than you’d like.  Obviously, that’s pretty lame evidence (and when you throw 100 it doesn’t matter AS much).  His pitch command was great, but I’m not sure he was throwing to the location he always wanted to be throwing.  It’s tough to tell that in Pitch F/X because we don’t know ‘where’ the ball was ‘supposed’ to go.  I think you describe the outing extremely well in your post.

It’s just a buzzword Tom.  When I say “filthy” it’s my way of saying “possessing about 5 inches more drop than the average RHH changeup while being thrown 7 MPH slower than his fastball with the same armspeed, all while being located an inch or two below the strike zone”.

I NEVER SAID STRASBURG HAD COMMAND LIKE MADDUX.

Ok, I needed to get that out of my system.

I think people would be surprised how often and how badly most pitchers, even the good ones, miss the catcher’s target.  Chart a few games and see if you don’t come to that conclusion along with me.

I don’t know about Maddux, but I’ve not seen any evidence that pitchers can place a pitch with pinpoint control.  They aim for one edge or the other and hope it comes close.  With curveballs they don’t even do that.  They aim for the middle and hope they get a strike or swing and miss.

Maddux excepted, of course.  I’m not going to impugn the great Maddux.  I’m sure he could take the left leg off a gnat at 60 feet.

Not to be a buzzkill, but part of the reason the Pirates were swinging and missing at so many pitches last night is because the Pirates suck eggs. 

Seriously, that was one putrid line-up they put out there.  It will be interesting to see what happens when he falls behind Shin-Soo Choo on Sunday and he can sit on a fastball.

"Sitting on the 1B side of the field, I can usually follow the ball out of the pitcher’s hand.”

Rally:  I was on the 1B side as well, section 133.  Where were you?  We should catch a game some time, and discuss fielding metrics. :>)

It was a damn impressive performance, especially the 6th and 7th innings.

I know it’s a buzzword.  I see it all.the.time.  It is to me what “alot” is to many people here.

Not to be a buzzkill, but part of the reason the Pirates were swinging and missing at so many pitches last night is because the Pirates suck eggs.

Looking at teams by wOBA, the Indians (.308) are bad, though a little better than the Pirates (.298). As it stands now, the Nationals aren’t scheduled to face a plus-offense team until the Reds (tops in the NL) in late July. Even if Strasburg doesn’t start that series, the Nationals follow that one by facing the Brewers, second in the NL.

It was a pretty bad offense, but not “putrid” bad or historically bad.  I had them around 53 runs collectively below average or 1/3 run per game.  Cle offense is equally bad.  53 runs below average.  The only two plus (better than average) players in their lineup are Choo and Branyan, according to my current projections.  And I have the NL and AL hitting around the same, maybe even a slight edge to the NL’ers these days.

I don’t think you can really “see” the difference between a -1/3 run lineup or an average lineup is 6 or 7 innings of play.  Now, if one team is a high K team and the other is not, that you can probably “see...”

And McCutchen is just as good of a hitter as Choo is.  He went 0-3 with a k against Strasburg (although he did line out hard in the first).

Choo may be able to sit on a fastball, but the likelihood of him squaring up a 99 MPH pitch up in the zone is low.

I’m pretty sure that projection doesn’t reflect the line-up Pittsburgh ran out last night.  I’m assuming you have pretty decent projections for guys like Jeff Clement, Ryan Doumit, and Akinori Iwamura (all guys that Chone/Zips liked well enough), none of whom played last night.  They were replaced by scrubs Delwyn Young and Jason Jaramillo. 

2010 OBP by batting order for the Pirates line-up last night:

.374, .370, .320, .333, .275, .309, .286, .245, Pitcher

That line-up blows.

Andrew McCutchen’s rest of season wOBA, per ZIPS: .359

Shin-Soo Choo’s rest of season wOBA, per ZIPS: .380

McCutchen is a righty.  Choo is a lefty. 

So, no, they’re not the same, and it’s not even close to the same thing.

.20 points of wOBA is “not every close” these days?  And is our confidence in projections so high that we can take that difference as significant?  Even so, most of the difference between the two via ZIPS is with regards to walk rate.  They each project for the same average and there is a .20 point difference in ISO.  Unless you are talking about Choo’s ability to walk while sitting on a fastball, the two are essentially as dangerous.

Dave, the May 3 San Diego lineup that you’re touting as the real deal had a projected wOBA of .290 according to Oliver, same as the players in the Pittsburgh lineup that Strasburg faced last night.  The Cleveland lineup last night had a .320 projected wOBA, but two-thirds of the difference between Cleveland and those two NL lineups is the DH replacing the pitcher.

I don’t see where the lineup is a significant factor here.

Yes, that was their exact lineup the other day: -53 runs per 150.

When you project two players to be X amount different in wOBA (or any metric), to me the uncertainty of each projection doesn’t matter in terms of who I think is better.  Each one is a mean estimate and so the mean estimate of the difference is simply the different between the mean estimates.  Obviously the uncertainty of the claim that one is better than the other is larger if the uncertainty around one or both players is large, but that goes without saying.

And yes, I would agree that 21 points of wOBA when one player is lefty and the other is righty is a lot versus a RHP.

MGL in #5 describing Maddux makes me wish there was some PITCHf/x system around for Maddux’s hey-day.

pitch f/x does NOT tell us much about a pitcher’s command.  Obviously it can tell us the percentage of each type of pitch to each location and from that we can infer something about command, but without “target” information, pitch f/x is pretty worthless for analyzing and evaluating command.

Not that you need command information to evaluate or analyze a pitcher’s effectiveness.  You don’t. Wherever his pitches end up, that is what determines his effectiveness, whether by design or not.  But if I were a team and I were trying to help my pitchers get better, I sure as heck would want to know their target on every pitch.  Plus knowing that can help you separate the luck from the skill in terms of actual performance…

MGL/32, I beg to differ.  Did you read my THT 09 Annual article on the topic?

On that note, I hate when we have two or more threads going at the same time on the same topic.  You can refer to my comments in the other Strasburg.  I don’t feel like constantly repeating everything in both threads.

I think I know what you mean and I think we differ only in semantics, perhaps in degree.  But it is tautological that we can’t FULLY determine a pitcher’s command/control without knowing their target. I can easily prove that, although it is so obvious that I shouldn’t have to.

Here are 3 pitches:

1) Fastball on the low outside corner.

2) slider low and in to a RHB from a LHP.

3) High inside fastball.

How is the pitcher’s control/command?  Obviously we have no idea unless we know the intended location.

You are going to respond with something about “sample size”, but clearly there is no magical point at which we can infer command/control from a sample of pitches.  If we know the intended location we don’t need any particular sample size to determine control/command.  One pitch or 1000 pitches will suffice.  So knowing the target will always give us much better info regarding command/control than not knowing the target.  Much better.

Now, I would not be surprised if you could construct an algorithm to infer command/control given a large sample of pitches. But…

One, you need a large sample otherwise any algorithm is going to be pretty worthless.  Two, no matter how large your sample, there is going to be some uncertainty as compared to if you knew the intended location.

Now, what part of THAT don’t you agree with?

Now, what part of THAT don’t you agree with?

This part:

Much better.

Maybe 80% of the time, identifying the pitch type is enough to know the intended location.  So you really don’t even need a huge sample.

Now, if you get a huge sample, you can get a better understanding, but that’s true with catcher target data, too.

I’d love to have catcher target data, but at this point that’s a pipe dream.  I’ve charted a few games by hand from video, but that’s extremely labor intensive.  So I work with what we have. And it really does do about 80% of the job.

"Maybe 80% of the time, identifying the pitch type is enough to know the intended location.”

I don’t see how that is going to help you all that much.

4 seam fastball.  Where was the intended location?  Unless the pitch was thrown down the middle, you have no idea if the pitcher did or did not hit his intended location!

Let’s say you had 10,000 of such pitches and you found that 1000 were down the middle (likely missed his location on lots of them, although not on most 3-0 counts or 2-0 versus a pitcher, etc.), 4000 were low and away, 1000 were high and away, 2500 were high and inside, and 1500 were high and away, how in the world would you know the command of the pitcher?

Please educate me.  I am willing to learn.

You have a copy of the 2009 Hardball Times Annual, yes?  Read my article.  It explains it.  I don’t feel like typing a whole page of explanation when I’ve already written it before.  If you read it and still have questions, I’ll be happy to answer them.

Unfortunately, my copy is out west and I am back east for the summer. No problem though.  I’ll take your word for it! Sort of.

"Rally:  I was on the 1B side as well, section 133.  Where were you?  We should catch a game some time, and discuss fielding metrics. :>)”

Too bad I didn’t know you’d be there.  We might have even bumped into each other at some point, I was in section 131.  Didn’t know where I’d be until I got to the park for will call.

Fielding metrics would have been the last thing on my mind, I suspect yours as well.  Just the kind of game where for me, all the issues of the saber world fade away, and just appreciate something special.  Besides, who needs fielding when you get that many whiffs?

MGL, send me your email address and I’ll email you copy of the article.

I looked at updated projections for the last 2 lineups Strasburg will face, plus the next one.  I’m using Cleveland’s lineup that kicked the crap out of Boston tonight, since it was against a righty starter.

The Pirate lineup comes out to 3.86 runs per game.  I’m using -65 as the projection for the pitcher/pinch hitter spot.  That is much better than the AAA lineup he faced in Buffalo, 3.45 RPG.  But he’ll face a much better lineup in Cleveland, 4.41 RPG.  He adjusted pretty well in the jump from facing a AAA team to a bad hitting MLB team, it will be a similar jump to face a better MLB lineup.

Most of the difference is because he’ll be facing a designated hitter, the game is in Cleveland.  Replace Travis Hafner with a pitcher/PH -65 projection, and the Indians are only barely better than the Pirates, 3.94.

I thought the Indians had a good offense coming into the season, but that is not the case.  A big part of my preseason projection was based on Grady Sizemore bouncing back from a down year and providing superstar production.  That didn’t happen, so for the rest of the year they don’t look too good.

I also have the PIT and CLE lineups around the same (after adjusting for the DH), with both being around .40 to .45 rpg worse than average.

You say the Pirates lineup is 3.86 rpg. Is that park neutral?  What is the league average?

I was using 720 runs/162 (4.44/g) as average, just a guess.  Looking at MLB so far, teams are scoring 4.47 per game.  Close enough.  And I was using the R/150 from my projections, so it’s park neutral. 

I agree that the lineups are pretty close when you adjust for the DH, but the DH is the big point here.  The next game is in Cleveland, and Strasburg will be facing one.  Having Travis Hafner, even if he’s not what he used to be, instead of Jeff Karstens makes it a tougher challenge.