THE BOOK--Playing The Percentages In Baseball

Monday, May 14, 2012

Sale the starter v Sale the reliever: why can’t he supposedly start?

Kyle speaks up.

Wednesday, May 09, 2012

What does it mean when you age?  It means you can’t throw as hard

Good stuff from Jeff, whereby he controls for the lack of year-to-year change in fastball, and notices that there is no change year-to-year in FIP.

Saturday, May 05, 2012

FIPer

This blogger is certainly correct that I shouldn’t have IP in the denominator.  I’m pretty sure I talked about this a few months ago.  You can just replace IP with PA/4.3, but it’s going to affect the coefficient of the K slightly.

Anyway, while it is very interesting that his version of FIP (that excludes BIP in the denominator) has as good or better correlation with next-year ERA, it really doesn’t make logical sense.  Consider a pitcher like Radke or Rueter that has 80%+ of his PA as BIP, and a pitcher like RJ that has 65% of his PA as BIP.  That knowledge is thrown out the window.  If they both happen to have the same “core” ratio, they get the same score.  So, that part doesn’t make sense. Now, the blogger is saved by the fact that most pitchers don’t have this problem.

Anyway, his basic idea is actually pretty good (and I’ve used a variant of it in the past for other things).  But what you also need to do is incorporate the rate of BIP per PA to get the fuller picture.

But we definitely need a better name!

Friday, May 04, 2012

Sometimes I think I can watch a pitcher for one inning and assess him pretty well…

This does not apply to most pitchers.

I saw Rafael Dolis throw an inning today in the Cubs game. The announcers said they are thinking about using him as a closer (their ace). He has never pitched above AA in the minors I don’t think.

I’ve never seen him before and I know nothing about him.

I am pretty sure he is terrible for one reason and one reason only. He has no idea where his fastball is going, presumably because he cannot repeat his (very simple) delivery.

He threw a 92-95 somewhat sinking fastball only. He had absolutely no command or control of it, and that did not look like a fluke to me (it could have been I suppose). It did not look like he had a lot of “life” or movement on the fastball other than some slight sink. If he has any secondary pitches, I would have to assume that he can rarely use them since he is so often behind in the count or if he does get behind in the count he cannot simply throw the fastball for a strike when he has to.

The absolute, number one thing a successful pitcher has to have, almost bar none, is command on his pitches, at least the fastball. Without that, unless you have a ridiculous fastball in movement, velocity or both, you cannot be a successful pitcher. The reason is obvious but I’ll spell it out. First, you will walk too many batters, second, you will make too many mistakes with location, and third, batters will be able to guess fastball and location because you are behind in the count too often.

Why would the Cubs ever think that this guy is even close to being an ace reliever? Or maybe I am wrong in my assessment. I would not let this guy pitch in the majors yet until if an when he gains some control/command.

Name this player!

He is one of the greatest players in the history of the game. A first ballot hall of famer, easily. If you told almost anyone, including those who play, manage, coach, and broadcast the game of baseball, that if he were to get injured and lose an entire season and you replaced him with a league average player at that position, his team would lose less than a win and a half in expectancy, they would think that you were a complete idiot who needs to go back to your mother’s basement.

Who am I talking about?

Wednesday, May 02, 2012

Selection bias

It’s not any more evident than seeing the aging curves for starting pitchers and relief pitchers.

Also, the unit of measurement should be per PA, not per IP.  Per IP is the same thing as saying per out, and there’s no reason to measure things in that manner.  The opportunity factor is the denominator, and that’s PA or BIP.

Anyway, the most interesting thing was the aging curve on fastball speed.

Friday, April 27, 2012

Are closers overpaid?

After reading this article on SBN:

http://mlb.sbnation.com/2012/4/26/2978326/jordan-walden-heath-bell-sean-marshall

I wrote this in the comments section:

A replacement level short reliever allows around .5 runs per game more than an average pitcher.

An elite closer allows around 1 run better than an average pitcher, or 1.5 runs per game less than a replacement reliever.

A typical elite closer pitches around 70 innings with an average Leverage Index (LI) of 2.0. That means that he pitches 140 “effective” innings.

For 140 innings at 1.5 per 9 innings less than a replacement reliever, that is 23.33 runs better than a replacement reliever, or 2.33 WAR.

In other words, an elite closer is worth around 2.33 WAR per season.

What is that worth in the FA market? Almost 12 million dollars. So, Brian Wilson, Heath Bell, et al. are actually underpaid.

Does that mean that a team cannot acquire a very good reliever who can function as a closer for less than market value, based on his expected (projected) WAR? No, it does not. In fact, it is much easier for a savvy team to acquire and/or use a cheap but good reliever as a closer than it is for a team to do so with any other position.

Still, closers are NOT overpaid as a class, especially the elite ones. In fact, you can make the argument that they are underpaid. Some are of course. But, so are players at any other position.

Now, I have not researched the salaries of the typical or average closer, as well as their average or typical WAR, so I could be wrong. But, it seems to me that that is a recently popular refrain among semi-sabermetric types - that closers are way over-paid, and I think it is wrong. My numbers above could be wrong too, like the typical LI for a closer, and the runs above average for a replacement short reliever. I was guessing at those.

Thursday, April 26, 2012

Is Greinke obsessed with controlling the strike zone?  And does that lead to a high BABIP?

Glenn asks the question of whether Greinke’s (seemingly) singular focus on a low-FIP leads to a high-BABIP.

The first thought I had was that Curt Schilling is probably the pitcher who had the best K/BB ratio in the previous generation.  (I didn’t check.  I’m just guessing that if he wasn’t #1, he was in the top 3.) Schilling also happens to have one of the highest BABIP of his generation.  This is doubly-interesting because there is a modest relationship between a low-BABIP and a high-K rate.  So, Curt really bucks the trend here.  Anyway, as I was reading Glenn’s article about Greinke, I kept thinking “Schilling !”.  Schilling is also friendly-ish with geekdom as well.

There may be something to Glenn’s point.  I said something about it in my last THT Annual article.  That pitchers who have alot of BB and alot of HR end up with a low BABIP because once you remove the “easy” hits (HR), you have a pitcher that batters end up either chasing balls close to the edge of the strike zone, or taking lots of walks.  But a Schilling for example likely pitches closer to the center than an average pitcher, and so has more “hittable” balls.  Not so hittable that they’ll be HR (and therefore are still part of the numerator and denominator of BABIP), and not so unhittable that Curt is pitching alot outside the strike zone, but hittable enough that that’s all a batter can do (make contact and get weak non-HR hits).

Anyway, just an untested theory that’s fun to think about.

Monday, April 23, 2012

Pitcher on Pitcher K

Good stuff from Eric, and I look forward to seeing career leaders in his next installment.

Friday, April 20, 2012

No-decisions

On the heels of Cliff Lee and Matt Cain came Felix Hernandez.  Those were three of the four best-pitched games (according to Bill James’ Game Score) so far this year that resulted in a no-decision.  That is, if you rely heavily on a pitcher’s W/L record, that metric suggests that these three games were.... neutral.  We all know that a pitcher’s W/L record is problematic for any given game, but the reality is that those problems are ignored by people who use the metric.  Rather than actually watch a baseball game, the users of pitcher W/L records rely on the cold numbers as if it means something more than it does.  They set aside the actual game that was pitched, and rely on the W/L record.  And in the cases of these three guys, they are given a no-decision, which is akin to them having not even pitched in that game.

Yes, yes, we know, “I use W/L in conjunction with ERA”.  Well, there’s no reason to.  The pitcher W/L record doesn’t add to what the pitcher actually did.  And yet, the prominence attached to the pitcher W/L record is enormous.  It’s there in every boxscore on TV, it’s there everywhere you look.  Stories will be written about the first pitcher to reach 10 wins and 15 wins and 20 wins this year.  Stories will be written when some pitcher is assigned 5 consecutive wins (without a loss, but could include a no-decisions, because, after all, a no-decision is like the pitcher didn’t even pitch in the game).

Why do we bother?  Because MLB has decided to treat the stat as official, we have the need to report on it, comment on it, and glorify it?  Did George Orwell invent this stat?  The pitcher W/L record should be treated as a sideshow stat, something that may be useful from time to time.

The reality is that for any given game, while the pitcher may be the single most important player in determining the winner and loser of a game, he is still not the entire reason for determining the winner and loser of a game.  He’s not even half the reason. The pitcher W/L record however gives the pitcher the entire credit, even though we know better.  The job of the record keepers is to reflect the story.  If the story is that they pitched great enough to win, but didn’t, then record that.  But, combining their starts with any other run-of-the-mill no-decision is a massive failure on the part of record keepers.

Cain, Lee, and Felix’s starts in the last 48 hours will be given a footnote in the annals of history, while their no-decisions will love on in infamy.  Is this really the best we can do?

Friday, April 13, 2012

Michael Schwimer: beyond PITCHf/x

Just fantastic.  Listen, every single pitcher (and every single batter) needs to do what this guy does, and then give their personal scouting reports to the systems team in the front office.  This is what we need, and what is missing from PITCHf/x and HITf/x and FIELDf/x and RUNf/x.  We need to know intent.  We need to know where you intended to throw a pitch, what you intended to throw, and what the batter was intending to look for and where to look for it.

Unfortunately, there’s a huge conflict of interest here.  A bad pitcher that self-reports will show up in the analysis as a bad pitcher, thereby forcing him back to the minor leagues.  If he gets traded, now we’ve got his personal diary, and he may be much easier to beat.

I remember Carlos Delgado saying he kept a journal of every one of his at bats.  Tim Raines said he kept a journal of every pitcher and how he would steal off them.  All I can say to Schwimer and the other guys like him: please, please, please, after your career is over, donate your journals to us researchers. We live and breathe to get this data.  This is scouting data.  This is how sabermetrics will get convergence between performance analysis and scouting.  As it stands, we’re left to resorting in inferring things like intent, and our uncertainty levels are going to be huge.

And a huge glove-slap to David for conducting the interview and presenting Schwimer’s thoughts.

Wednesday, April 11, 2012

Ace versus elite closer, 4th time through the order…

Of course this one game anecdote means nothing other than it is a perfect illustration of the “times through the order” effect and why you want to avoid having your starter face the lineup for the 4th time, even if said starter is an elite pitcher.

Verlander was pitching a 2-0 shutout in the 9th, and he gets to the top of the lineup, the 4th time through, with a runner on and 1 out. I would think that Valverde or another elite reliever is better than Verlander at this point, even though Verlander is probably a top 5 starter in baseball, maybe top 3.

But, Verlander is throwing a shutout, so surely he is going to continue throwing above his own true talent level, even though he has faced the lineup 3 times already. Not.

He proceeds to give up a single, wild pitch, walk, and a single, is taken out, and the Tigers blow the game…

Are hit batters an inevitability of pitching?

I looked at all pitchers born since 1931, who faced at least 10,000 batters (120 pitchers).  The average number of hit batters per 38 batters faced (roughly 9 innings) was 0.21.

The best pitchers (based on ERA+ at BR.com) were:
Pedro (0.42 HBP/38 batters faced)
Clemens (0.30)
Doc (0.25)
RJ (0.42)
Maddux (0.25)

So, the five best pitchers, perhaps ever, were all above average in hit batters.  The next guys on the list:
Schilling (0.15)
Seaver (0.15)

Then back to normal:
Gibson (0.24)
Hudson (0.36)
Brown (0.39)

Then finally:
Palmer (0.09)

So, for the best pitchers, most had higher than average hit batters, even though they had better than average control.  Pitching inside was part of the strategy, and the inevitability of hit batters.  Except for Jim Palmer, and maybe Schilling/Seaver.

The leaders in lowest HBP/38 batters:
0.04 Cueller
0.06 Vida
0.07 Hooten
0.08 Fernando
0.08 Renko
0.09 Palmer
0.09 McCormick
0.09 Steve Carlton
0.10 Br Hurst
0.11 Marichal

Their ERA+ was 109, which was identical to the 120 pitchers from the sample.  So, it’s definitely possible to maintain a low hit batter rate, and still get success.  Jim Palmer would be the model, with one hit batter per 424 batters faced.  Steve Carlton at 409 and Marichal at 356.  Hitters knew that these pitchers “self-policed” themselves by either pitching in “responsibly”, or not pitching in much.

What would happen if you make it harder for pitchers to pitch inside?  Maybe we’ll get more Mike Cuellers and Bert Hootens, with 2.7 BB, 5.3 SO, and 0.7 HR per 9IP, about 5-10% lower rates than the sample average, while maintaining the same ERA.

So, where is the evidence that enacting rules that lower hit batters will lead to a drastic balance of power in favor of the hitters?  And why is it that pitchers who enacted their own rules (via their behaviour) to limit hit batters still performed quite well? 

Are we suggesting possibly that we have a ton of pitchers who did have low hit batter rates who were drummed out of the league, becauase they wouldn’t pitch inside, and therefore, did not make the study?  That’s possible.  I’d like to see someone else pick up the slack here, and look for pitchers through age, say, 25, with low hit batter rates, and see if they were able to survive beyond that.

I’d like to know why Schilling and Seaver’s 0.15 hit batters per game can’t be the top-end of pitching inside successfully.  Why does it have to be 0.25 - 0.42 that Doc - Pedro get as the requirement?  Why couldn’t the overall average be 0.10 instead of 0.21?

Tuesday, April 10, 2012

Batted Ball FIP

I don’t really talk about bbFIP much, mostly because it’d be easier if Fangraphs and BR.com would track it for us (though I understand they’re overloaded with stats already).  It’s good to see others bringing it up though.

Does experience matter for a closer?

Jay Jaffe attempts to answer that in this BP article:

http://www.baseballprospectus.com/article.php?articleid=16222

Unless I am misunderstanding his methodology, I don’t see how his data are anything but the result of selecting sampling. If you pitch lousy in a save situation for your first few save opportunities, you are less likely to ever get another one than if you pitch well, regardless of your talent or experience.

I wrote this in the comment section:

Jay, correct me if I am wrong, but I assume that the pool of pitchers in each group are not the same. In other words, some of the pitchers who got 1-5 or 1-7 saves did not go on to have any more opportunities and hence they were not in any other groups. Obviously anyone in a larger group was also in a smaller group.

If my assumption is correct, then your results are simply due to selective sampling, no? Some of the pitchers in group 1 pitches badly (blew a good proportion of their save opportunities) and were not allowed to have a save opportunity anymore, right? So, of course, the lower groups will have a lower save percentage. This tells us nothing about their talent withe respect to being able to close out a game. Nothing whatsoever.

So, if you are trying to find some evidence that experience matters, you are not going to, using this methodology. I’d have to say that this statement you made:

“The way many managers, mainstream media types, and even fans discuss the role, it would seem to, and to a limited extent, the data backs that up.”

is false. The last section at least.

In fact, if you make sure that you use the same pitches in each group - in other words, if you worked backwards by only using the pitchers in the last group, you will likely see the opposite result - that these pitchers, as a group, who went on the get at least 22 or 26 save opportunities, were lucky in the beginning. You should find their save percentage in the first 5 or 10 games (group 1) to be higher than later on (the latter groups).

Friday, March 23, 2012

Start v Relief: Max

Max throws his hat in the ring.

In Max’s second table, the one that follows this, he STILL has a selection bias:

To have a fairer look at the issue, I compared things within each pitcher/game combination. For each pitcher, I calculated the runs per PA as above for every game played. Then I compared the R/PA each time through the order in the game to the R/PA for the entire game, thus obtaining a delta R/PA (negative when the pitcher allows fewer runs).

The following table better reflects reality:

Can you spot it?  It’ll help if you try to think of extreme scenarios.

Also note two other things we’ve discovered recently that Max didn’t address:
1. Time of day
2. 9th inning effect

If you start with the idea that the peak time for hitting (and therefore the worse time to pitch) is at 5-6pm, then a starting pitcher pitching at 10pm the third time through the order is getting a huge benefit.  So, you REALLY need to use time of day here to determine if there is an impact, and how much.

As for the 9th inning effect: when it’s a close game in the 9th inning, especially bottom of the 9th, everything changes.  How fielders field, how batters bat, and how pitchers pitch.  The payoff for each event and each action is different enough in the 9th inning that you can’t just make a straight comparison.

Friday, March 16, 2012

No, really, this time, Jeremy Hellickson will be the guy who will buck the BABIP-trend

Every year, it’s the same thing.  Some guy has an obscenely low BABIP.  But that guy has “something” that makes him immune to the laws of BABIP regression.  This year, it’s Jeremy Hellickson.

In the year 2000, Pedro Martinez had an equally obscenely low BABIP of .236.  And he was coming off a year with an ERA of 1.74, 284 strikeouts (!) and only 32 (!!) walks.  Now, if anyone can buck the BABIP regression, it’s THAT GUY.  In 2001, his BABIP was .307.  Somehow in 1999, his BABIP was .323.  Listen, this is ground that Voros covered in the beginning of DIPS.

Here, this is what you can do.  Find me EVERY pitcher over the last 15 years (through 2010) with at least 150 IP, and a BABIP of at most .250.  Then, tell me what the average BABIP of that group of pitchers was in the NEXT year.  And tell me what percentage of those pitchers had a BABIP of under .250.

Now, find me EVERY pitcher over the last 15 years (through 2010) with at least 150 IP, and a FIP of under 2.75 (*).  Tell me what the average BABIP of that group of pitchers was in the next year, as well as what percentage of those pitchers with a BABIP of under .250.

(*) Set this value so that the number of pitchers who meet the FIP threshold matches the number of pitchers who meet the BABIP threshold.

I haven’t run this, but I can guess as to the answer.  But, surprise me.

Wednesday, March 07, 2012

Pitch speed, injury, aging, and changing roles…

I’ve been doing some interesting research on all of the above - actually how they relate. The data I am using is this:

Average seasonal fastball pitch speeds from FG, which uses BIS data, not pitch f/x, I think. DL data, which for this research, was just number of days spend on the DL in a season. For most of this research, I only noted “on DL” or “not on DL” for a particular season. Finally, I used my own NERC (normalized component ERA) data which are ERA looking numbers produced from a BaseRuns formula on the underlying components, including WP I think. They are all adjusted for context - defense (as best as I can, using UZR data), park, opponents, and league. 4.00 is defined as a league average pitcher in both AL and NL and a 4.00 in the NL is equivalent to a 4.00 in the AL since I do league adjustments. Anyway, those details are not that important.

Here is some really interesting data and discussion:

Read More

Tuesday, March 06, 2012

Do pitchers who are hurting allow a higher BABIP?

Conventional sabermetric wisdom is that a pitcher’s BABIP has very little to do with his overall talent.  IOW, if we look at good or great pitchers and then at bad or terrible ones, both groups will have around the same BABIP - what separates good pitchers from bad ones is their HR, BB, and K rates.

What about when a pitcher is hurt and clearly not throwing nearly as effectively as when he is healthy?  Does that affect his BABIP?

I looked at all pitchers from 2004-2011 who spent time on the DL. From observing baseball for many years, my instinct told me that pitchers who go on the DL for non-acute types of injuries typically pitch awful for a while and then it turns out that they were nursing an injury and they go in the DL.  Obviously this is not always the case and it depends on the injury.

I did not distinguish between types of injuries or length of stay on the DL.  A pitcher in my sample could have gone on the DL for a strained hamstring running to first, a broken hand smashing a water cooler, or Tommy John surgery.

I looked at performance in the 14 days prior to their “effective” DL date. I am not exactly sure the difference between “effective” and “official” DL date. Maybe someone like Jeff Z. can explain it.

Anyway, these pitchers did indeed pitch awfully in that time period - around 1.5 runs per 9 higher than their “normal” performance. You may be surprised at how bad they pitch - I was. You would think that a team should be able to figure out sooner that a pitcher is hurt!

But, what about their BABIP?  Take a second to think about what your answer might be…

...
...
...

OK, here it is. Their BABIP is substantially higher than “normal” during this time period. While they walk around 6% more batters (per PA), K 1.2% fewer batters, and allow 15% more HR, their BABIP is .324.

This is further evidence that it is not that pitchers, in general, have almost no control over the ball when it leaves the bat and stays in the park. It is that MLB pitchers are specifically and partially chosen on the basis of not being hit hard by major league batters. If you get hit hard by major league batters, you don’t pitch in the major leagues.

BTW, those percentage numbers above are “as compared to this pool of pitchers’ seasonal stats in the same year,” and those seasonal stats include periods of time when these pitchers are pitching while hurt, as well as when they come back from the DL, assuming that they do come back.

Also, when pitchers come back from the DL and have pitched for at least a month, what is their BABIP after that?  .288.

Sunday, February 26, 2012

Should ERA+ and ERA- adjust for the role of the pitcher?

ERA+ is League ERA / pitcher ERA.
ERA- is pitcher ERA / League ERA.

Now, you would think “league ERA” is easy enough.  But, it may or may not be based on AL/NL (as opposed to MLB).  It may or may not be based on the pitcher’s home park.  It may or may not be based on how often a pitcher pitched at each park (say, if Latos pitched alot more at Petco than the standard 50%). 

But, let’s go further.  Can we adjust for the different quality of competition between AL and NL?  Or, do we want 100% to be the average for the NL and AL pitcher?  And, someone could or could not also adjust for the identity of their hitters, and heck, their fielders.

But, let’s go even further that that.  Because all of the above is nothing compared to the adjustment for the role of the pitcher. 

Suppose the league average for all pitchers is an ERA of 4.00.  If you take a perfectly average pitcher, and put him in relief, his ERA will be around 3.40.  You put that same pitcher as a starter, and his ERA is 4.30.  (Roughly, more or less, give or take.)

As a result, the “role adjustment” is far more significant than adjusting for year-league, quality of league, quality of competition, or park.

Roy Halladay had a 2.35 ERA while his mate Ryan Madson had an ERA of 2.37.  Their ERA+ both checked in at 164, and ERA- of 61 and 62 respectively.

Now, are we happy with that?  Or, would we like to adjust for their role as well?  Which in this case, would mean Halladay goes to 176, while Madson goes to 135 for ERA+.  For ERA-, it would be 57 and 73.

To me, it’s clear that we want to do the role adjustment.  Why waste all our efforts doing all the little adjustments, the park, the league, even the quality of league, and then ignore the one thing that has the most impact?

And, if you are on the side that says to not adjust for role, but are happy with the park-et-al adjustment, then tell me the reason.  Explain to me why you would adjust for Petco and Safeco, even though you have NO IDEA how Latos and Felix were exactly, or reasonably, affected (after all, the park adjustments are global-generic).  And at the same time, you don’t want to adjust for the role?  Because I believe that your argument here will be a political one: you will present the half of the argument that makes park factors preferable, and the half of the arguments that make the role factors not preferable.

The use of ERA+ and ERA- only exists if you make the adjustments at some point.  You would never consider ERA+ of Halladay and Madson equals, and you would adjust anyway.  So, why not just adjust them first, like park et al?

Go.