THE BOOK--Playing The Percentages In Baseball

Tuesday, October 18, 2011

Times through the order with the 9th inning removed…

In light of the new research presented on this blog which suggests that when starters pitch the 9th, the score tends to be lopsided in favor of the pitching team and wOBA tends to be lower than expected given the true talent of the pitchers and batters (and other things that affect offense), I have recalculated the “times through the order” wOBA for both day and night games, with indoor games not in the sample, removing all 9th inning data.

In The Book, this is what we presented:

Times through the order expected actual

1 .353 .345
2 .353 .354
3 .354 .362
4 .353 .354

As you can see, the more a starting pitcher faces the lineup the better those batters do, due to familiarity or pitcher tiring, or both (or some other reason or reasons).  However, the 4th time through the order, the trend seems to stop and batters actually perform the same as the second time through the order.  This seems to make no sense.

We have speculated two things that might be causing the 4th time through the order depression:  One, 2/3 of all games are at night and it is colder the 4th time through the order.  Two, and more recently, the 4th time through the order sometimes happens in the 9th inning, and as we have just found, 9th inning wOBA versus starters gets depressed because the score is usually lopsided in favor of the pitching team.

So I reran the numbers separately for day and night games (and ignored indoor games), and I also ignored the 9th innings.  The wOBA is adjusted for the pool of pitchers and batters in each bucket.  The first row is 1st time through the order in the 1st inning only.  The second row is 1st time through the order in all other innings.  We see a real depression in the first inning.  Although the data is for home and road teams combined, it is actually the road team batting that is heavily depressed in the first inning for some reason. Either home team pitchers are already used to the mound, the road batting team starts out “cold” or there is some other reason or reasons.

Night games

1 (1st inn) .339
1 (other inn) .341
2 .352
3 .359
4 .350

So, again, we see a depression the 4th time even though we are not using 9th inning data.

Day games

1 (1st inn) .330
1 (other inn) .341
2 .349
3 .357
4 .367

Here we see a large jump from the 3th to the 4th time.  It does appear that either temperature or pitcher tiring during the day (but not so much at night), or perhaps shadow issues during the day, greatly affect the “time through the order” penalty…

What The Book has said about starting pitchers

While I encourage everyone to read The Book, via their local library, Amazon’s free Look Inside, buy-and-return, or outright buying The Book (if you click the link in the top left, we’ll make about 2$ profit), I know there’s still a significant group of people that still won’t do that!  Why? I don’t know.  In any case, I’m going to clip the “The Book Says” box from Chapter 7, which deals with Starting Pitchers.  Note that when we wrote The Book (published Feb 2006), the data available was less.  The data from this chapter comes from 1999-2002.  Ideally, I’d redo this chapter using 1993-2011 data, plus all the other stuff we’ve learned since then. 

Until then, here are the conclusions of each mini-study.  If you want to know the reason for these conclusions, then read the relevant part of The Book.

The Book Says:
As the game goes on, the hitter has a progressively greater advantage over the starting pitcher.

The Book Says:
Don’t bank on some magical pitch count level, such as 100 pitches, to tell you when a pitcher’s tank level is low.

The Book Says:
Pitchers perform best with five days of rest, and worst with three days of rest. To manage our entire starting rotation effectively, four days of rest seems to be the optimal point. The current MLB pattern of scheduling the starting rotation works.

The Book Says:
You can’t tell if a pitcher is on based solely on the results of the first nine batters he faces.

The Book Says:
You can indeed tell if a pitcher is off based solely on the results of the first nine batters he faces. Such pitchers will perform somewhat worse than their true talent levels the rest of the way.

The Book Says:
If a pitcher is getting hammered early, there is a huge carryover effect for an inexperienced pitcher. For an experienced pitcher, there may be some evidence of a carryover effect.

The Book Says:
We can’t rely on the poor results of late-game performance to establish the fatigue level of a starter. Observation of a pitcher’s mechanics would be the preferred method.

The Book Says:
If a pitcher is still getting outs late in the game, keep him in there; he may have a bit more left in the tank.

The Book Says:
Starters have a lot going against them, as compared to relievers. They pitch a lot longer, forcing them to pace themselves. They face the same batters multiple times in a game, giving the batters an advantage. Relievers can come in and put all their efforts into a very short stint.

The Book Says:
Selective sampling has a profound effect on interpreting the performance numbers of pitchers as starters and relievers. Be careful with any interpretation.

Summarizing late inning starter and reliever performance…

I realize that lots of people have chimed in or lurked a little with regard to the data presented about starters and relievers pitching in the later innings, especially the 9th, and that the data can be confusing.  In an effort to summarize an analysis of the data, here is what it suggests:

If managers and coaches can tell whether a starting pitcher is “on” in any given day based on how he has pitched in the early and middle innings, we should see gradually increasing performance in the later innings, relative to a starter’s overall performance.  IOW, if we look at all starters who pitch the 7th, we should see a collective excellent performance in 1-6 (otherwise they would not be allowed to continue).  We do.  We see a wOBA against of .322 (23 points better than their seasonal wOBA) in innings 1-6 (and presumably a low pitch count and few runs allowed) when they are allowed to start the 7th.  Presumably their managers and coaches are thinking or saying, “He is pitching well, and his stuff is good today.  We’ll let him start the 7th - at least.”

Since a pitcher’s overall wOBA against is a combination of days when his stuff is good and his stuff is bad, we should see better than average performance in inning 7.  We see, however, a wOBA against of .354 while these pitchers’ overall wOBA is .345, which is almost exactly what we would expect the 3rd time through the order.

So it seems that managers and coaches are not able to tell that their stuff is good that day and they will continue to pitch particularly well, since they don’t.  In fact, if they had let these pitchers pitch the 7th when they were pitching horribly in 1-6, they would have had to pitch the same in the 7th as they did when they were pitching exceptionally well in 1-6, since the two samples have to average to their overall seasonal numbers.  IOW, if they are overall .345 and they pitch at .354 the third time through the order in the 7th, when they have been pitching great in 1-6, they have to also pitch at .354 when they are pitching badly in 1-6, since we expect .354 overall!

For inning 8, we should see the same phenomenon but even stronger.  Since pitch counts are higher and short relievers are now readily available, we should see a manager only let his starter start the 8th when he has pitched even better in 1-7 and the manager believes that he is really “on” that day.  He has to be even more “on” than in the 7th inning since his pitch count is higher and short relievers can now come in to pitch.  In fact, going into the 8th inning, these pitchers who are allowed to pitch the 8th have pitched at a .297 level in innings 1-7!  Surely they are “on”!  So we expect his 8th inning performance to be really better than their overall performance.  While it is true that these guys who are allowed to pitch the 8th have been even better in 1-7 than pitchers who were allowed to pitch the 7th, it turns out that in inning 8, these guys once again, like the 7th, pitch at their expected seasonal levels.  They were at .345 in the 8th and they were .343 pitchers overall.  Again, given that they were pitching the 3rd or 4th time through the order, that is about what we would expect, if there were no predictive value associated with their prior innings.  Remember these numbers, like the .345, are adjusted for the pool of batters in that inning.

Once again, managers simply cannot tell whether their starters are “on” or not, or perhaps there is no such thing as being “on.”

What about inning 9?  Again, we should see the same phenomenon, but even stronger than inning 7 and 8.  Pitchers who are allowed to pitch the 9th pitched at .283 in 1-8.  This time we do see better than expected pitching in the 9th!  Starters who pitch the 9th not only show exceptional performance in 1-8, but they continue to some extent that exceptional performance in the 9th.  They are .330 in the 9th even though they are .342 pitchers overall.  They do around 13 points better than expected (the 4th time through the order, pitchers typically do 1 point worse than overall).  How can they all of a sudden do that in the 9th but not at all in the 8th or the 7th?  Surely managers just don’t let any starter pitch the 7th and 8th yet all of a sudden they decide that only starters who are “on” that day will pitch the 9th?  That makes no sense!

So what is the explanation?  It is simple once we look at how wOBA is recorded in different score situations (whether the batting team is tied, up by a little, up by a lot, down by a little, or down by a lot).  As it turns out, even if the actual quality of the pitching (and hitting) is the same, the wOBA can change radically because of the approach of the batters, pitchers, and fielders, depending on the score in the 9th (or later) inning only and because the wOBA weights are based on average values (across all innings) of the various events.  In fact, we see that for all pitchers in the 9th, including relievers, who actually pitch an overwhelming majority of 9th innings of course, wOBA is much lower when the pitching team is ahead by 2 or more runs, and much higher when the game is tied or the pitching team is ahead by only one run or is losing.  This is evident from looking only at the relievers, who aren’t pitching the 9th inning because they are “on” that day.  They are pitching the 9th because they are primarily closers or late inning relievers in general.  And please remember, when I say that wOBA is lower in games where the pitching team is ahead by a lot and higher when the game is close or they are losing, I mean relative to the pool of batters and the quality of the pitchers in that “bucket.”

So, for example, if in blowouts, the average pitcher (reliever and starter) is a .350 pitcher and the batters are average, then we might see a wOBA of .330 (20 points lower than expected).  If in close games, the average pitcher is a .320 (again, with average batters), we might see a wOBA of .340, 20 points higher than expected.

Here are the numbers in the 9th inning for all pitchers based on score differential of the pitching team.  I’ll only use score differentials when the pitching team is winning so we can use home and road numbers (the numbers for when the pitching team is losing are similar to when the game is close).  Look at the pattern.  It is obvious.  The first number is the wOBA adjusted for the batting pool. The numbers in parentheses are the seasonal talent levels of the pool of pitchers in that bucket.  Remember these numbers are for all pitchers, which are mostly relievers of course, since 90% or so of all PA in the 9th inning are pitched by relievers.

Up 4 or more runs: .325 (.337)
Up 3: .305 (.318)
Up 2: .304 (.318)
Up 1: .318 (.317)
Tied: .356 (.330)

When up by 2 or more runs, wOBA for all pitchers is around 13 points lower than “expected” (seasonal numbers).
When up by 1, it is 1 point higher.
When tied, it is 26 points higher!

So that still doesn’t explain why all of a sudden in the 9th inning, we see starters doing exceptionally well.  It does if you know this one important fact:  Most of the time that a starter pitches in the 9th, he is pitching with a large lead.  In fact, 76% of the time that a starter pitches in the 9th his team is ahead by 2 runs or more.  55% of the time, his team is ahead by 4 or more runs.  Only 9% of the time that a starter pitches in the 9th is the game tied, which is when wOBA is the highest by far, for all pitchers.

Contrast that to relievers.  Only 48% of the time do they pitch with a 2 or more run lead, 27% of the time their team has a 4 or more run lead, and 13% of the time they pitch when the game is tied.  They also pitch considerably more often than starters when their team is losing, The wOBA is also quite high in games (in the 9th of course) in which the pitching team is losing.

And that is why starters seems to pitch so well in the 9th. It is not that they are really pitching well. It is just that the way wOBA is figured it understates what is really happening (the weights of the events are not correct, and other things probably occur more often, like sac bunts and IBBs) in the 9th inning of games in which starters tend to pitch (close games, not losing) presumably because of the approach of the batters, fielders, and pitchers.  Again, we know that the low wOBA against for starters in the 9th has nothing to do with the starting pitchers themselves because we see that when the pitching team is ahead by a lot with a reliever in the game, the wOBA is just as low.  Again, it is just that starters tend to pitch in the 9th when they are leading by 2 or more and relievers tend to pitch the 9th when their teams are losing or the game is close.

In fact, when we look at games in which the score is close, the starters’ wOBA against is around the same as the relievers (relative to their overall wOBA against) and when the pitching team is ahead by a lot, the starters’ wOBA against is also the same as the relievers’.  No difference.  Starters and relievers pitch the same in the 9th relative to their true talent.  Starters do not pitch well despite having pitched exceptionally well in innings 1-8.

A simple but perfectly apt analogy would be this:

Let’s say that during the day wOBA was 10 points higher than at night, given the same pool of batters.  And let’s say that pitchers with blond hair pitched mostly at night and pitchers with brown or black hair pitched mostly during the day.  What would we find?  We would find that blond haired pitchers appeared to pitch a lot better (by almost 10 points) than dark haired pitchers.  This would be an illusion.  We would find of course that blonds during the day pitched the same as brunettes during the day and that both groups also pitched the same at night.

In this case, starters are blondes and relievers are brunettes and night games are games in which the pitching team is ahead by a lot and day games are games in which the score is close or the pitching team is losing.

So, there does NOT appear to be any predictive value to pitchers who have pitched great in 1-6 or 1-7 or 1-8 AND their manager considering them to be “on” that day (I presume) and thus leaving them in the game.  We only saw that in the 9th inning anyway, and it was an illusion created by the fact that starters tend to pitch when they are ahead by a lot and ANY pitcher will have a low wOBA against, presumably because of a non-typical approach by batters, pitchers, and fielders, and relievers tend to pitch in closer games (or losing games) when ANY pitcher will have a high wOBA against, again, presumably because of a non-typical approach by batters, pitchers, and fielders.

I hope I have explained this well…

Monday, October 17, 2011

Starters and Relievers in the 9th Inning and Score Differential

There has been much discussion (and research) on the relative performances of starters and relievers in the 9th inning.  Preliminary results suggest that starters perform much better in the 9th inning, relative to their overall performance, than do relievers, even considering that they are facing the opposing lineup for the 4th time through the order.  The speculation is that these starters are having “on” days, their managers and coaches can recognize this, which is at least one reason why they are allowed to pitch (at least start) the 9th, and that their “onness” continued into the 9th inning.  (It is true that starters who are allowed to pitch the 9th inning – and 8th, 7th, etc. – have pitched exceptionally well in prior innings.)

While this is a reasonable assumption and certainly comports with conventional thinking, it is somewhat out of step with what we found in The Book – that early in a game success and failure does not have much predictive value.

I don’t necessarily have any particularly strong reason to disagree with this (new found – at least on my part) conclusion (that starters who are having exceptional games up to the 9th inning will continue to pitch at an above-average – for them - level), however, there are two things that I am uncomfortable with:  One, the small sample sizes of starting pitchers in the 9th – in my research I only looked at 2007-2010, and two, the fact that starters and relievers likely have different distributions of score differentials at home and on the road, and that these score differentials alone may considerably impact opponent wOBA (because their approaches may change and thus the wOBA weights are not correct).  This was brought up by Guy in one of the threads on this subject.

I was also troubled by one more thing:  When I looked at the 07-10 data for all starters and not just elite ones, so that I had a much large sample size, they did not pitch very well in the 8th, but pitched exceptionally well in the 9th.  You would expect that if starting pitchers who were “on” continued to be “on,” we would see this effect in the 8th (and 7th) as well as the 9th inning, although perhaps not to the same degree.  In fact, we should see a gradually increasing effect in every inning.  After all, a starter pitching the 8th, on the average, has pitched exceptionally well through the 7th inning, and presumably the manager and coaches allowed him to pitch the 8th not only because he has pitched well thus far, but because they deem that he is ”on” (and not just lucky).

So, one, I expanded my research to examine data from 1993 to 2010 (18 years), and two, I looked in more detail at how wOBA differs for relievers and starters in the 9th inning as a function of the score differential.

I included all relievers and starters in the analysis since I think that whatever happens in the 9th inning for all starters and relievers will also happen with elite starters and relievers (relative to their overall talent of course), and of course most relievers in the 9th ARE elite relievers and even starters in the 9th will tend to be the better starters.

Here is some of the data:

Read More

Tuesday, October 11, 2011

Starting Pitcher on a great day v Closer

Max Marchi weighs in with a great article.  Read it all.  This is the conclusion:

I looked at games played in the past 20 years, thanks to the invaluable Retrosheet data. I selected all the instances in which the starting pitcher has completed eight innings giving up one run at most. These should be the circumstances when the manager can believe his starter “has it” and can complete the game.

I removed the games in which the offense had provided the pitcher more than three runs. Thus, we are dealing with situations in which the game is still on the line, and the manager should be trying to maximize his chances. (In a blowout the skipper’s choices could be dictated by having to rest the bullpen or wanting to try a young arm.)

The games were then split in two groups: Games with the starter beginning the ninth (STARTER) and games with a reliever beginning the ninth (CLOSER).

Here’s how the two groups fared, with more than 1,000 games represented in each group.

runs        percentage
allowed      CLOSER  STARTER
  0          76     74
  1          14     16
  2            7       5
  3            2       3
  4+            0       1

Looking at the numbers above, the decision on whether leaving the starter in or removing him appears as a coin flip. However, the above table can suffer from selection bias, with three possible sources of bias coming to my mind.

Monday, October 10, 2011

Goodbye to you, WPA for starting pitchers

In a game for the ages, Roy Halladay and Chris Carpenter did what they did best.

In a poll on my site, asking for which pitcher threw the better game, nearly 50% called it a draw.  20% went with Carpenter and 30% went with Doc.  Basically, it was 52/48 for Doc, with plenty of uncertainty.  For all intents and purposes, it looked like a draw to the readers.

I introduced four Game Score metrics modelled on the Bill James original, and it gave a slight advantage to Carp over Doc.  If you had to give it odds, it was probably 55/45 in favor of Carpenter.  (Game Score like innings pitched, and Carpenter went the whole nine.)

In terms of trying to have an overall evaluation of their performances, you are left to conclude that it was pretty close.

Except for WPA. 

WPA called it +0.77 wins for Carpenter and +0.24 wins for Halladay if you were to follow Fangraphs, and Baseball Reference says +0.81 for Carpenter and +0.28 for Halladay.  Either way, the gap is 0.53 wins.  A gap of 0.53 wins is equivalent to 5.3 runs in a typical game!  (Of course the higher the leverage, the lower the number of runs it takes to turn into a win.)

So, what happened?  Why did WPA simply blow it here?

The first thing is we have to understand what is WPA is doing.  As it pertains to pitchers, and as Fangraphs and Baseball-Reference calculate it, the pitcher gets 100% of the credit, and the fielders are ignored.  This is not because these two sites believe this as a philosophy.  It is simply a data limitation: if no one is telling them how much impact a fielder actually had on each play, then we simply ignore the fielders altogether.  It’s not the best solution, but it is the same solution as any other metric used.  If you look at a pitcher’s OBP allowed or SLG allowed, the impact of the fielders are ignored, and the pitcher is credited with all of it.

Basically, WPA is simply based on a pitcher’s runs allowed, and his innings pitched.  (We’ll ignore discussion of partial innings for this article.)

Ok, so if WPA is based on a pitcher’s runs allowed and innings pitched, then how does one guy giving up 1 run in 8 innings and another guy giving up 0 runs in 9 innings cause a 0.53 difference in wins?

WPA has a second component: timing.  Let’s take a look at how the game progressed, which started with Halladay giving up an early run, and then, it was zeroes the rest of the way:

What we see right off the bat is that the Phillies chances of winning goes down immediately from the starting point of 50/50.  And as the game progresses to its eventual conclusion of 100% win for the Cardinals, then each out that Carpenter was getting had a bit more win value than Halladay’s outs.

Basically, Carpenter’s outs were feeding off each other, and he’s rolling down the hill toward a win.  Halladay’s outs were simply trying to stem the tide as much as possible.  It’s a case of Halladay setting Carpenter one step back, and then Carpenter getting two steps forward.  Time was running out on Halladay.  And it’s all because of that initial run allowed.

Another way to consider it is this way: imagine that each pitcher was getting a shutout through 8 innings.  Each one step forward was matched by the other guy, such that at the top of each inning, the chance of winning was always 50/50.  For every +0.05 inning for one pitcher, it was exactly matched back with his opponent’s +0.05.

And then you get into the 9th inning.  You are in the 9th inning, it’s still a 50/50 game, both pitchers have earned say +0.52 wins to that point.  Doc gives up a run in the top of the 9th, sending the Phillies chance of winning down to 20%.  That single run costs Doc 0.30 wins, taking him from +0.52 heading into the 8th, down to +0.22 heading into the bottom of the 9th.  Carpenter, with an 80% chance of winning shuts down the side, and he gains +0.20 wins for the 9th inning.  That shutout inning takes him from +0.52 at the start of the inning to +0.72 to end the game.

So, the final tally, as far as WPA is concerned, becomes:
+0.72 Carpenter
+0.22 Halladay

That difference is 0.50 wins.

In this respect, it’s exactly like a pitcher’s Won/Loss record: Carpenter gets the win, and Halladay gets the loss:
1-0 Carpenter
0-1 Halladay

The difference here is 1 win exactly, but that’s only because we don’t consider the rest of the team at all.  If we give Carp and Doc half-credit for their W/L record, then the gap between the two, using a pitcher’s Won-Loss record is 0.50 wins.  And that’s just like WPA.

WPA for starting pitchers simply hides in alot of number crunching what a pitcher’s Won-Loss record already does cleanly and obviously.

Now, does this mean that WPA for starting pitchers is redundant?  Not totally.  While the gap between the above two pitchers was 0.50 wins using WPA, you can see that they both ended up with a positive number.  Had the pitchers given up 1 run per inning, the gap between the two pitchers would STILL have remained at 0.50 wins, except now, each pitcher would have posted negative WPA.  In the case of the starting pitchers each allowing 8 runs through 8, they would have been standing at -0.85 wins each.  Even the winning pitcher would end up in the negative.

So, WPA for a starting pitcher is more like a combination of W/L record AND ERA.  The losing pitcher of a 1-0 game will end up with a higher WPA than the winning pitcher of a 9-8 game.  Which, one would think, sounds right.

The main problem is that it treats both metrics, the W/L record and the ERA, with similar value.  One metric shows the gap at a full win.  The other metric shows the gap as extremely close (1 run is 0.10 wins).  WPA takes the middle position, and gives the gap between the pitchers at 0.50 wins.

Nonetheless, it’s not very appealing to see a 1 run game resulting in a 0.50 win difference. But, that’s at the very essence of any game, that you have a winner and a loser, that you end up with a binary outcome.  It’s not like the game ended with 0.52 wins for Philly and 0.48 for the Cards, or that it ended 0.55 wins for the Cards and 0.45 wins for the Phillies.

WPA is a story stat, as studes calls it.  But, I don’t think WPA for a starting pitcher really adds much to the story.  I think it describes the Cards/Philly as a game perfectly.  But in terms of including the identity of the starting pitcher in that story?  I think WPA fails to tell that story.

Since WPA for a starting pitcher is very reliant on the context of whether the team happens to be winning or losing, it will overshadow the performance of the pitcher to some degree.  While it’s not as bad an indicator as a pitcher’s W/L record, neither is it better than a pitcher’s ERA.

And so, I say, goodbye to you, WPA for starting pitchers.

(Note that WPA for a relief pitcher is another matter entirely.  As untelling WPA for a starting pitcher is as a metric, it’s quite the opposite for a relief pitcher.)

Game Score however tells a much better picture. I said I had 4 versions, each one looking at a particular subset of a pitcher’s performance:
Game Score 1: Runs Allowed
40 + (6.4 * IP) – (10 * R)

81 Doc
98 Carp

Game Score 2: K/BB
40 + (0.4 * IP) + 3 * (SO – BB)

61 Doc
53 Carp

Game Score 3: FIP
40 + (2.5 * IP) + FIPcore

71 Doc
69 Carp

Game Score 4: Linear Weights
40 + (8.4 * IP) + LWTScore

74 Doc
101 Carp

So, depending how you look at their performance, we can see different stories being told here.  The runs allowed story favors Carpenter, but when you exclude balls in play, the story slightly favors Halladay.  Given the lack of consensus by the fans who watched the game, that kind of story seems much more accurate than anything WPA can tell us.

And I say, Hello Game Scores!

Sunday, October 09, 2011

How do starters perform by inning?

First thing I did was look at elite relievers.  I defined an elite reliever as the top 20 relievers or so according to my projections going into year X.  Then in year X, I looked at how they did in the 7th, 8th, and 9th innings (in all games), and what the batter pool was in each of those innings (based on those batters’ overall stats for that year), as well as the platoon situation for that batter pool in each of those innings buckets.

Here is the data:

Elite relievers

6th inning or earlier

Performance in 6th or earlier, .291 wOBA against.
Batters in this bucket were .336 batters overall.
53% batted from the same side as the pitcher.
The pitchers in this bucket were .307 pitchers (wOBA against) for the year.

7th inning

Performance in 7th, .315 wOBA against.
Batters in this bucket were .338 batters overall.
56% batted from the same side as the pitcher.
The pitchers in this bucket were .289 pitchers (wOBA against) for the year.

8th inning

Performance in 8th, .287 wOBA against.
Batters in this bucket were .339 batters overall.
51% batted from the same side as the pitcher.
The pitchers in this bucket were .289 pitchers (wOBA against) for the year.

9th inning

Performance in 9th, .289 wOBA against.
Batters in this bucket were .338 batters overall.
47% batted from the same side as the pitcher.
The pitchers in this bucket were .283 pitchers (wOBA against) for the year.

It looks like against elite relievers (presumably the first and only time through the order - because they are generally short relievers, although I did not track times through the order), we expect a normal pool of batters in the 8th and 9th innings, with slightly more favorable platoon situations for the batting team (presumably because of pinch hitters), hitting around .288 in wOBA, which is somewhere in the neighborhood of these relievers’ overall seasonal stats.

Here is the list of the (elite) relievers going into 2010:

Mike Adams
Betancourt
Broxton
F. Cordero
Devine
Francisco
Hoffman
Kuo
Nathan
Papelbon
Rhodes
Rivera
K-Rod
Saito
Soria
Soriano
Street
Thornton
Valverde
Wagner
B. Wilson

Now what about elite starters pitching in the 9th, presumably while facing the lineup for the 3rd or 4th time?  I used the top 10 or so starting pitchers for each year, 07-10, according to my projections.

First we’ll just look at innings, like I did above with the relievers:

6th inning or earlier

Performance in 6th or earlier, .313 wOBA against.
Batters in this bucket were .340 batters overall.
39% batted from the same side as the pitcher.
The pitchers in this bucket were .315 pitchers (wOBA against) for the year.

7th inning

Performance in 7th, .293 wOBA against.
Batters in this bucket were .332 batters overall.
39% batted from the same side as the pitcher.
The pitchers in this bucket were .313 pitchers (wOBA against) for the year.

8th inning

Performance in 8th, .305 wOBA against.
Batters in this bucket were .335 batters overall.
37% batted from the same side as the pitcher.
The pitchers in this bucket were .312 pitchers (wOBA against) for the year.

9th inning

Performance in 9th, .267 wOBA against.
Batters in this bucket were .338 batters overall.
41% batted from the same side as the pitcher.
The pitchers in this bucket were .308 pitchers (wOBA against) for the year.

Looks to me like the pool of batters in the 9th inning that a starter faces is roughly an average batter, so there probably is not a bias toward the top or bottom of the order.  The pool of pitchers that pitch in the 9th is obviously better than the overall pool of elite starters by around 7 points in wOBA against.  The starters face slightly more same side batters in the 9th than earlier innings, despite there likely being some pinch hitters in the 9th, suggesting that starters tend to pitch the 9th when there are same side batters coming to the plate, but only slightly so.

Somehow though the performance of these pitchers in the 9th is almost 40 points better than expected (.267) given the overall talent of the pitchers (.308) and the quite typical talent of the batters (.338).  This is despite it being the 3rd or 4th time through the order.

I also looked at average starting game time temp in each bucket as well as the average park factor.  All buckets were around the same, so it is not like our starters who pitches the 9th tended to do so in cold weather and in pitchers parks.  In fact, the average game time temp in bucket 4, the 9th inning, was one degree warmer than in the other buckets for some reason (including indoor games).

So unless I am missing something, it looks to me like there is something special about a starter pitching the 9th, at least an elite starter.  It is looking more and more like their stuff is “on” that day and/or their pitch count is low, or something else that the manager and pitching coach can identify.

That may even be the case in the 8th inning. The starters’ performance in the 8th is around what we would expect based on their overall talent, so we don’t seem to see the 3rd (or 4th) time through the order penalty that we expect.

It may be that the “times through the order” penalty may only be applicable when a pitcher is having a bad game and those 2nd, 3rd and 4th times through the order occur early in a game.

Interesting stuff and it appears that I may have been wrong for a long time about taking out a starter who has pitched well and bringing in a reliever.

I would like to look at all starters or non-elite starters to see if the same or a similar thing occurs.

I would also like to look at 3rd and 4th times through the order in early innings (suggesting a bad game for the pitcher) compared to the late innings (suggesting an average or good game)…

How do pitchers do each time through the order?

I can’t believe I’m doing this at 2:30AM, but, there you have.  All data from 1993-2010.

I looked at all starting pitchers, minimum 200 PA the 4th time through the order at night.  Here’s their simple average wOBA each time through the order (meaning each pitcher equally weighted, and in each pool):
.320 1st
.331 2nd
.341 3rd
.341 4th

That was for 105 starting pitchers.

I repeated for day time (min 100 PA 4th time thru).
.317 1st
.327 2nd
.340 3rd
.344 4th

83 pitchers

Saturday, October 08, 2011

How do great starting pitchers pitch the 4th time through the order?

I have my data setup for 1993-2010, so that’s what I’ll be using.

I looked at all starting pitchers, and came up with their wOBA, at night, for the first three times through the order, min 2000 PA.

Here are the top 15:
linct001
martp001
maddg002
johnr005
smolj001
zambc001
santj003
hamec001
peavj001
clemr001
schic002
hardr001
browk001
hallr001
cainm001

All the names should be recognizable.

The totalled 77,530 PA for the first three times through the order at night.

Their average wOBA was .293.

Now, how did they do the 4th and later time through the order?

They had 5,676 PA (which is 7% of all their night time PA).

Their wOBA?  .313.

That’s ridiculously good, especially since the overall league average for all starting pitchers was a bit over .340.  The league average for all relief pitchers was .330.

Motte (through 2010) had a nighttime wOBA of .317 in relief.

Among the 199 relief pitchers with at least 1000 PA at night, their wOBA was also .317.

So, if the choice is between Carpenter or a pretty good reliever, it’s probably a wash. 

Ryan Madson has a .301 career wOBA through 2010, so, his quality level would be the slightly preferred choice.  Motte had a great 2011, and if you add it to his career through 2010, he’s probably going to be close to Madson’s career wOBA.

My call based on this evidence?  A slight preference to bring in Motte over Carpenter.  But, there’s definitely enough uncertainty there that allowing Carpenter to pitch in the 9th is a very reasonable choice.

The tougher choice for Larussa though was letting Carpenter bat in the 8th inning in the first place.  He played with fire there, but he did not get burnt.

***

It’s clear that quoting 9th inning runs allowed was terribly deceiving.  I don’t blame the poster for doing that, because we’ve all been ensnared by the 9th inning partial-inning rules of baseball.

UPDATE: this data has been corrected later on in the comments section.

As an example, here is the runs scored per 27 outs of each half-inning, since 1993:

So, from innings 1 through 8, the home team scores 0.44 more runs.  But in the 9th, they score 0.13 LESS.  That’s a 0.57 RA9 swing.  Why is that?  Because of partial innings.  Once the home team wins, the bases get cleared.  Those runners on base have the effect of being considered as “out on base”.

This is why you can’t look at ERA in the 9th inning.  But, you can look at component numbers, like wOBA or OPS.

Tuesday, October 04, 2011

Rick Porcello and his “sinker”

I have watched Porcello off and on for 3 years now.  He is a mediocre pitcher who was touted as a very good prospect - a first round draft pick in 2007.

I don’t think I have ever seen a sinker ball pitcher throw so many sinkers up in the zone.  If you are a Tiger’s fan, that must be very frustrating.  He seems to have decent command with a nice, smooth delivery, unlike someone like, say, Fausto Carmona.  There really shouldn’t be any reason why he can’t throw that pitch consistently down in the zone. In fact, he doesn’t walk that many batters.  He should be throwing so far down in the zone that he ends up missing low more often such that he walks a few more batters.  It almost seems to me that he wants to strike batters out and that he overthrows.

I have never seen any of his pitch f/x data, but I would bet that he throws his sinker too far up in the zone, on the average, compared to other sinker ball pitchers.  On top of that, I don’t think that he has THAT much of a sinker, so he really needs to pitch down in the zone.  Again, I have never seen any “spin” pitch f/x data, but I would also bet that his sinker does not sink as much as other sinker ball pitchers, like Carmona, Lowe, Wang, Webb, etc.

What say some of you pitch gurus and Tiger fans?

Wednesday, September 28, 2011

Kershaw v Strasburg

In their last 4 starts:
IP: 24, 28
K: 24, 26
BB: 2, 4
HR: 0, 2
H: 15, 14
R: 5, 4

Which is who, and who is which?

(They are also born 4 months apart.)

In their last 90-something innings (i.e., Strasburg’s career v Kershaw since July 20):
IP: 94, 96
K: 116, 93 <--
BB: 19, 17
HR: 5, 5
H: 71, 67
R: 30, 17 <--

Matt “anti-DIPS” Cain

Josh’s great investigation.

Sunday, September 25, 2011

WOWY and catcher framing

I used Mike Fast’s list of the best and worst framing catchers from 2007-2010 and did a WOWY between them and all the starting pitchers they caught in that time period. If you have not read Mike’s recent article on quantifying catcher framing using pitch f/x data, you should.

Anyway, I compared the wOBA against, NIBB, and SO rates of all the starting pitchers (when they started) matched up with each of the catchers in my sample and those same starting pitchers when they pitched (again, only as a starter) with another catcher (regardless of the team).  I weighted each of the differences by the lesser of the two PA and summed and averaged all the data (the weighted differences) for two groups - the 10 best framing catchers and the 10 worst, using regressed (adding 60 games of league average framing) runs per 120 games.

The 10 best (at framing, per Mike Fast) were

Alex Avila
Hanigan
Lucroy
Russell Martin
Jose Molina
Montero
Ross
Saltamacchia
Torrealba
Zaun

The 10 worst were:

Doumit
Hill
Hundley
Ianneta
Johjima
Rob Johnson
Kendall
Laird
Posada
Treanor

Here is an example of the methodology:

Say there are 2 catchers in the good group, A and B.

And say that A caught pitchers I and J, and B caught pitchers X and Y.

Say that pitchers I and J had a wOBA against of .300 and .310 with catcher A behind the plate and .290 and .306 with all other catchers behind the plate.

And say that the number of PA for pitcher I with catcher A was 100 and for pitcher I and all other catchers it was 300.  The difference in wOBA is .300-.290, or .010, weighted by 100 (the lesser of the two PA).

For pitcher B, the difference is .310 - .306 or .004, weighted by the lesser of those two PA, say, 150.

We do the same thing for catcher J.

We take the weighted average of all the wOBA differences.

I did the same thing for BB and K rates (per PA).

We would expect that the good framing catchers would have a better wOBA with their pitchers than with those same pitchers and another catcher.  We also expect them to have a better K rate (higher) and BB rate (lower).  Vice versa for the bad framing catchers of course.

Keep in mind that I did not control for anything else which might effect the offensive environment, such as park, weather, strength of the offensive opponents, etc. Also keep in mind that any WOWY differences found would include pitch calling skill by the catcher.

Anyway, here are the results:

Higher wOBA by 2 points, which is around .07 runs per 9 innings.  Lower K rate by .23 K per 9 innings.  Higher BB rate by .19 per 9.

Obviously there is some noise in these numbers.  I don’t know the standard error off the top of my head.  The total number of PA for each group for purposes of weighting (IOW, summing all the “lesser of the PA") is around 50,000, which is a lot.  There also may be biases, as I said, caused by different parks, opponent offenses, weather, etc. in the withs and withouts. 

Again, these numbers include pitch calling.  In fact, one reason why the bad framing catchers only had 2 points of wOBA worse may be that they are superior pitch callers otherwise they would not be catching given their poor framing skills.

I did the same test using out of sample data - 2003-2006 - for the same catchers.  Obviously some of these catchers had little or no playing time in those years.  Here are the results for them for this time period, 03-06.  If what Mike is capturing is indeed a “skill,” (and his y-t-y correlations suggest that it is a strong skill) we should see similar results in the out of sample data.  Indeed we do!

Good framers

Higher wOBA by a whopping 14 points, which is around .46 runs per 9 innings!  Lower K rate by .46 K per 9 innings.  Higher BB rate by .15 per 9.

Here there is no evidence that the bad framers are good game callers.  Just that they are terrible catchers overall!

Wednesday, September 21, 2011

Talking pitching

Loved the exchange between Eric and Dave.  This is exactly the kind of stuff that is great for those that aren’t in the deep end of sabermetrics to follow along.  You have two people that engage in a debate, listen to each other, and acknowledge the other’s viewpoint as being reasonable.

Monday, September 19, 2011

Tennis/golf rankings… for pitchers!

I’ve always been fascinated by the world rankings for tennis and golf.  I was especially interested in Serena being ranked #28 for the US Open.  The best way to see how a system works is to look at extreme cases.  So, if someone doesn’t play for a period of months or longer, how does that effect the ratings?  How far back of performances do you consider?

Basically, this is day-by-day Marcel for tennis and golf.  I have no idea how they calculate things.  Presumably Chess rankings are similarly affected, though presumably it won’t need to rely so much on recent data.

Bill James wondered: well, why not for starting pitchers?  And he proposes a system that is simple enough: start everyone at some baseline that you can’t get below (he uses 300), and then remove 3% of his score entering a game, and add in 30% of his Game Score.  Then a daily penalty for missing a start over more than 7 days, and yet another penalty for missing a start more than 200 days.  All very sweet.

I’m not sure it works with those particular numbers, but, the basic framework seems good enough.

Basically, this is day-by-day Marcel for starting pitchers, but without being so computational heavy like Marcel.

In my view, the best way to test James’ particular implementation is against the daily Marcels. 

***

This is how James’ system sees the rankings for #1 over the past two years:
Tim Lincecum August 1, 2009 to September 6, 2009
CC Sabathia September 7, 2009 to September 13, 2009
Tim Lincecum September 14, 2009 to September 18, 2009
CC Sabathia September 19, 2009 to September 30, 2009
Tim Lincecum October 1, 2009 to April 15, 2010
CC Sabathia April 16, 2010 to April 19, 2010

Roy Halladay April 20, 2010 to May 3, 2010
Tim Lincecum May 4, 2010 to May 17, 2010
Roy Halladay May 18, 2010 to September 22, 2010
Felix Hernandez September 23, 2010 to October 5, 2010
Roy Halladay October 6, 2010 to July 30, 2011

Justin Verlander July 31, 2011 to the present

So, Lincecum and CC trade off the #1 spot for a few months, then it’s Halladay with some blips from Lincecum and Felix, and then Halladay keeps control until Verlander.

Like I said, I like the basic premise of it.  Now it’s just a matter of whether everything is as balanced as it should be.

Tuesday, September 13, 2011

Draft Order and Major League Pitching Performance…

I know there has been similar research published on the web, but I am too lazy to look it up right now.  I took the 1998-2010 draft list from BA and looked at how the various pitchers did in the major leagues, breaking the rounds down into various buckets.  It was a quick study and I just matched names from the draft lists with my major league databases.  I probably missed 3-5% of the players because the names, especially first names, did not match up exactly.

First I looked at the rookie years for all pitchers in each draft round 1-3+, for a total of 4 buckets. Each round included the supplemental rounds, so, the first round actually has 60 picks in most years.  The rest of the rounds pretty much have 30 picks each.  Perhaps I should consider the first 30 picks of the 1st round as the 1st round and then the next 30 picks in the 1st round supplement as the 2nd round, etc.

Does anyone know if there is anything special about the supplemental round after the first 30 picks or is it just the next 30 best players and then the second round is 61-90 best players?

For each pitcher, I looked at whether their primary role in their rookie major league season was as a starter (S), a reliever (R), or mixed (N).

The currency I used was my normalized, component ERA (nERC), which is an “ERA” based on a pitcher’s raw stats (s, d, t, hr, nibb, hp, outs, and wp) adjusted for park, opponent, and defense, and normalized to his own league, where the average pitcher in each league, weighted by TBF, is 4.00.  I weighted the aggregate nERC in each bucket by each pitcher’s IP.  If I used the simple average of all the pitchers (weighting each pitcher exactly the same regardless of how many IP they threw), the numbers would be much higher, as the pitchers who had the worst true talent generally had the fewest IP.

Read More

Monday, September 12, 2011

Poz on WAR and pitcher wins

He pummels it like no one else can:

But some Brilliant Readers did get my intention, and one in particular made a well-reasoned argument that wins, flawed as they are, do tell us with a pretty decent sense of accuracy whether a pitcher is good or not, especially over a long career. OK, put that thought away for a minute.

The next email was from another Brilliant Reader [BR] who had myriad complaints about WAR. This too was well-reasoned, and it made the point that WAR is far from perfect, that the formula between Baseball Reference and Fangraphs is quite different, that it’s ridiculous to take out the human element from baseball analysis and simply determine who is the best player by the decimal points of WAR.

It was good to read those back-to-back, because in just two emails I felt like I had seen the arc. The first BR wants too little from stats. The second BR expects ways too much.

On a single-season basis, WAR is better than pitcher wins.  On a career basis, WAR is better than pitcher wins.  That’s apples to apples.

Career pitcher wins may be better than single season WAR.  Career WAR is far far better than single season pitcher wins.  Apples and oranges.

So, on an apples-to-apples comparison, WAR wins. 

Cain v Shields, part 2

In the off-season, I noted the similarity and differences of Cain and Shields.  They both had a similar FIP entering 2011, but Cain had a much better BABIP, and sequenced his events better, hence Cain allowed runs at a much lower rate than Shields.

This is why god invented the bet.  If you really believed that there is in essence no difference in overall talent between Shields and Cain, then you should be taking bets to that effect, that Cain and Shields are expected to be on the mound a similar number of times when runs are scored.  But, if you believe that Cain is in fact more talented in getting positive outcomes when the ball is in play, and/or Cain is in fact more talented in stringing events to minimize their impact to run scoring, then you should be taking bets to that effect.

So, how’s 2011 been treating them?  They’ve both faced a similar number of batters (825 for Cain, 850 for Shields).  Both have the same number of walks (excluding IBB, including HBP: one has 57, the other has 58).  Cain has 41 fewer K, but he also has 13 fewer HR.  The result is that Cain has a much lower FIP.

As for Cain’s BABIP: he continues to maintain his very low BABIP, as he’s at .260 (compared to his career .265).  Shields however matches him at .261 (compared to career .300).

However, the real test in all these things is runs allowed.  Remember: betting on FIP entering 2011 meant that you’d bet them to be about equals.  Betting on also including BABIP and sequencing would mean that you’d bet on Cain.  So, how did the test go?  How many runs have they allowed?

Shields has allowed 72 runs, while Cain has allowed 73. 

So, in a weird way, betting on FIP was right.  Except of course that the individual components to those runs allowed is not what we were trying to sell.  We thought Cain’s BABIP would increase, except Cain maintained his superlow level, and Shields met him there.  We thought their FIP would remain the same, but Cain crushed him by simply not allowing HR.

Saturday, September 10, 2011

What happens to a pitcher’s HR rate year to year?

In this case, I’m looking at pitchers with at least 300 PA year to year in the same park, 1993-2010.  So, what we see here is not only pitcher skill, but also park effects.  Follows the same form as the BABIP chart.

In this case, regression is about 65%.  Obviously, if I control for park, it would regress even more.

***

UPDATE: This is when I look at road parks year to year.  This effectively controls for the park bias, leaving us only with the pitcher as the sole variable.  Regression is also around 65%, indicating that the park is not really a bias in the data.

Friday, September 09, 2011

What happens to a pitcher’s BABIP year to year?

I took all pitchers with at least 400 balls in play since 1993 and calculated their BABIP (batting average on balls in play… for my purposes, I include reaching on error).  I put the pitchers in buckets (e.g., the .300 bucket was all pitchers with BABIP of .295 to .305).

The first column to the left is the Year 1 BABIP bucket.  The columns on the right are the Year 2 BABIP buckets. 

n is the number of pitchers in the Year 1 bucket.

The % are the % of pitchers that landed in the Year 2 buckets.

In blue is the median.

We see therefore a slight effect that as your Year 1 BABIP was bigger, you have a slightly bigger chance in landing in the higher BABIP buckets in year 2.

Those who were in the .260 bucket in year 1 had a 26% chance of being in the .320 bucket or higher.
Those who were in the .360 bucket in year 1 had a 45% chance of being in the .320 bucket or higher.

The very last column is the average BABIP in year 2.

Roughly speaking, there’s about an 80% regression toward the mean.