The Yu Controversy -- Learning to win, holding leads, and random variation

Who knew a home run by Adam Dunn in a random August game would end up triggering so much drama?

It was six days ago, on August 24, that Adam Dunn launched a two run bomb off of Yu Darvish in the bottom of the sixth inning in Chicago, ending Darvish's shutout bid and tying the game at two.  The Rangers had just taken the lead in the top of the sixth on a two run Alex Rios home run, and Darvish's failure to execute a shutdown inning, his allowing a homer in a game the Rangers eventually lost 3-2 when Tanner Scheppers allowed a walk-off single in the 9th, has become Exhibit A in a suddenly hot debate over Darvish's ability to hold a lead and perform in close games.

Eric Nadel had some comments during the radio broadcast of that game that were perceived by some as being overly critical of Yu, and then followed that up with a series of Twitter exchanges on the subject.  The next morning, Jamey Newberg sent out a report simply entitled "Yu", which included the following:

Do you ever wonder how good Matt Garza would be if he had Yu Darvish’s stuff?

Do you?

Garza’s mound temperament isn’t perfect, but I kinda like imagining Darvish if he had some of that.

Can he make that different? Can he make that better? Is that possible?

Isn’t it fair to say Garza’s stuff plays up at least a little bit because of it?

Is it fair to ask whether Darvish’s stuff actually plays down, at least from time to time?

* * *

Pitcher wins are not direct functions of pitcher effectiveness. Obviously, team wins when a pitcher starts aren’t either, as the bullpen factors into every no-decision, by definition.

But is it fair to expect an ace to reach for that next gear when given a lead, when facing the bottom third, when the instant situation calls for a beast to put everyone in the same uniform on his back?

I, as those of you who have read my stuff for a while could probably guess, came down on the side of those who said, it isn't reasonable to expect an ace to overperform in those situations, that the idea that certain pitchers can reach for that next gear (Jack Morris being the classic example people reference) does not have much support in the statistical data available.

That being said, there are quite a few folks who share the view that Yu needs to, well, learn how to win.  Richard Durrett took that position in a blog post on ESPN Dallas:

I consider Darvish an ace, but to be mentioned consistently with Hernandez, Justin Verlander and other top pitchers in the AL, Darvish must win more of the close games. The Rangers took a 2-0 lead in the top of the sixth on Saturday after both teams put up zeroes to that point. Darvish then gave those two runs back in the bottom half of the inning. Dunn's homer came on a good pitch, frankly. But Darvish had other pitches in that inning that weren't good, and he allowed a hit to make Dunn's shot the tying homer rather than a solo shot.

* * *

Don't get me wrong: Darvish is terrific. He's going to be a Cy Young candidate year in and year out, if he stays healthy. And Saturday proves he can still get even better. He'll learn to shut the other team down more consistently late in close games. When he does, he'll be collecting those Cy Young trophies. That's coming, if you ask me.

* * *

Darvish, already on an extremely high level, has another level he can reach. He's got the stuff, the fight and the talent to do it. It's not easy. But is it fair to hold him to that standard? I think it is, based on what we've seen. He'll learn from starts like Saturday, and, come playoff time, he'll be even better for it.

T.R. Sullivan's article yesterday echoed the sentiments Durrett expressed:

From the moment the Rangers paid a record-breaking $51.7 million posting fee to the Nippon Ham Fighters and then signed Darvish to a six-year, $56 million contract, the anticipation has been he would be the No. 1 starter on the staff.

There is evidence to suggest that Darvish is close to achieving that. There is also evidence that he is still shy of that mighty badge of pitching honor.

* * *

Washington said there are two issues Darvish needs to overcome to reach the level of a true No. 1 starter. He needs to learn to hold on to leads and he needs to cut down pitch counts to get deeper in a game.

Darvish has had four starts this season where he had the lead at some point during the game and the Rangers ended up losing. The latest was on Saturday against the White Sox, when Alex Rios hit a two-run home run in the top of the sixth to give the Rangers a 2-0 lead. But Darvish gave up a two-run home run to Adam Dunn in the bottom of the sixth, and Texas ended up losing the game in the ninth.

"That separates the grand pitcher from the one who is still learning," Washington said. "That's what he has to close. He's got to close that gap. He'll get better at it."

And this highlights the struggle I have with this argument.  I think we all agree that Yu Darvish is awesome.  I think we all agree that Yu Darvish could be better.  The post Brad did yesterday highlights that Yu Darvish hasn't been awesome, relative to his peers, when it comes to avoiding late-game "meltdowns" (though he has still been well above-average).  And the Rangers' record in games Yu Darvish starts aren't as good as you would expect the team's record to be, given Yu's numbers.

But does that really mean that Yu needs to "learn" how to shut down the other team in key situations?  Can you "learn" to hold a lead?  Is that even a skill?  Is it something that can be learned?  What, exactly, is Yu supposed to "learn from starts like Saturday"?  If there isn't a skill or ability to being able to find that extra gear and go to 11 in a close game or after getting a lead, if there holding a lead isn't a "skill," then you obviously can't learn how to do that.

I acknowledge that there are those who disagree with this point.  Clearly, Ron Washington seems to think it is a skill, is something that can be learned.  And he's not alone.  Eric Nadel, whose comments seems to have been a catalyst for much of this discussion, talked about the issue yesterday morning on the Ticket.  One of the things he emphasized was that, while sabermetrically inclined outsiders suggest that keeping or losing leads late the in games is primarily due to random fluctuations, he doesn't believe it is random, and neither do former players he has talked to.  I encourage you to listen to the whole segment here, since he expands on and clarifies his position regarding Darvish, but there are also a few items I've transcribed as follows that I want to talk about:

Well, first, let me say, I love Darvish, I love watching him pitch, I think he's really good, but there are areas he needs to improve.  I don't want to get into the debate anymore, but the comments that I made about how he pitches late in games, games that are close, shutdown innings, the comments that I made were based on seeing every pitch that he's thrown, and my memory of things that I've observed, and things that I've said such as something about failing to execute a shutdown inning.

* * *

But one of the websites where people were attacking me, and some of the attacks became personal, is Lone Star Ball.  And Lone Star Ball this week has been doing a study to see if what I was saying about the late inning performances actually had validity . . . and it will prove my point about the issue of performance late in close games.  So it turns out I was in fact correct . . .

* * *

Some say that the performance late in close games is totally random -- that's the point of some of these people (editor's note -- such as me) and you know, that's their opinion.  But all the players, all the ex-players I've talked to about this, a substantial number of people believe this is not random.  That there is a "clutch" element to performance.  And I tend to listen to them on this point, rather than people who didn't play the game, and of course I didn't play the game either.  So that's why I tend to listen to players when they tell me, maybe the player is trying to hard, maybe its not a matter of feeling pressure but trying too hard, maybe guys are consciously trying to relax because its a pressure situation and they don't perform as well.  Who knows?  But I can't believe that its random.

Eric eloquently sets out what is really, I think, the crux of the issue here, the interesting item for us to think about:  not whether Yu Darvish is good, or fits someone's definition of an "ace," but the philosophical difference between someone like me, who believes that most of what we think of as "clutch" can be explained by random variations, and the view of those who don't believe that is random (and secondarily, the question of whether Yu Darvish's performance in late and close situations this season is due to Darvish not knowing how to hold a lead/handle these sorts of situations, or random variation).

The point that people who play the game believe that its not random, that there is a clutch aspect to this, isn't easily dismissed, though.  Given that, I sought out someone who is in the overlap of the former player and sabermetrically inclined Venn Diagram -- Gabe Kapler, the former major league outfielder and minor league manager who has done a couple of pieces recently for Baseball Prospectus -- and ask him his thoughts on this, and he was gracious enough to speak at length with me on the topic:

"I don't think that you learn it.  I don't think Yu Darvish is going to learn anything that is going to help him close out games, or have a two run lead in the 9th and dial it up.  I think he has the physical ability to hold a two run lead, when necessary, and I don't think he's going to learn anything going forward.  I think you could point to a similar number of games for Justin Verlander, or C.C. Sabathia, or any other big game power pitcher where they gave up the lead late in the game.  So no, I don't think Yu Darvish needs to learn anything -- I think its in there, and its like a hitter who starts his career with a bad stretch of at bats.  It doesn't mean that he's a bad hitter -- it just means he's having a bad streak of at bats.

And I think there have been enough studies about the way that people are motivated to know that some players are motivated by fear and others are motivated by positive reinforcement.  So in that case, you can make the argument that environment counts, and that some players are going to have more difficulty if their backs are against the wall because they are scared and perform better when they have easier situations.  And vice versa -- for example, stepping into a situation like Boston or New York, after having played in Seattle or Tampa.  I do think there have been enough studies around psychology and on adrenaline where there are scientific elements of this discussion worth looking into.

But in this case, what I think we're talking about here is a sample size issue.  We don't have enough data here to make any sort of determination with Yu Darvish.  Analysts in baseball tend to try to look at baseball in a nutshell, rather than in a huge sample size, which is really how we can best analyze players, and try to predict performance.

Yu Darvish is going to dominate situations like this going forward, and everyone is going to say, see, he learned how to do it.  But he's always known how to do it.  He's got arguably the best slider in the game, he's got top of the class fastball velocity, he's striking out more batters per 9 innings than any other starter in the game right now.  Its an absurd discussion, frankly.  He's going to do what everyone is asking him to do, but I don't believe its going to be the result of his having learned it -- he's performed admirably to this point, and the concerns about whether he knows how to win or close out games can be thrown out because there simply isn't enough of a sample size to judge him on."

I think what Kapler says is fair, and makes sense.  I'm not globally going to dismiss the idea that there are factors at work that can impact how a player performs in the clutch, or in blowout situations.  At the same time, though, as Kapler also emphasizes, the sample size to make the signal to noise ratio meaningful in evaluating those traits in individual players is enormous, and we aren't close to that level with Yu Darvish.  Jack Morris has long been the poster child for "pitching to the score," "knowing how to win," and the other traits that people have been questioning whether Darvish has at this point.  However, Joe Sheehan, having gone through every game of Jack Morris's career, pretty well debunks that notion.

If a tree falls in the woods, but it doesn't make a sound that is statistically significant compared to the background noise of the woods, does it really make a sound at all?  If a pitcher pitches to the score, knows how to win and has an innate understanding about how to hold a lead, but it still doesn't make an impact that is larger than the margin of error for the sample size in question, does it matter if that clutchness exists?

Ben and Skin talked to Jon Daniels about this issue as well on their radio show yesterday, and asked JD his thoughts on the discussion.  Again, check out the whole segment, but this is what I thought was particularly noteworthy among what JD had this to say:

I think an ace, or a #1, is something that is earned over time . . . you can't just anoint someone as that . . .

* * *

I think its also important to remember, you look at some of the greatest pitchers of all time, Roger Clemens and Nolan Ryan and Greg Maddux -- a couple of those guys we work with right now -- and at different times in their careers, they had the reputation of, "Can't win the big game," or "Not a playoff winner."  And now you look back in retrospect at how silly those kind of statements seem, because you get tied up in a game, or a guy gives up a hit or a run, and you kind of want to draw conclusions, or you think you've just seen something, but the reality is that you make these judgments over a long period of time, and I don't think you can really sit here and say somebody is or isn't -- bottom line is, Yu's a pitcher we're thrilled to have, he's one of the best, most talented guys out there, and every other team would love to have him.

So Daniels hits on a similar theme to what Gabe Kapler was talking about -- when analyzing, there is a tendency to over-emphasize what happens over a short period of time, when the reality is that you need to have a much large sample to really be able to weigh in on such an issue.

Finally, one of the trump cards that those who question Darvish have played is that, well, the team's record isn't very good when he pitches.  The Rangers have a worse record with Darvish on the mound than without Darvish starting...how can that be, if he is an ace?  Particularly given his run support this season has been very good -- over 4.5 runs per game?

Ah, but the run support average can be misleading.  The Rangers have scored a lot of runs for Darvish in total, but many of those are the result of huge offensive nights by the bats -- in 9 of Darvish's 25 starts, the Rangers have scored 7 runs or more.  On the other hand, in 13 of Darvish's 25 starts, the Rangers have scored 3 runs or fewer.

In only 3 of the 25 games Darvish has started have the Ranger bats scored between 4 and 6 runs.  And that distribution goes a long way towards explaining why the Rangers have won 14 out of 25 games he's started, rather than significantly more.  You're going to win most of the games you score 7+ runs in.  You're going to lose most of the games you score 3 or fewer runs in.  The games that can really swing a won/loss record are the 4-6 run games.  And there simply haven't been that many of those when Darvish has been on the mound...the Ranger offense has been either feast or famine in support of Darvish.

In further support of this point, I'm going to quote -- wait for it -- Randy Galloway, from his column this morning:

From here, the frustration level with Darvish remains extremely low. Actually, when there’s frustration, it centers on the lack of run support for Yu, and that’s an area totally out of his control.

If he’s not an ace now, he’s an ace waiting to happen.

* * *

But do consider the following:

Darvish has an L when he lost games by scores of 3-1, 1-0, 3-l and 1-0.

In team losses when he started, and in games when Yu went six-plus innings, he got two runs and gave up one (the bullpen blew up), he got three runs and gave up three, he got three runs and gave up three, he got three runs and gave up three and he got two runs and gave up two.

So in nine of the 11 team L’s when Yu was on the mound, he gave up three runs or fewer.

Sorry for that kind of numbers-whipping, but, yes, I’m grumbling about the Yu grumbling.

So yes, look at what that seemingly generic Adam Dunn home run has wrought.  It has brought me and Randy Galloway together in a common cause.