Has the League Figured Yu Darvish Out?
During Yu Darvish’s first eight starts, he faced eight different teams, making his first regular season start against each of them. His last four starts, however, have been repeat performances, as he’s faced the Mariners, Angels, Blue Jays, and Athletics for the second time. During those four starts, he’s been awful.
May 21st, @ SEA: 4 IP, 5 H, 4 R, 6 BB, 5 K
May 27th, vs TOR: 5 IP, 7 H, 3 R, 3 BB, 3 K
June 2nd, @LAA: 6.1 IP, 6 H, 3 R, 3 BB, 7 K
June 7th, @Oak: 5.1 IP, 6 H, 6 R, 6 BB, 4 K
You don’t need to know much about statistical analysis to know that 18 walks in 20 2/3 innings pitched is not good, and any pitcher issuing that many free passes probably isn’t going to be successful. But, is the recent failure to throw strikes related to opposing batters learning how to approach Darvish after getting an earlier look at him?
That appears to be a fairly popular theory at the moment, but let’s look and see whether the evidence supports the idea. Let’s start with the plate discipline stats, which seem like the most likely place where a change in batter approach would be the most noticeable.
In his first eight starts, 46% of the pitches Darvish threw were defined as being in the strike zone according to PITCHF/x. Opposing batters swung at 43.2% of these pitches and made contact on 74.4% pitches that they swung at. All of these marks were a bit below the league average, but not extremely far from what the norm.
So, what about in these last four starts? Darvish has actually thrown a higher rate of pitches that PITCHF/x thinks have been strikes (48.3%), but opposing batters have kept their bats on the shoulders more often. The swing rate against Darvish since May 21st is just 40.1%, with the decline coming both on pitches in and out of the zone. Three of the four teams swung the bat less often in their second match-up against Darvish than they did in their first.
SEA: 41.8% Swing% on 4/18, 39.6% on 5/21
TOR: 45.4% Swing% on 4/30, 42.4% on 5/27
LAA: 41.9% Swing% on 5/11, 44.1% on 6/2
OAK: 44.9% Swing% on 5/16, 34.6% on 6/7
Patience has never really been the Angels thing, and it’s probably not a coincidence that his start against Anaheim is also the only one of the last four where Darvish has actually performed decently. The A’s, Blue Jays, and Mariners were all content to let Darvish pitch himself into trouble, and he responded by doing just that.
However, we can’t ignore the fact that PITCHF/x thinks Darvish’s rate of throwing strikes has actually gone up lately. In fact, his starts against Settle and Toronto at the end of May are his only two starts all year where the system believes that more than half of his pitchers were in the strike zone, and yet, he issued nine walks and only had eight strikeouts in those two starts. What do we make of that?
Maybe the batters aren’t the only ones adjusting to Darvish’s reputation for poor command. Take a look at this strike zone plot from TexasLeaguers.com for those two games against the Mariners and Blue Jays:
There are 16 pitches within the boundaries of the strike zone as defined by PITCHF/x that were called balls and exactly one pitch that is outside the defined borders that was called a strike. The starters for the Mariners (Felix Hernandez) and Blue Jays (Kyle Drabek) combined to have seven pitches within the PITCHF/x strike zone called balls in those same games with those same umpires, and they got three pitches out of the zone as called strikes.
This isn’t to say that umpires are definitely squeezing Yu Darvish. PITCHF/x has a margin for error, and most of the pitches that were called balls are close enough to the edge that they could have legitimately been low or outside. And, of course, the umpires had nothing to do with Darvish’s disaster start yesterday, where nearly 40% of the pitches he threw were outside the strike zone. However, there is some evidence to suggest that Darvish hasn’t been getting some calls lately, and just as the league has picked up on his command problems and is swinging the bat less, it is also possible that umpires are less likely to give Darvish the benefit of the doubt on marginal pitches that could go either way.
As Patrick Kilgo noted in his presentation at the SABR convention last year, there is some evidence that umpires favor veteran players when it comes to calling balls and strikes, and better pitchers get more calls than others. It’s a widely held belief in the game that guys who are consistently around the zone get the benefit of the doubt on borderline calls, and Darvish is certainly not a guy who can paint the corners with consistency.
So, are big league hitters adjusting to Darvish? It does look like they’re trending towards swinging less often, which makes sense considering his command problems. However, they’re not the only ones who seem to be adjusting to Darvish, and he’s going to have to be aware of the fact that he’s probably a bit less likely to get calls on the corners until he establishes himself as a guy who can throw strikes with regularity.
The batter/pitcher match-up is a game of constant adjustment. If opposing batters are going to make Darvish throw strikes, he’ll need to adjust and be more aggressive in getting getting ahead in the count. The best way to get hitters to chase pitches out of the zone is to put them in defensive counts, and if they’re going to take a more passive approach at the plate against him, he has an opportunity to throw first pitch strikes and start hitters off 0-1 until they adjust.
Darvish has Major League stuff – and despite his recent struggles, his command doesn’t appear to be getting worse – but Major League hitters aren’t going to keep getting themselves out if he doesn’t start giving them a reason to. Whether it’s a mechanical adjustment or simply a decision to actually use the entire plate (right now, he just doesn’t throw inside against left-handers), Darvish is going to have to make some changes in order to counter the reputation that he garnered in his first two months in the big leagues. Good stuff and bad command is a better place to start from than vice versa, but if Darvish is going to become the guy he was billed as, he’s going to have to attack the strike zone in a way that he hasn’t so far.
Dave is the Managing Editor of FanGraphs.
Umpires favoring pitchers or the home team is nothing new. Scorecasting had a bunch of interesting stats on that.
Honest opinion? I have to believe people overrated him coming into the league. The last high-profile Japanese starting pitcher (Dice-K) had the same exact problems, but nobody wanted to make the comparison because they were scared to be labeled as semi-racist for making the comparison only because they were Japanese.
At some point, though, do you consider whether there is a cultural difference in the way Japanese pitchers are trained, practice, etc. that may make them more likely to have command/control problems in the majors?
At some point you can consider your hypothesis, but that point isn’t probably at n=2.
n=1 is enough to make certain claims. Darvish had great control for several years in Japan. This year his BB rate is 3 or 4 times higher than it was in Japan. That can’t be chance, and something has to have changed…either Yu or his pitching context. I’d say the latter is more plausible, which would mean there’s something different about pitching in Japan compared to here. What is it? That’s where we get into the hypotheticals…
Whatever it is, it doesn’t seem to affect every pitcher the same way. Matsuzaka’s walk rate suffered, but if you look back at Nomo’s stats, for instance, his walk rate actually improved significantly the year he crossed over. Kuroda’s rates stayed identical.
If anything, I’d say the lesson is that it’s very hard to predict how a Japanese pitcher’s control will translate to the MLB, from his Japanese stats. That’s not a trivial point, because generally walk rate is considered something that is fairly stable for a starter.
My intuition (having never watched a Japanese baseball game) is that the strikezone over there is likely just different. So for some pitchers, they come over here and, effectively, they’re throwing to one strikezone, and the umpire is calling another. It doesn’t help that the hitters also are privy to where the strikezone is and therefore have an advantage in commanding it during an at bat.
For other pitchers, the MLB strikezone might actually help. Maybe Nomo’s arsenal hit the MLB strikezone with greater regularity than it hit the Japanese strikezone, without him ever having to change his approach (the way Yu may have to).
Aren’t you cherrypicking Japanese pitchers who happen to have command issues? Keep in mind that that there are Japanese pitchers who definitely do not have any such issues: Kuroda, Uehara (he of the 0.89 BB/9), etc. So it appears that there is a range of ability among Japanese pitchers when it comes to control….just like among the rest of MLB pitchers of all nationalities…..
He should definitely be trying to emulate Kuroda, who has a similar arsenal of pitches but commands his FBs well. The problem is that Yu Darvish was not hyped as the next Kuroda, and nobody would have offered anywhere near what Texas paid if he was. Which is funny, because a young Kuroda is easily worth what Darvish signed for. Expectations should probably be adjusted, but by no means does expecting young Kuroda numbers pitching in Arlington make him a poor value or a bad pitcher.
I’ve read that the size of baseballs used in Japanese leagues are slightly smaller. That always struck me as being a huge difference for pitchers that could easily lead to control problems, but the size difference is hardly ever mentioned on this site.
Another article fellating the Rangers. FG continues to ignore upstart teams like the Blue Jays!
It shouldn’t be surprising he’s not getting pitches called out of the strike zone balls. If the Rays could hit their way out of a paper bag, even they would score off this guy.
Ah I’ve got an imitator! nice. I must be doing something right.
You wish you were the real Everdiso.
There was an impostor Well-Beered Englishman comment one time. After deciding against the possibility of my posting and not remembering, I found it both bizarre and flattering.
It should be even more flattering that W-BE impostery stands out like spotted dick on an Olive Garden menu.
nobody wishes they were the real everdiso. your constant trolling/jealousy of your AL east rivals is played out enough on your regular blog.
for the record:
Everdiso = whiny red sox excuse-maker
everdiso = super-awesome, brilliant, and good looking jays fan
What? Who am I? How did I get here?
Wow… someone is having fun with name it appears. Awesome.
dude anthopolous and the jays get slurped here on the reg.
It would be a shame if Texas ended up with the new Jonathan Sanchez.
No, it wouldn’t.
It baffles me why teams are still so willing to throw egregious amounts of money at Japanese players. It’s much too early to evaluate Darvish, but how many examples do you need of imports that completely bomb relative to the value of their contracts?
Yeah no Japanese players are ever worth their contracts. 3 of 4 of Yu’s most recent starts were poor? Stop buying Japanese players!
I +1’d your post because of your username.
Of course not, but everyone should realize that when ‘film arrives’ stuff is overrated if you can’t hit a target and don’t know the hitters.
Of the 25 qualified starters who throw 6 pitches or more, Darvish has the second highest percentage of four seam fastballs thrown next to Guthrie. His values for that pitch are very poor, and they’re not great for his other fastballs that total over 70% of his pitches thrown. Compare to Jake Peavy, who throws a similar percentage of both fastballs overall and four seamers, and the difference is stark.
To expand on Dave’s excellent observation, it looks like he just isn’t commanding the four seamer or two seamer at all. You could argue that it’s a veteran bias, but I’d argue that established pitchers get more calls because they have learned to command their fastballs and overcome the fear of MLB hitters blasting it since they’re good and all. If you can’t do that, you’re a straight-up junkballer in the long run, and no pitchers starts out wanting to be that. It will be interesting to see if Yu can make that adjustment. Interesting, he should be trying to emulate Kuroda, whose success is very much associated with command of a few different fastballs.
The article is on point. From a batter’s standpoint, umpires used to
do the same with Frank Thomas, the Chicago White Sox slugger.
He was such a great hitter, that he established a reputation for
having a keen eye, and his stats proved it. For years the umpires
would give him the benefit of doubt on close, or berderline,
pitches which only served to reinforce his stats. Baseball is a
a game of adjustments. Players who adjust go on to great
careers. Even the great Babe Ruth went through a similar
problem. Batters discovered a habit he’d gotten into while
throwing a certain pitch. They’d wait for it and clobber it.
When he corrected that flaw, Ruth was again unhittable.
I’m not sure that’s true about Frank Thomas. Oh you mean the Chicago White Sox slugger? Not the 1950’s Pirates outfielder? Then I agree.
Robot umpires.
Robot umpires.
I said this in an earlier thread on Darvish, his fastball in the zone is very ordinary and hittable for his velocity. He has extraordinary breaking stuff on par with the absolute best, but there are times when he loses complete control of the fastball and has to throw his straight fastball down the middle to get a called strike. That leads to getting hit.
Who cares about Darvish? I want an article about Lance ‘9 win’ Lynn.
Yu must be new here. We don’t fancy wins as much as Yu think we do.
C’mon Bryz, don’t be a (9 win) Dickey.
I think a Dickey article is on the horizon. He’s taken the knuckleball to a whole new level, and he’s putting up ace-like numbers. To have that walk-rate as a knuckleball pitcher is amazing. And he’ll make, what, 1/3 of what Yu makes?
It’s like expecting an article on Melky Cabrera and Vogelsong, two players guilty of failure to regress.
I have to wonder if some of those pitches called balls inside the zone are not because of his getting squeezed at all, but maybe because what ended up as an inside fastball to a LH hitter was supposed to be 3″ off the plate outside. I know that technically a strike is a strike and the umpire should call it as such, but it’s hard to get the call if the catcher has to make an emergency move 18″+ to his right to snare an off-target pitch.
I’m not sure how this would be documented aside from a very comprehensive visual review of the video, but just a thought I had.
I thought Mike Fast had documented this prior to his Astro career. That active catcher gloves don’t get calls.
It definitely appears as though Darvish is getting squeezed (many of his pitches that are in the strike zone are called balls). But when he misses so wildly at times, it just makes it tough for umpires to call those close pitches. If he were consistently off the plate slightly (never missing by wildly), those balls would more likely be called strikes by the ump – let alone the pitches that actually are in the strike zone.
Dave,
You assume that the difference in strike percentage between Darvish’s first 8 and last 4 starts warrants an explanation. You have not shown that they are statistically different! It is almost certainly the case, if you were to actually do the statistical test, that the difference you are trying so hard to explain, can best be understood as random sampling variation. You have to reject the null hypothesis before you start offering up complicated explanations.
Fair point, so I took the liberty to do the basic stats. The 95% confidence interval for Yu’s BB% over the first 8 games is ~.10-.20. The 95% confidence interval for Yu’s BB% over the next 4 games is ~.14-.33. Enough overlap there that you wouldn’t reject the null hypothesis using a conventional statistical approach.
If you presume a null hypothesis that Yu’s “true” BB% is his cumulative rate across all 12 games, then both his early returns (first 8 games) and later returns (next 4 games) produced plausible walk rates. About 20% of the time, under this null hypothesis, you would get data at least as extreme as observed in Yu’s last 4 games, by chance alone.
So there is no evidence that anything has changed in Darvish’s ball %. That ought to settle the matter with respect to umpire conspiracy theories (it won’t of course).
I’m pretty sure most of us understand that we’re unlikely to yield a statistically valid result with such a small sample. You could most often say the same thing about a pitch chart that shows a guy didn’t throw as hard as the game wore on and happened to get hammered when he did. I do appreciate your demands for high standards in the statistical analysis realm, but I think it’s misplaced in this case (and probably at least half of the things we look at here where a valid result subjected to that standard is unlikely.
How is it misplaced if it’s true? I’ll cut the author some slack if you want, but save the word count next time and just say “Cut ’em some slack.”
Besides, it’s not the size of the sample per se that’s the issue. If the guy went from 0-1 BB a game to 6 a game, that would still yield a statistically significant difference with the same sample size.
Jason H is just advocating for the completely defensible position that if there is no phenomenon to explain, then there is no explanation needed—let alone a complicated one.
It is true that there are real phenomenon that are not possible to detect statistically (Your example of lost velocity at the end of a start is not one of them, since we have thousands of starts worth of data). This does not, however, justify throwing rigor out the window. Limitations have to be acknowledged and fully discussed when they exist.
In the present example, Dave has put the cart before the horse and assumed that the difference in walk percentage between his first 8 starts and last 4 is due to some change. Whether it is detectable statistically or not, he has to at least address the issue that nothing has changed, since that is always the simplest explanation. Dave should just state, there is an X probability of observing this pattern simply by chance. It is an easy calculation to do. If he observes that there is a high likelihood of observing this pattern randomly, and he still believes something is actually going on, he might then try to argue that you’d need an effect size of Y to detect a statistical difference with such a small sample size and that much smaller effect sizes are still meaningful. But, regardless of how he justifies it, he has to attempt to justify why he is making up stories to fit data that is easily explained by the null hypothesis.
In the end, if Dave were made to justify why he thinks there is really something going, it would probably become very apparent to the reader that this is a lot of bullshit. The great irony to me is that SABR people pride themselves on the idea that they are data driven, but so much of it is narrative masquerading as analysis. Small sample size is universally acknowledged as a problem by authors and readers of this site. That is fantastic. However, it is also routinely ignored in order to write articles, such as this, for which the data does not justify the conclusions.
I am not arguing that these phenomenon should not be addressed. I think they should, because as you correctly point out real phenomenon can be missed for lack of statistical power. Just be fully transparent about the limitations of the data. State the null hypothesis and state the alternative. If you support the alternative, despite inconclusive data, as Dave has done here, state what you expect to see heading forward. Treat the data on it’s own terms.
For example, Dave’s hypothesis predicts that Darvish will continue with an increased walk rate going forward (if nothing again changes). The null hypothesis, which is currently supported by the data, predicts his walk rate will regress towards the mean.
Jason – I mentioned this below but the same thing can be said about swing %. With the exception of the Oakland game the “difference” in swingrates is 2-3 per game….given the variables involved (lineup, time of day, batter’s eye in the park?) hard to say hitters are being more patient when the absolute level is so small.
I’m all for rate stats but I think folks in the SABR community sometimes tend to get caught up on %’s and don’t always realize how small a seemingly large difference is.
You make some fair points, and where analysis like this is based on such small sample sizes it would be very useful if the writer were to acknowledge the percentage chance that such a distribution is random.
Oh but for the record, “data” may be accepted as a plural noun that is treated like a singular one in everyday usage, but “phenomenon” is definitely singular.
/pedant
could it be that the average Japanese batter is shorter than the avg MLB hitter? So he’s trying to utilize a bigger strike zone and just missing?
“There are 16 pitches within the boundaries of the strike zone as defined by PITCHF/x that were called balls and exactly one pitch that is outside the defined borders that was called a strike.”
—
That’s amazing. Thanks!
So glad someone delved into this topic. Anecdotally I felt like he had a lot of bad calls force him into hitters counts. The most aggregious yesterday was a count that should have been 0-2 that turned into 2-0 with the bases loaded. Is there any data by at bat that determines these types of sequences (i.e. flipping from a pitchers to a hitters count based on missed calls)? They dramatically change the outcome of a given plate appearance.
I don’t know if he nibbled in Japan or not, but he refuses to give in, even on 3-0 counts– he always throws for the corners. Until the umps figure him out, he’s going to be severely handicapped.
Man, doesn’t it sound like we’re talking about Dice-K? I know he’s not just total garbage, but he had a million pitches that were supposed to all be plus-plus, he worked slow, he never gave in, etc.
It makes sense that umps tend to call more strikes with a pitcher who pounds the zone. I wonder if they also favor pitchers who work quickly. In either event, it helps them stay focused, and I think the bias for an ump on a borderline call would be a ball.
I dunno…. he’s probably trying to through strikes in counts that favor the batter, no? He might be getting more fouls, too — it should at least be considered in the analysis. You know, foul ball %, looking strike %, swinging strike %. Then maybe break in down into counts that favor the pitcher vs. favor the batter…
What about lurking variables? It’s entirely possible for a pitcher to locate better while issuing more walks depending on the distribution of his walks (i.e., what count the walk came on). It also depends on contact on pitches outside the zone.
Gio Gonzalez syndrome. He will be fantastic as soon as he gets traded to the NL.
Shocker, a douchey commenter on a Cameron article.
the way he shut them down on may 21st?
he shoots he scores
Problem A: The size of the plate in thew Texas Leaguer plot is 24″, actual plate is 17″.
So pitches that look just off the plate that are balls are actually significantly off the plate (4+”). Pitches “in” the zone on the edge are actually 50/50 calls that are 2-3″ off the plate (that are typically strikes, but don’t have to be)
Problem B: The height has a fair degree of error. Not sure if TexasLeaguers is using a normalized zone or raw location but even if it is using a normalized zone the height can be tricky. Also the type of pitch low in the zone is also a factor… fastballs tend to get called for strikes low in the zone more than breaking stuff. Pitch f/x is also looking at height (and width) at one specific plane as well… the plate has depth so this can also be tricky with breaking pitches or even location of the pitcher on the rubber/release point.
Problem C: There is a fairly well documented lefty/right hitter bias on how the umps call strikezones. Righty hitters tend to have a symmetrical zone, lefty hitters generally have anything even marginally off the plate inside called balls, while pitches up to ~4-6″ off the plate outside called strikes. The charts should be broken out lefty/right to reflect this impact. Mike Fast has done a lot of work on this and his “typical” zones are on the plots at BrooksBaseball.net
Also with the exception of the Oakland game, the swing % looks signficant but we are talking 2-3 swings per game with ~100pitches… hardly statistically meaningful despite the rate stats looking like it might. That coild just as easily be different players in the lineup.
The Oakland game was very different, but if you look at the pitch f/x plot from that game, Darvish was missing the zone a lot more; not sure if it was hitters being more selective.
Fair enough. Couldn’t some of this be investigated by comparison with the other pitchers, though? In the article it mentions
“There are 16 pitches within the boundaries of the strike zone as defined by PITCHF/x that were called balls and exactly one pitch that is outside the defined borders that was called a strike. The starters for the Mariners (Felix Hernandez) and Blue Jays (Kyle Drabek) combined to have seven pitches within the PITCHF/x strike zone called balls in those same games with those same umpires, and they got three pitches out of the zone as called strikes. ”
Obviously more comparison could be done to see if there’s a trend, but those numbers at least suggest there could be.
I’m glad the Jays decided we didn’t want much of this bum and passed on giving him a real offer. Astute move by the FO as always.
Then again, after the butchering of Drabek, I’m not really confident in their talent evaluation of pitchers. Who knows. When Adam Lind inevitably turns it around, they’ll still be a force.
How is Adam Lind relevant to this article? I’m starting to become unsure which is the real everdiso and which is the parody.
That there would be my adorable imposter. be nice to him. he’s trying so hard and all.
“for the record:
Everdiso = whiny red sox excuse-maker
everdiso = super-awesome, brilliant, and good looking jays fan”
You still lose due to the last two words here.
I am everdiso
No, I am everdiso.
No, I am Spartacus.
Also, please god someone link me to everdiso’s blog.
I don’t have a blog…but if I did, people would read it.
Heh. nice. at least i got myself a GOOD imposter. that is even more flattering. i like this guy.
How do we know that YOU’RE not MY impostor.
you are adorable.
;p lol fuk yu
Important to keep in mind: Napoli sucks at framing pitches. Not saying it explains it all, but that it’s certainly a factor to keep in mind regarding Darvish not getting borderline calls.
Looks like Napoli was the catcher for three of those four starts. On June 2, Napoli started at first base and moved to catcher later in the game (not sure when).
Well, I’m sure this isn’t what the Rangers are hoping for. Looks like hes been a real chink in the chain!
Can this previous comment be removed please?
Can this previous comment be removed please? I mean, really.