Yu Darvish, Being Great and the Killer Instinct

This weekend Yu Darvish pitched a game which he gave up 2 runs in 7 innings on 6 hits, 3 walks and 11 strikeouts, but the offense only scored 2 runs and the Rangers ultimately lost 3-2.  The catch being that the 2 runs allowed by Darvish occurred on an Adam Dunn home run immediately after being staked to a 2 run lead in the 6th inning.  This prompted some comments from some very respected folks in the Rangers universe on whether Darvish lacked a Killer Instinct and needs to kick it into an extra gear when it really matters.  Of course obligatory internet arguments ensued.

First, let's go through how great Darvish has been in 2013.

Metric	Value	AL Rank	MLB Rank
IP	168.0	12	25
ERA	2.68	3	8
FIP	3.05	4	10
xFIP	2.65	1	2
SO	225	1	1
SO / 9	12.1	1	1
ERA+	156	2	6
bWAR	5.1	4	9
fWAR	4.5	5	8

In both looking at run based results or peripherals, Yu Darvish has been a top 5 pitcher in the AL and a top 10 pitcher in all of baseball.  Many words have been written on the amazing strike out numbers he's posting this year; just in terms of run prevention, he's been excellent.

In fact, if Yu Darvish maintains his current 2.68 ERA and 156 ERA+, those marks will both be the best in franchise history by a right handed starting pitcher, besting Bert Blyleven's 1977 season.  Additionally, they would be third best for any handed pitcher after Rick Honeycutt in 1983 and Jon Matlack in 1978.  The last Rangers right handed starter to post an ERA under 3.00 is Nolan Ryan's 2.91 in 1991, meaning it's never happened in the hitter friendly Ballpark in Arlington which is especially nice to left handed batters.

Simply put, Yu Darvish is having one of the best years in the 42 year franchise history, and arguably the best ever for a right handed starting pitcher.

But... as noted above, there are concerns about his season when it comes low scoring, close games and having the mentality to carry a team.  The Rangers record when Darvish starts is only 14-11 for a .560 winning percentage, lower than the team's current .577.  Even though Darvish has brilliant overall numbers, the ultimate team success has not been what you would expect.  One major reason is the offense has scored 0 twice, 1 twice, 2 twice, 3 three times, 4 once and 5 once in the 11 losses.  In the 14 wins, they've scored 3 or less 4 times but are 10-2 when scoring at least 4 runs and 13-5 when scoring at least 3 runs.  It's hard to blame a pitcher for the team going 1-6 when they score 2 runs or less, but in the perception of aces, coming up with 1-0 or 2-1 wins is part of the game for a lot of people.

In fact, in those 7 games where the Rangers scored 2 runs or less, Darvish has a 2.54 ERA, even better than his season total of 2.68.   Split by Run Support, Darvish's season has been:

Yu Darvish Run Support for 2013

In terms of run prevention, he's been almost the same in games with no support as he's been in games where he receives plenty.

But while that shows that Darvish is a good pitcher regardless of the run support, it doesn't really address the sentiments about kicking it into another gear, giving up leads or doing it when it counts most.  There is a metric called Win Probability Added (WPA) that looks specifically at the impact a player has on the probability of winning a game.  It uses game context such as the current score and point in the game compared to historical results to see what the impact of every play is.  Events like giving up a lead or putting your team in a hole affect the WPA score a lot more than giving up a run when you're up by 10.  Late game events count more as well because there is less time to change the outcome if, for example, you give up a go ahead run in the 8th inning compared to the 2nd.

Win Probability Added for 2013

Darvish still acquits himself well even when context is heavily factored in by WPA.  He's 9th in WPA, just like he's 9th in bWAR, but notice that those two metrics don't closely correlate.  Getting a high WPA value is driven by pitching well in close games, thus pitching in a 5-4 win where you hold a lead will be more valuable than pitching in a 3-4 loss where you give up the go ahead or tying runs, even though your run prevention was the same.

Looking at the gap between Darvish's 2.4 and Kershaw's 4.9 or even Kuroda and Scherzer's 3.6, it starts to touch on the difference in feeling that some have in how Darvish has performed compared to other pitchers.  Darvish has been really good, but is a bigger step behind in WPA than he is in WAR.  Looking at pitcher seasons since 2010 (and prorating 2013 to a full season), only 20 pitchers have had a season with 3.5 or more WPA.

Win Probability Added for 2010-2013 with at least one season of 3.5 WPA

Pitcher	2010	2011	2012	2013*	AVG
Clayton Kershaw	3.5	3.7	3.7	6.0	4.2
Jose Fernandez				4.0	4.0
Felix Hernandez	4.8	1.4	4.2	3.8	3.6
Jered Weaver	2.8	5.1	3.4	1.4	3.2
Cliff Lee	2.2	4.1	2.0	2.3	2.7
Justin Verlander	1.3	5.0	4.1	-0.2	2.6
Cole Hamels	2.5	3.9	2.6	-0.1	2.2
Yu Darvish			1.4	2.9	2.2
Matt Harvey			0.6	3.7	2.1
Roy Halladay	5.1	4.3	-0.4	-0.8	2.1
David Price	2.4	0.1	3.6	2.0	2.0
Hiroki Kuroda	0.8	1.1	1.1	4.4	1.8
Tim Hudson	3.7	1.6	1.3	0.3	1.7
Max Scherzer	1.5	-0.2	0.7	4.5	1.6
Patrick Corbin			-0.8	4.0	1.6
James Shields	-2.2	4.2	0.5	2.6	1.3
Dan Haren	1.2	4.2	-0.6	-0.9	1.0
Josh Johnson	3.6	1.4	0.8	-2.0	1.0
Roy Oswalt	3.7	0.0	-0.3	-0.7	0.7
Ian Kennedy	1.0	4.3	0.0	-2.9	0.6
Ubaldo Jimenez	4.0	-2.0	-2.8	-0.6	-0.3
* 2013 is prorated

The list is filled guys considered to be big game pitchers or aces, but also a handful of guys who just had one great year, and almost all of them but Clayton Kershaw fluctuate from year to year, not always coming through in adding WPA.  Some, like Justin Verlander and Cole Hamels this year, are considered to be aces in their prime who actually have a negative WPA value for 2013.  Darvish's pro-rated 2.9 is good, but not quite in amazing territory, yet still right in line with the non elite seasons of a guy like Cliff Lee or Felix Hernandez, and way, way better than that of Cole Hamels or Justin Verlander's off year.

WPA shows the overall numbers and adds the context of how it impacts the probability of his team winning, but we can look even further into leverage splits to see specifically how hitters have done against him split out by how meaningful the at bat was:

Yu Darvish Leverage Splits for 2013

The last two columns, tOPS+ and sOPS+ indicate how he's performed compared to his overall numbers (tOPS+) and how's performed compared to the rest of the league in the same situation (sOPS+).  He's absolutely murdered opposing hitters in a laughable way in low leverage (large margin in the score either way) situations, been a bit better than average in medium leverage and pretty good in high leverage.  That seems fairly in line with what I would expect having watched all of Darvish's starts this year, and also resembles some of the comments by folks who feel like he only dominates with a big lead.  He's certainly been much, much better when it's low leverage and still pretty good even when it's not, but not quite dominant.

Digging even further into a specific split, Baseball Reference tracks "Late & Close" defined as the 7th inning or later with the team tied, ahead by one or at least the tying run on deck.

Yu Darvish Late & Close for 2013

Here's a specific item where Darvish is 6% worse than league average (sOPS+) and 37% worse than his overall performance (tOPS+) which probably fuels a lot of the anecdotal memories of folks.  League average is going to include a lot of late inning relievers pitching with the lead, so even though a .699 OPS is pretty good, it's not great late in the game.  Here's some examples of other starters in 2013 in the same situation.

Top 10 MLB Starting PItcher Late & Close for 2013 (min 9 IP, 15 GS), plus selected pitchers

Darvish is well behind a lot of his top pitcher contemporaries, but then again, he's peformed better than Cliff Lee this year in late, close game situations and Cliff Lee is universally considered an ace.  To even be on this list means your manager trusts you enough to let a starter pitch in a late close game enough to accumulate at least 9 innings in this situation.

Looking at all of this data, I feel it's fair and accurate to say that Darvish has pitched extremely well this year, but hasn't been at his best in non-low leverage situations or in late, close games.

But!  Does that mean he truly lacks a Killer Instinct or that he needs to learn how to pitch in those situations?  Now we're hip deep in the "clutch" argument.  Intellectually, I have a hard time believing in clutch.  One issue that that stands out about the nature of looking at clutch statistics is that they are usually very small samples.  Judging Yu Darvish on 11 innings of late and close pitching seems like it's not really worth giving much consideration.  That said, emotionally I sort of want to believe in clutch because it feels like it might exist.  People get nervous or lose focus or have a bad day, but then again scouting types have mentioned many times that every guy at the ML level has already passed the "can you handle it" stage.

On the overall argument of clutch, I'm not 100% in either direction, but lean towards it's usually just small sample noise.  In the specific case of "Is Yu Darvish lacking the clutch gene?" I submit yet another set of small samples for you to consider:

Yu Darvish Leverage Splits for 2012

Yu Darvish Late & Close for 2012

Last year, Yu Darvish was at his best in higher leverage situations and in late and close games.  In both cases he was very strong compared to league average numbers and had a hard time in low leverage games when it didn't matter.  I recall it often being talked about how he elevated his game last year when it really mattered... he just had a hard time with the first couple of innings.  So if he does lack killer instinct or the extra gear, did he have it last year and misplaced it?  Neither sample is very large and I don't know what's in the mind of Yu Darvish this year, but given that in two years he's had one very good high leverage year paired with mediocre low leverage and one mediocre high leverage year paired with mind blowingly good low leverage, I would lean towards this year being more noise than substance when it comes to his actual ability.  Further, I expect his future performances to regress to his overall numbers, as I would expect comparing almost any small sample to a larger one without compelling evidence to argue otherwise.

Yu Darvish is a great pitcher having a great year and I believe one of the best pitchers in the game.  His late game and high leverage results haven't been as good as his low leverage this year, but I wouldn't feel comfortable saying that it's because he's lacking something considering it was a strength last year and we don't have a lot of data to draw a meaningful conclusion.