Investigating The Hyun-Jin Ryu Rest Effect

As Hyun-jin Ryu returns from a strained gluteus muscle, he will be an integral part of the Dodgers’ success in 2014. He pitched extremely well on Sunday, so everything looks positive so far. Ryu remains one of the most unheralded good starters in the game. His FIP- this year is 75, the 12th best in all of baseball. The two pitchers behind him on that leaderboard are Adam Wainwright and David Price. His adjusted FIP is better than Zack Greinke and Madison Bumgarner. His FIP (both raw and adjusted) is better than the FIP Clayton Kershaw produced in 2012. Obviously, Ryu’s ERA doesn’t quite match up with his FIP this year, but it’s still above average and is disproportionately inflated by two disaster starts. You’d expect his FIP to be a bit higher than his ERA long term, since he controls the running game so well, or a neutral gap at worst if you think he’s easier to get hits off of than an average lefty.

One thing that is said often about Ryu, especially during broadcasts, is that he pitches better with long rest than he does with four days of rest. It makes intuitive sense, since the pitching schedule in Korea involves longer gaps between starts. His W-L record is usually the first thing cited, which we know is irrelevant. But looking beyond that, Ryu has a 3.52 ERA while pitching with four days of rest and 2.71 ERA while pitching with five or more days of rest through his young career. There’s some truth to this. Pitching well on regular rest is especially important since Ryu will need to do so during the playoffs.

The samples are small-ish (153-2/3 innings on regular rest, 182-2/3 on long rest), but enough to have some sort of significance. Even so, it’s worth looking into to see how real the effect is. Even if the ERA is higher, ERA isn’t exactly a perfect statistic, even in samples of moderate size. Peripherals are important to look at as well. We can also use Pitch FX data to look and see if there are any identifiable causes to the splits or if the difference is just noise.

2013 rest splits

Rest	IP	ERA	FIP	K%	BB%	K-BB%	HR/9	HR/FB	xFIP
2013, 4 days	88-1/3	3.26	3.81	20.06%	7.24%	12.81%	1.02	11.36%	3.69
2013, 5+ days	103-2/3	2.78	2.76	19.34%	5.42%	13.92%	0.43	5.88%	3.25

In 2013, Ryu was pretty definitively better on longer rest, matching the perception. His strikeout rate was actually better with shorter rest, but that’s where the upside ended. He walked more batters on regular rest, leading to a K%-BB% of about one percent lower than it was on long rest. However, the biggest difference was home runs. Ryu allowed home runs at about double the rate on regular rest. The ERA difference between splits was about half a run, but the FIP difference was a full run. The difference in xFIP was pretty close to the same as the ERA difference, but the base was higher since Ryu’s average HR/FB ratio was lower than the league average.

2014 rest splits

Rest	IP	ERA	FIP	K%	BB%	K-BB%	HR/9	HR/FB	xFIP
2014, 4 days	65-1/3	3.86	2.87	21.38%	3.62%	17.75%	0.69	8.20%	3.04
2014, 5+ days	79	2.71	2.66	21.94%	5.33%	16.61%	0.34	4.17%	3.18

This is where things get a bit interesting. The ERA difference has actually widened between splits, now over a full run. However, many of Ryu’s peripherals are better on regular rest this season. The strikeout rate is nearly identical, but Ryu is walking significantly fewer batters on regular rest. Ryu’s walk rate on four days of rest in 2014 is almost precisely half of what it was last year. Ryu’s K%-BB% is better than it was in 2013 across the board, but the regular rest split is two percentage points better than the long rest split.

Putting the peripherals together counters the big ERA split. Ryu’s xFIP is actually a bit worse on long rest this season. Like 2013, the biggest difference between the splits is home run rate, which is almost double on four days of rest once again. Even after including the home runs into regular FIP, the gap between the two stats is 0.2, about 80% closer than the ERAs.

Summing everything up, it seems clear that Ryu’s strikeout numbers aren’t really impacted by rest. His walk numbers were worse on regular rest, but now they’re better. The biggest difference is that Ryu seems more hittable on regular rest. This is evidenced by the higher home run rate in both seasons and the big ERA-FIP gap this year. Even this isn’t consistent, since Ryu had a lower BABIP on regular rest in 2013 and his FIP was lower than his ERA in the same split, but it’s still worth investigating further.

Causes of the split

The issue with saying that batters hit Ryu better on regular rest is that the hittability statistics (ERA/FIP gap, HR/FB ratio, etc) take an extremely long time to stabilize. This could very well be just noise, but a check of Pitch FX data can get us a more specific idea of how a shorter layoff impacts Ryu. First, let’s check velocities and location:

	Fastball Velocity	% Grooved Total
2013, 4 days rest	91.28	3.87
2013, 5+ days rest	91.09	5.14
2014, 4 days rest	91.44	3.96
2014, 5+ days rest	91.79	4.61

Ryu’s fastball velocity isn’t really impacted by rest duration. Last year, it was faster with regular rest by 0.2 MPH. This year, it’s a bit slower. Average both seasons out, and the magnitude of velocity difference is 0.01 (91.35 on regular rest, 91.36 on long rest). Clearly, this isn’t a factor.

Location is a bit trickier to check. In the table, I have an overall percentage of “grooved” pitches per Brooks Baseball, defined as pitches thrown in the middle third horizontally and vertically in the strike zone (if you split the strike zone into a 3×3 grid, it’s the center area). In both 2013 and 2014, Ryu throws more grooved pitches on long rest; about one extra pitch per start. It isn’t universal between pitches, but there isn’t a pattern to which pitches were more or less grooved than others and the magnitudes were all small. Again, this seems like something which would not impact the ERA gap.

I also checked for average vertical pitch location to see if Ryu tended to leave his pitches up more often on regular rest, accounting for the home run rate spike. That isn’t the case, either. The average location of all pitches were actually slightly lower on regular rest. Much like the groove rate, the magnitudes are small enough to be insignificant (about 0.1 inch gap per pitch).

Pitch break, particularly in the vertical direction, changed a bit more. Ryu’s curveball is breaking about an inch less on regular rest this year, but he hasn’t allowed a home run on a curve this year. Most of Ryu’s home runs in 2014 are on his fastball, which isn’t really effected by rest duration. His change is the other pitch which occasionally gets taken deep (3 of 8 home runs this year, 2 on regular rest). However, the changeup “drop” (the difference between the vertical break of fastball and changeup) is not statistically different between the two splits (95% confidence).

So, a glance at the Pitch FX data doesn’t reveal any cause of the increased home run rate on the starts with shorter rest. This just seems like a case of data arbitrarily clumping into one side of the sample. And that’s where this falls apart. Yes, Ryu has allowed a better ERA on short rest than on long rest, both in 2014 and through the course of his career. Some peripherals are actually better on regular rest, and the ones more susceptible to small sample variation are the ones which are higher on short rest. When looking into specific pitch information, there isn’t any data backing the split, either. There just isn’t any evidence that Ryu will be worse on four days of rest going forward.