Saturday, January 01, 2005

2005 Dodgers: Kazuhisa Ishii

So I've spent countless hours chasing Bluto's rainbows, trying to figure out this riddle wrapped in an enigma doused in maple syrup and topped with pebbles, chocolate sprinkles, and mud gravy. I've mixed more metaphors than a bartender at a poetry school in my aborted attempts to come up with some cogent and coherent analysis.

What's so maddening about him isn't that he's inconsistent. It's that he's just so darned consistent in his inconsistency. One day he's Senator Blutarsky and the next he stuffs his mouth full of cafeteria food and prepares his hands to expel said food all over the cafeteria, inciting a full-scale food-riot. There's not a lot of in-between, and there aren't any stretches where he gets his act together or stretches where he falls apart so much that you can isolate any trends. He's like Neil Young from 1980-1986. Ishii will put out an album with the promise of Little Wing and The Old Homestead, but that album's also going to have insufferable tunes like Lost in Space and Stayin' Power on it. And unlike with Young, there's no reason to expect that someday Ishii will release his Freedom album, full of solid games/songs and a couple of masterpieces. But, as you'll see, there may be reason to expect a relative gem like Young's Life.

This overwhelming inconsistency isn't just the way it feels; this is the data. Over almost any five game stretch in 2004, Ishii had at least two quality starts (6+ IP, 3 runs or fewer allowed) and at most three, with only two exceptions when he had five-game stretches with only one quality start. Even using the modified "good start" 6 IP/2 R standard, only creates one additional three-start stretch where he'd only mustered one good start in his last five. If you look at his FIP's (a stat that mimics ERA based on walks, strikeouts, and home runs and which correlates with future ERA better than actual ERA) and look at each five-start period (1-5, 6-10, etc), his low is 4.33 and his high is 5.68. That's not a substantial amount of variation. But it's not being achieved by consistent mediocrity; it's being achieved by bouncing from good to bad to terrible, and it's not much fun for those of us who have to watch it.

What's more, there's not something obvious like Ishii's rest pattern to finger the blame at. I broke down Ishii's starts by the amount of pitches he'd thrown in his previous outing. There was a slight pattern toward improvement based on the length of the previous outing:

101+ pitches: 5.84 ERA, .33 "Good Start" percentage
91-100: 4.94, .38
89-90: 4.23, .4
80 or fewer: 4.14, .5

As promising as that looks to support the notion that pitch count caps could solve this problem, two facts mitigate that: first, that looks more like ERA randomness, as his FIP in those four categories of starts looks different: 6.27, 4.60, 6.27, 4.58. Now, maybe you're the type of person who sees that and thinks "Okay, we can let him pitch up to 80, but once he gets to 80 he's got to get up to 91 but then stop after 100." I, on the other hand, think that indicates small sample size randomness, although a cap at 100 might not be a terrible idea. Second, Ishii's value largely derives from the fact that when he's on, he's on; his splits indicate that he's the rare pitcher who seems to improve the longer he's in the game, although that's more of a selective sampling issue than anything-- he lasts longer on those days when he's good, and on the days when he's bad it's evident from the start.

Another thing I discovered was that Ishii performed poorly on five days rest:

4 days rest: 4.44 ERA, 4.73 FIP, .44 "Good Start" percentage
5 days rest: 6.43 ERA, 7.77 FIP, .29 gs%
6+ days rest: 4.23 ERA, 4.43 FIP, .4 gs%

My intuition says that's small sample size foolery, but if I were working for the Dodgers I would look into either altering his between starts routine when he's working on five days rest or just having him be the guy whose slot in the rotation is skipped when there's a day off. I might check into these splits for his past seasons, but I haven't done so yet. This isn't huge, but there's at least some reason to believe this is an issue, and if it is that would be helpful as this is something that's probably correctable.

But let's dig into how Kaz' performance is distributed. Ishii is like a vision out of Michael Wolverton's dreams, at once alarming, terrifying, and beautiful. His good starts are truly good, and his bad starts are truly bad. According to the support-neutral stats at Baseball Prospectus, Ishii's pitching was such so that the average team could have expected to win 15.3 games and lose 15.7 games in Ishii's 31 starts. That means that, on balance, Ishii's performance was only a hair below average, and his schizophrenia is precisely the kind that tends to get undervalued just by looking at a player's rate stats. That's the good news.

The bad news is that Ishii's 2004 rate of run allowance versus his component/peripheral stats is at the outer bounds, and it's somewhat unrealistic to expect him to continue being as lucky on balls in play and the timing of the hits he's allowed as he has been. (I should note here that support-neutral stats, in their current incarnation, are based on runs allowed and runners bequeathed, not component or peripheral stats). That's not to say he's due for some sort of crash, but he'll need to actually improve his performance to remain in the neighborhood of average. While he has traditionally been better than average on BABIP and runs allowed versus runs created, if he repeats his 5.35 FIP chances are his ERA will reside north of 5.00, rather than cooling off down at 4.71. Using the DIPS and DIP% data at ESPN.com (it's not on the player cards; you have to look at the entire league stats to get it), Ishii's DIP% - the ratio of his Defense Independent Pitching to his actual ERA - has resided in the lofty 1.16-1.24 range in each season of his career. In 2002 and 2003, he outperformed his DIPS mainly by good timing, as his component ERA (ERC) to ERA ratio was in the same range. In 2004, however, his ERA was actually higher than his ERC, so the relatively low ERA was accomplished through some permutation of strong defense and his own ability to induce batters to hit balls in a way that would be less conducive to them becoming hits. This is perhaps becoming more opaque than I intended, so let me take a moment to summarize: in the past, he's outperformed his peripherals two different ways, but he hasn't demonstrated consistency in the way he's outperformed his peripherals, so we can rule out the likelihood of improvement vis-a-vis his peripherals and Bill James' famous "plexiglass principle" indicates that the stronger likelihood is that he'll soon underperform his peripherals.

Now, you might be thinking that Ishii could maintain his relative success with balls in play because a) the Dodgers have excellent defense and b) he changed his approach to get more ball in play outs in 2004 than in the past. Well, both points have evidence to support them, and I'll be getting more into the nuts and bolts of balls in play and the Dodgers defense pretty soon, so stay tuned. Did Ishii change his approach to take advantage of the Dodgers' D? Well, the optimistic narrative is that yes, he did. His pitches per plate appearance was down from 3.91 (2002) and 4.07 (2003) to 3.71, and whereas he walked 15.4% of batters faced in 2002 and 2003 he walked only 13.1% of his opponents in 2004. His groundball to fly ball ratio also dropped substantially, which contrary to popular perception means fewer hits on balls in play unless the pitcher in question has a real nice sinker (Ishii does not). Obviously, though, that comes at a cost, as he used to strike out nearly a batter per inning and in 2004 his K rate was awful. Nonetheless, it can be said that the apparent new approach worked to an extent, as his component ERA has dropped from 4.94 to 4.79 to 4.50.

There is, however, a pessimistic counter-narrative that looks like this: Ishii's stuff is declining, and batters started to get more aggressive on him but he was bailed out by a combination of luck and good defense. This counter-narrative would suggest that Ishii is due for further decline in the future, as one can't really regain those absent K's and that luck won't hold out forever. Furthermore, it could be argued that Ishii's ability to exploit the strength of the Dodgers' defense is far from unique, so this skill doesn't hold weight in comparison to potential replacements.

Which story is the right one? I have no idea. My first guess was that it's somewhere in the middle. Since, I've come around to the optimistic narrative a bit, and the reason why is Ishii's platoon splits. In 2002-2003, his left-handedness was fully evident, as his defense-independent performance would indicate a 3.23/5.40 FIP split, with a .291 BABIP from RHB and a .288 BABIP from LHB. In 2004, however, he had a reverse platoon split, with those FIP figures at 5.46 for LHB and 5.30 for RHB. His performance against RHB actually significantly improved because they managed a paltry .254 BABIP. Against southpaws, however, the BABIP saw no drop (.289) while the walks increased and the strikeouts disappeared; while his K:BB against RHB only fell from 1.10 to 0.99, against LHB it was cut in three: 3.03 in 2002-2003, 1.07 in 2004. Yikes.

Now, a starting pitcher's splits present two major sampling issues, the first being sample size and the second being selective sampling (they'll typically tend to face only left-handed hitters who are good enough to start against southpaws). That's an important caveat, but I can't help but look at the data and come to this conclusion: Ishii (or Jim Colborn or some combination of the two) decided to pitch to contact more in 2004, and this strategy was beneficial vis-a-vis right handed batters but disastrous against left-handed batters. The data supports this, and it has the ring of intuitive sense to it, since to the best of my knowledge the platoon advantage rests mainly in the ability to disguise the type of pitch and the pitch's location from the batter, something a pitch-to-contact strategy doesn't access.

Thus, my humble recommendation for building a better Ishii is for Kaz to maintain his apparent 2004 approach of trying to induce flyballs against right-handed batters while reverting to his previous strikeoutcentric philosophy against left-handed batters. What might Kaz look like if he did that? If Ishii faces left-handed batters 23% of the time (his 3-year average) and does what he did against southpaws in 02-03 while doing what he did against RHB in 2004, his ERC comes out to about 3.87. That's a) not fantastic given the defense it assumes as well as the hitting environment it's derived from and b) a high-end estimate since it relies on selective sampling. At the same time, I don't think it's overly optimistic to argue that, if indeed Ishii's performance in 2004 reflected the change in approach I've observed, Ishii could be slightly above average ERA-wise if he goes back to looking for the K when he faces soupaws. In concert with the belief that ERA somewhat underestimates his value due to the inconsistency of his outings as the support-neutral stats show, it's possible that Ishii could, overall, be solidly above average.

When the Dodgers give a contract to their 2005 pitching coach, could someone have him drop me an email?

(note: edited back to initial posting; the slight data error I "fixed" in the final paragraph was actually not an error)

Comments:
So would you keep Ishii or trade him? Is he over-paid? How much is he worth?
right now the rotation is Penny, weaver, ishii, jackson, dessens. let's say we get Lowe and halsey. do you still keep ishii or do you trade him off for whatever you can get?
 
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?