What in the World is Going on with James Shields?

Here at FanGraphs, it is gospel to say that a pitcher’s ERA is related to both skill and luck. The skill comes from being able to get batters to swing and miss or to induce weak contact, while limiting walks and home runs. The luck comes from how well the other players defend, and also the sequencing of events. That last element of luck, sequencing, merits a brief conversation.

There is little difference, from a pitcher’s skill point of view, between consecutive hits and hits in separate innings. That is to say, a pitcher’s skill is related to how many hard hit balls he gives up; a pitcher’s luck is related to when those hard hit balls occur. So, ERA is affected by the timing of hits, which we can measure easily using LOB%, which is the percentage of base runners that do not score at the end of an inning. It’s not this simple, but basically, a low LOB% rate means the pitcher has been unlucky, and a high LOB% rate means he has been lucky.

The average LOB% in 2015 so far is 72.4%. James Shields‘ LOB% is 87.8%. This is significant! Seven out of eight runners that reach base on him get stranded! His ERA should be anemic, right? Wrong. His ERA is a respectable 3.74, but this is unexpectedly high given what I have told you so far. Clearly, I haven’t told you everything.

There was concern during the offseason that James Shields’s fly ball tendencies would be problematic in the spacious Petco Park with a highly questionable outfield defensively. I guess his home ballpark isn’t spacious enough, because Shields is allowing a career high 2.28 home runs per nine innings, and 25.5% of the fly balls he surrenders leave the ballpark.

Meanwhile, Shields is also striking batters out at a significantly higher rate than his historical numbers indicate he should be. In fact, Shields is striking out batters at a greater rate than any other qualified starting pitcher (and most unqualified ones too!). Opposing hitters are also swinging and missing against Shields more frequently than any other pitcher, even more than highly sophisticated robot and Rust Cohle impersonator Corey Kluber!

When contact is made against Shields, though, it’s been hard contact. According to our new quality of contact statistics, only three starting pitchers have given up a higher percentage of hard contact than Shields. Batters rarely make contact, but paradoxically, when they do make contact, they’re hitting ropes.

This is confusing, and I don’t know why it’s happening, though I can speculate. Shields picked up a knuckle curve a few years ago, and he’s throwing it this year almost a quarter of the time. He’s a good pitcher, and it’s probably a good pitch, which explains the swinging and missing. However, it’s also a new pitch, and he’s probably also making a fair amount of mistakes, which hitters are taking advantage of.

That all made sense to me until a quick PitchF/X search told me that only one of the dingers off Shields were on curveballs. Back to Square 1; I have no idea why this is happening, and it will probably take someone smarter than me to figure it out, or it’s just a sample size issue.

In conclusion, let’s look back at the definition of LOB%. It measures the amount of batters that are left on base when an inning is over. Shields’s high K% probably helps inflate that LOB%. But, it’s also small sample size, and I’m not talking about early season small sample size (although that is probably also a factor). When a high percentage of hits given up are home runs, there are no runners to leave on base in the first place! James Shields is striking out and walking batters, and giving up home runs, all at a career high rate. And it’s kind of working.



Print This Post

Clinical Research Coordinator at University of Cincinnati, but never will surrender my Cleveland Baseball fandom.

newest oldest most voted
Jim S.
Guest
Jim S.

His changeup simply isn’t as good as it used to be, and he’s trying to compensate. He’s not the same guy he was with the Rays.

fyi
Guest
fyi

per the fangraphs glossary:
LOB% = (H+BB+HBP-R)/(H+BB+HBP-(1.4*HR))

basically, it seems like fangraphs assumes all homeruns give 1.4-runs. essentially, if you allow a walk, hit a batter, and strand both runners before giving up a solo homerun the next inning, you have:

LOB% = (1+1+1-1)/(1+1+1-(1.4*1)) = 2/1.6, i.e., over 100%

It looks like Fangraphs has fixed this quirk so that pitchers who would be over 100% are corrected back to 100% (see Will Harris). But it still assumes all hr provide 1.4 runs regardless of how many runners are on base (see Soria who would have a 90% LOB by your method but has a 98% per fangraphs

Mitchell
Guest
Mitchell

Interesting, thanks for pointing that out.

fyi
Guest
fyi

sure thing and i liked the article by the way. but what we’re seeing here is partially shields’s lob-rate being overly-inflated by giving up solo homeruns. shields has given up 16 runs on 14 homeruns. basically, by the fangraphs method (1.4 runs per homer), that’s 19.6 runs, so it’s assuming that he’s stranded an extra four baserunners.

now, there are some ways the methodology makes sense, of course — if you assume that whether an homerun is solo or not is random, by allowing solo homeruns, the pitcher is stranding runners from other innings

as an explanatory hypothesis, shields may be changing his pitch mix with runners on, particularly reducing use of a four-seamer in favor of a two-seamer in order to get grounders. lob-rate is also highly dependent on k-rate as well as babip (fewer balls in play obviously increases strand-rate) so this may also be part of the quirk you found.

hebrew
Member
Member
hebrew

this is strong analysis. good work.