- FanGraphs Baseball - http://www.fangraphs.com/blogs -

Complete Game James: James Shields Unlearns Us

Education consists mainly of what we have unlearned.

-Mark Twain

In 2010, James Shields led the league in hits allowed, home runs allowed, and earned runs allowed. Many among the Rays faithful protested last year when Joe Maddon tabbed Shieldsy to start game two of the ALDS, wherein he failed to complete five innings.

“He’s broken,” said some. “He shouldn’t even be on the playoff roster!” fumed others.

He finished the 2010 season somehow once again topping 200 innings, but his ERA was over 5.00, his FIP was above 4.20, and his fans were frustrated. Despite his career-best 3.55 xFIP, a career-low LOB%, and career-high BABIP (a whopping .341), many — even among the sabermetric-slanted — doubted he was merely a product of bad luck.

Well, in 2011, his statistics have sung, “Cy Young!” all season long: 2.96 ERA, 3.36 FIP, and 3.11 xFIP. And most impressively: a league leading 10 complete games and 4 shut outs.

For a guy who has pitched 200+ innings in 5 straight seasons, he sure never showed a knack for complete games before:

He went two years without a single complete game, and then BAM! suddenly he has 10 in one season (at least; he starts again tonight). How’d he do it?! Is it just a luck swing-around?! Did he change his approach?! Is his Bradley Woodrum-esque quasi-Amish beard to blame?

The answer: Yes to all.

Firstly: Let’s give Shields some proper credit for 2010. He may have had a ghastly ERA and a below average FIP (105 FIP-), but the dude had the major luck components — BABIP and LOB% — going in the nasty direction from his career norm.

Despite that, Shieldsy still managed an above-average start length:


NOTE: The decimals represent percentages of the innings, not outs. That is: 6.2 means 6 and 20%, not 6 and 2/3 innings.

The average MLB start length since 2004 is around 5.9 innings, and this year it’s 6.1. What with his absurd 10 complete games, Shields is now flirting with 7.5 innings pitched per game. However, his previous seasons have all been above average as well — he certainly doesn’t Oops! his way in 200 innings year-after-year.

But 2011 is clearly different. Not only is he pitching like a madman, he’s pitching like a different man, giving hitters more curves and changes than they can handle:

James is throwing his fastball — his most pedestrian pitch — and his cutter much less, choosing to pump his knee-buckling curve and Jose Bautistainfuriating change-up instead. This is the right choice, one must think.

Holding game theory constant (in other words, assuming the players don’t adjust to his new approach and then suddenly learn how to hit his curve and change), it makes sense that Shieldsy would want to use his best two pitches more.

So does this explain it? More good stuff, less feh stuff equals New James Shields? Well, not quite. The truth is — though his change-up has always been a doozy and his curve has usually been dominant — they did him few favors last year:

He could have thrown more changes in 2010, but he couldn’t throw only changes. None of his other pitches were apparently working, but as a starter, throwing one pitch is akin to watching Corky Ramano: There’s just no hope of survival. So pitch selection has been a part of Shields’ newfound awesomeness, but it alone cannot explain everything.

In 2010, only his change was working; there are two possible explanations for that:

1) Location: Fans roundly critiqued Shields for placing fastballs, cutters, and curves in the heart of the strike zone last year. A quick look at his heat maps, however, does not seem to provide the evidence that ought to be pretty damning. So, then, we must consider…

2) Bad luck: Yeah, those pitch values have a pretty bad shot at being positive when over 33% of the balls in play go for hits and nearly 14% of each fly ball goes for a home run.

This present James Shields is a product of several factors: (a) A good, new approach, (2) superior facial hair, and (d) a LOB% 10 colossal percentage points higher than last year (68.4% vs. 78.4%) and a BABIP 22% lower than last year (.341 vs. .267; sweet Moses, that’s a big swing!).

So what have we unlearned here? Simply this: There is always a perfectly logical rationalization or narrative for an unlucky or extraordinary season — but that doesn’t make it worthy or capable of forecasting.

Joe Maddon was right to start Shields in the ALDS; he rightly knew that all the predictive statistics Shields had were pointing up. Now, they’re pointing kind of obliquely downward, but in the meantime, Shieldsy fans: Enjoy the ride; it may never get better than this.