How On Earth Did Buzz Capra Win the ERA Crown?

If the sabermetric movement could be condensed into one sentence it would be this: We have better ways to analyze the performance of players and teams, and better ways to predict their future value, than the traditional statistics.

New research demonstrates how nearly useless one year of batting average and ERA are in revealing a player's actual ability. Bill Petti of Beyond the Box Score looked at every player year in baseball history to determine how well one year's number correlate with the following year's. In other words, if Braves rookie first baseman Freddie Freeman hits .291/.352/.461 this year, what are the odds that he's roughly a .291/.352/.461 hitter, and not a .261/.312/.398 hitter getting good breaks or a .325/.398/.517 hitter getting bad ones?

The answer is that batting average correlates pretty poorly from year to year, largely because BABIP, which is a key component of batting average, correlates even worse from year to year. On base percentage, which combines batting average and walk rate, correlates about 50% better year-to-year than batting average because walk rate doesn't change much year-to-year. Slugging and OPS are about the same as OBP when it comes to year-to-year consistency.

In other words, Freddie may or may not be a .291 hitter, but whatever he hits, add about 60 points to get his on-base percentage and about 170 points to get his slugging percentage. (A .354 BABIP and a 3-1 K/BB rate suggest regression next year, though he's just a rookie, so natural improvement may  offset some of that decline.)


What correlates well year to year? Contact rate, patience metrics, strikeout rate and walks are the best. For what it's worth, line drive rate is the worst. Here's the whole chart:

.	Hitter Metric	Year to Year Correlation
.	Contact %	0.90
.	SwStr %	0.89
.	Swing %	0.84
.	K%	0.84
.	Z-Swing %	0.83
.	O-Contact %	0.81
.	Z-Contact %	0.80
.	BB%	0.78
.	BUH	0.77
.	GB/FB	0.77
.	GB%	0.76
.	O-Swing %	0.75
.	ISO	0.73
.	HR/FB	0.73
.	FB%	0.73
.	SLG	0.63
.	OPS	0.63
.	OBP	0.62
.	wOBA	0.61
.	IFH	0.59
.	IFFB%	0.56
.	F-Strike %	0.56
.	Zone %	0.52
.	IFH%	0.44
.	Batting Average	0.41
.	BABIP	0.35
.	BUH%	0.24
.	LD%	0.22

It's worth repeating what Petti says in the article, that ERA has about the same correlation as batting average, which is why seamheads are working to derive fielding-independent statistics for pitchers. As it turns out, even in their relative infancy, those measures have a much better year-to-year correlation than ERA.

It explains how Red Sock Bill Mueller could lead the league in hitting in 2003 (.326) after batting .260 the year before and before hitting .283 the year after. And how Buzz Capra could lead the league in ERA for the Braves in 1974 (2.28) the year after sporting a 3.86 ERA and before posting a 4.25. (Capra started just 61 games in his seven-year career and, other than that one season, was a below-replacement level pitcher.)

All this can get ridiculously complicated, but even if you don't understand the underlying methodology you can appreciate the information. If you read that Jair Jurrjens has a 2.96 ERA so far this year, but a 3.96 FIP, that means he's been a 3.96 ERA pitcher with great fielding, relief support and/or luck and can be expected to tail off the rest of the way. (Of course, in Jurrjens' particular case, he's laid up, so no regression for him right now.)

This is useful data to general managers, fantasy owners, baseball obsessives and fans, in that order, and that is the order in which it has been embraced.
b