Last season, one of my favorite baseball reads that became useful fantasy knowledge was this piece by Rich Lederer at Baseball Analysts. What he laid out is something that I’ve recommended and used in previous years as a quick and dirty way to look for potential targets at the end of drafts. If you believe in simple regression to the mean, it makes sense to target pitchers that were well below their personal and/or league average, since logic dictates they should do better the following season. As Lederer put it:
While strikeouts, walks, and home runs play a large part in determining ERA, the latter is also a function of defensive and bullpen support, as well as performance with bases empty vs. runners in scoring position. As a result, the difference between ERA and FIP is almost entirely accounted by strand rate (LOB%*) and batting average on balls in play (BABIP). Each variable has a coefficient correlation of nearly 80 percent with the delta between ERA and FIP.
When you put the two together (LOB% divided by BABIP), the coefficient correlation jumps to 90 percent. Accordingly, the coefficient of determination or R² is 81 percent. In other words, more than four-fifths of the difference between ERA and FIP is due to LOB% and BABIP. As such, in addition to SO, BB, and HR rates, it makes sense to study LOB% and BABIP to understand why a pitcher's ERA may be better or worse than his FIP.
The graph below shows the 138 pitchers who threw at least 100 innings last season. The red lines represent the league averages for BABIP and LOB% while the slanted line is the line of best fit and represents a correlation coefficient of -0.43.
Quadrant I represent the pitchers who had both a BABIP and a LOB% that were better than league average. The table below shows the most extreme members of this group:
|
PITCHER |
2011 BABIP |
2011 LOB% |
CAREER BABIP |
CAREER LOB% |
ERA-FIP |
|
.252 |
83% |
.276 |
77% |
-0.83 |
|
|
.224 |
82% |
.230 |
82% |
-1.52 |
|
|
.237 |
80% |
.285 |
73% |
-0.63 |
|
|
.249 |
80% |
.290 |
72% |
-0.72 |
|
|
.260 |
80% |
.299 |
73% |
-0.63 |
|
|
.245 |
79% |
.285 |
75% |
-1.31 |
|
|
.259 |
78% |
.280 |
77% |
-0.23 |
|
|
.254 |
76% |
.283 |
75% |
-1.11 |
Hellickson’s career numbers mean nothing since his career is nearly entirely comprised of 2011 statistics. That said, he and Weaver are very similar in many ways. Both are mainly fastball/changeup pitchers that induce plenty of swings and misses along with bad contact in the form of infield pop-ups. Largely as a result, Weaver has had a BABIP below the league average for most of his career, and Hellickson has shown a propensity across one season to do that. Hamels is another pitcher who has been able to out-perform the league average several times, and he has been a Quadrant I pitcher for two straight seasons. Also worth noting is that five of the pitchers above also feature one of the best changeups in baseball, which helps them with both swings and misses as well as inducing weak contact.
The ERA-FIP column shows why many seem to be running from Hellickson in drafts this month and why his BABIP has been quite the talking point in recent weeks. If you regress Hellickson’s BABIP to league average using the 3700 balls in play figure both Tom Tango and our own Derek Carty found to be correct, his BABIP comes out to .282. However, if you regress it to the Rays’ team BABIP of .266, Hellickson’s BABIP would have been .260. There is no doubt that Hellickson’s BABIP will climb in 2012, but looking at numbers like that should help lessen the worry about it climbing all of the way back up to the league average (especially since, as an extreme flyball pitcher who induces pop-ups, his BABIP should be expected to be lower to begin with).
The largest ERA-FIP differences in Quandrant I were:
- Jeremy Hellickson: -1.52
- Ricky Romero: -1.31
- Johnny Cueto: -1.11
- Joe Saunders: -1.05
- Jair Jurrjens: -0.99
- Ryan Vogelsong: -0.92
- Jeff Karstens: -0.88
- Jered Weaver: -0.83
- Josh Beckett: -0.72
- Kyle McClellan: -0.70
When pursuing pitchers in Quadrant I, some will come at a discount because people fear regression, but the bigger names often find their way into that quadrant because they are simply that good. Conversely, Quadrant IV present bargains in the form of pitchers coming off rough seasons thanks to worse-than-average BABIP and LOB%. The table below shows the most extreme cases in that area:
|
PITCHER |
2011 BABIP |
2011 LOB% |
CAREER BABIP |
CAREER LOB% |
ERA-FIP |
|
.340 |
64% |
.309 |
72% |
1.67 |
|
|
.332 |
66% |
.295 |
70% |
1.38 |
|
|
.337 |
66% |
.309 |
68% |
1.17 |
|
|
.334 |
66% |
.301 |
73% |
0.92 |
|
|
.324 |
66% |
.298 |
70% |
0.67 |
|
|
.326 |
65% |
.317 |
64% |
0.63 |
|
|
.326 |
65% |
.302 |
71% |
0.63 |
In the case of Reyes, he was essentially pitching to his typical (poor) career averages, but the rest of the pitchers were much worse than their career averages. Lackey is out for the 2011 season, so he is out of consideration, but Lowe changing to a better infield defense, Nolasco moving indoors, and Garcia all present intriguing options to pick up at a discount.
Here are the 12 best ERA-FIP totals in Quadrant IV:
- Jake Peavy: 1.67
- Derek Lowe: 1.38
- Ricky Nolasco: 1.17
- Jon Niese: 1.07
- Brandon Morrow: 1.05
- Ubaldo Jimenez: 1.01
- Ryan Dempster: 0.93
- Brian Duensing: 0.92
- Zack Greinke: 0.88
- Travis Wood: 0.81
- Danny Duffy: 0.78
- Felipe Paulino: 0.75
Thank you for reading
This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.
Subscribe now
Good article.
And it brings to mind the mini-discussion last week from Jay Jaffe's NL rotation-ratings article --- how projected WARP is based largely on pitcher independent outcomes (Fair Run Avg.) and its limitations.
My question: is there a pitcher-rating based on something like FRA that also incorporates these next-level insights into LOB/BABIP, FIP/ERA disparities?
Since I'm guessing not, isn't this a stat calling out to be invented?
Also, I must have missed this. Do we have any indication how the Miami Aquarium will rank in parkish factors?
I especially liked the reference to change-ups after Table 1, as I believe that some pitchers do have an exceptional ability to induce weak contact, and had noticed that many of the exceptions were pitchers with plus change-ups.
In fact, I was surprised that Matt Cain missed your first list, given that he is the poster boy for low BABIP's and beating his FIP's, and of course his meanest offering is el cambio. Then I saw that his 70.6% LOB was too low for inclusion, so I understand the omission.
I appreciate research that detect trends in the collective data set, but the power of such studies often breaks down when applied to individual cases. I contend that "regression to the mean" is often mis-applied, either in the context of league-wide vs player-specific regression, or in the underlying assumption that a player's "True Value" is static. It is necessary to look deeper at individual player context, as you have well-exemplified in this article!
I have been writing an article series on the subject of True Value over at BDD, so this topic is right in my zone of thought these days. Thanks for adding more thoughtful analysis to the discussion, and for addressing some of the underlying issues "Between the Numbers."
When you regress Hellickson's BABIP against the team defense, a .270 doesn't look as bad as the .290-.310 range everyone assumes will be the case. I'd like to dive more into the FB/CH combo to see how that helps because I go back and look at a guy like Tom Browning who had BABIPs below .280 in six of his eight full seasons.
Why don't more people do that? Regress Matt Cain and Jered Weaver's HR/FB to their respective teams instead of to the NL average? Seems to make more sense to me to do it that way.