Regression to the Mean

I have read several posts by various authors that all say the same thing - pitchers with either too high or too low BABIP (Batting Average of Balls In Play) will 'regress to the mean', meaning they'll get better or worse. I've never seen a single one of these authors try to regress a pitcher or two and alter their stats - or rerank the whole league based on this "regression." There are other stats that account for fielding (FIP) or ballpark factors (ERA+) but none to my knowledge that even out BABIP. Keep in mind - major league pitchers are not all equal. They might appear that way, but it's mostly because the talent level is very high amongst all players (except for Bobby Ayala).

In most fantasy leagues, WHIP and ERA are stats to be concerned about. In these fits below, I took stats through May 7th and isolated just the starters (6+ GS).

The fit is close to perfect - It could be totally perfect if we substituted HR/9 for HR/FB (we have to add the HRs back in to the subtracted factor missing from BABIP) but then we wouldn't be able to 'regress' that stat.

ERA is trickier to model - at this point there are still some statistical outliers (yay Cliff Lee) that throw the fit off. Even excluding these guys, the predicted values can get out of control. In this fit I've added GB/FB.

The ML BABIP so far this year is .292, and the HR/FB % is 9.6. Here are your adjusted leaderboards:

	WHIP	WHIPf	WHIPm
Cliff Lee	0.604	0.611	0.934
Roy Halladay	0.982	1.012	1.093
Dan Haren	0.992	0.977	1.094
Johan Santana	1.014	1.024	1.102
Tim Hudson	1.047	1.039	1.160
Jesse Litsch	1.35	1.346	1.177
Brett Myers	1.364	1.315	1.180
John Danks	1.038	1.036	1.186
Jake Peavy	1.027	1.029	1.196

* WHIPf - model fit; WHIPm model adjusted to league average

Name	ERA	ERAf	ERAm
Edinson Volquez	1.06	1.23	2.38
Johan Santana	2.91	2.81	2.51
Tim Lincecum	1.49	2.10	2.66
Roy Oswalt	5.33	5.08	2.72
Felix Hernandez	3.04	3.55	2.89
Jair Jurrjens	2.84	1.98	2.96
Jake Peavy	2.22	2.41	2.97
Cliff Lee	0.81	1.00	2.97
Brandon Webb	2.49	2.28	3.11

* ERAf - model fit; ERAm model adjusted to league average

Cliff Lee is legitimately one of the top-10 pitchers in MLB, probably top-5. Jurrjens is a surprise - but the ERA formula doesn't look to fit him well.

		WHIPd
1	Cliff Lee	54.6%
2	Ryan Dempster	35.6%
3	Ben Sheets	34.8%
4	Shaun Marcum	34.0%
5	Adam Wainwright	24.6%
6	Tim Wakefield	19.5%
7	Brandon Webb	19.5%
8	Zack Greinke	17.3%
9	Jake Peavy	16.4%
10	Ian Snell	-16.3%
58	Ian Snell	-16.3%
59	Clay Buchholz	-16.8%
60	Mark Redman	-17.3%
61	Kevin Millwood	-17.6%
62	Roy Oswalt	-18.9%
63	Matt Chico	-22.4%
64	Manny Parra	-22.6%
65	C.C. Sabathia	-25.4%
66	Andrew Miller	-31.5%
67	Bronson Arroyo	-33.9%

		ERAp	ERAe
1	Cliff Lee	266.5%	0.23
2	Edinson Volquez	124.5%	0.16
3	Carlos Zambrano	111.3%	0.72
4	Zack Greinke	111.2%	0.69
5	Tim Lincecum	78.3%	0.41
6	Ben Sheets	63.7%	0.11
7	Adam Wainwright	52.3%	0.09
8	Tim Wakefield	50.2%	0.28
9	Ryan Dempster	50.1%	0.06
10	Vicente Padilla	39.9%	0.48
58	Jered Weaver	-22.7%	-0.06
59	Clay Buchholz	-23.5%	-0.08
60	Brett Myers	-23.7%	0.11
61	Mark Buehrle	-31.3%	-0.05
62	Matt Chico	-37.3%	-0.05
63	Mark Redman	-43.7%	-0.26
64	Andrew Miller	-44.4%	-0.06
65	Roy Oswalt	-49.0%	-0.05
66	C.C. Sabathia	-49.3%	-0.09
67	Bronson Arroyo	-56.3%	0.04

* Bold indicates the error might be pretty high in that prediction.

The most artificially inflated, and deflated, numbers in the game. Negative means you're playing better than ERA and WHIP indicate, and it looks like Bronson Arroyo is the poster child of screwedism thus far (due for the biggest correction in a good way). I'm pretty sure I haven't been reading any fantasy blogs telling you to go pick that dude up.