Saturday, February 13, 2010

SIERA Watch: Day 5 (FIP and xFIP)

Above, we have SIERA vs FIP and xFIP. The line is approximately 45 degrees. In other words, the line represents a 1-to-1 equality. If SIERA and FIP/xFIP were by-in-large similar (which wouldn't necessarily be bad), then they would follow along this line. We can see, particularly in the higher numbers, how xFIP tends to be a touch lower than SIERA. It's hard to say, just yet, what this means.

Anyway, let's continue profiling our beloved Cubs and see if this coincides with our other knowledge.

First: I've added xFIP to the master dataset. Instead of two versions, I'm just going to link a .csv version, so even STATA users can play right away:

SIERA Test, with xFIP (csv)

If we constrict the data to 40 IPs, we get these summary stats:

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
       siera |       385    4.183974    .7120269       2.04       6.46
         fip |       385    4.302961    .8680156       2.03       7.11
        xfip |       385    4.035429    .6244117       1.94       6.31
         era |       385     4.23361    1.245259       1.73       9.64

Unsurprisingly, ERA is the most volatile (st. dev. 1.25, min. 1.73, max. 9.64). This is unsurprising because ERA has elements of defense, ballpark-factor, error-opinion, etc. -- all swirling around wildly. In turn, xFIP and SIERA are pretty steady, which makes sense: both are regression-based and meant to normalize numbers according to our common expectations.

Anyway: On Day 1, we observed that these pitchers didn't really gain or lose too much, in terms of xFIP, from SIERA:

The Middle Pack (SIERA difference between -0.49 and 0)
Rich Harden: SIERA 3.24, xFIP 3.70
Tom Gorzelanny: 3.41, 3.64
Ryan Dempster: 3.79, 3.81
Carlos Zambrano: 4.11, 4.27
Angel Guzman: 4.13, 4.19
Aaron Heilman: 4.16, 4.20
David Patton: 4.72, 4.78
John Grabow: 4.93, 4.96
Jeff Samardzija: 5.10, 5.16
Kevin Hart: 6.12, 6.20

Do any of these pitchers profile as someone who SIERA should reward or punish?
Rich Harden receives a pretty nice bonus (SIERA = xFIP - 0.46), but this seems appropriate given his above average SO rate (10.9 K/9) and and super high HR rate (1.47 HR/9; career HR/9 = 0.84!).

Tom Gorzelanny also got a nice boost (-0.23), which is okay, given his 9.00 K/9 and 40% GB rate, but I'm a little iffy on this one.

Carlos Zambrano, I'm thinking, should have been punished for an unrealistic 0.5 HR/9 and his signature, not-so-great walk rate, 4.15 BB/9. Should Zambrano be so near to the mean SIERA? I don't know. I think it would make sense if it was a little higher. He's got an above-median xFIP, I expect at least the same with SIERA.

David Pattion had HUGE walk rates and HUGE ground ball rates. Result? They pretty much cancel each other out.

John Grabow makes me wonder the same thing I did about Big Z: high walks, high HRs -- shouldn't his SIERA be higher than his xFIP? I guess his GB rate must nullify that effect (GB 43.3%).

Jeff Samardzija had average BB rates, high HR rates, yet SIERA leaves him over a full st. dev. away from mean... Hmmm...
Anyway, that's what I've found. Let me know if I missed someone or thing. Next, we'll look at the SIERA losers.

1 comment:

  1. LOVE how this shows the volatility of ERA...gotta love our friend "standard deviation".

    ReplyDelete