Posted by JFerg on 2/22/2012 11:37:00 AM (view original):
Posted by gjello10 on 2/22/2012 11:26:00 AM (view original):
Posted by jvford on 2/22/2012 11:17:00 AM (view original):
Posted by gjello10 on 2/22/2012 11:03:00 AM (view original):
Posted by jvford on 2/22/2012 11:00:00 AM (view original):
For starters, this is wrong:
"1. Balls not in play (K, BB, HBP, HR) These are almost 100% under the pitcher's control."
Assuming both pitchers faced roughly similar hitters (on aggregate over the course of a long season), then it's not wrong. His conclusions are still wrong, but that statement is correct in the way he meant it.
No it's wrong. Home Runs are not "almost 100% under the pitcher's control." Luck plays a HUGE role in the variability of home runs season to season. Which is why xFIP is used more often than FIP.
Well, this is semantics because I also prefer xFIP to FIP, but random variance over a small sample size is not the same thing as lack of control over the variable. Over a large enough sample, assuming 2 pitchers face similar quality aggregate hitters, a difference in HR/9 is basically entirely down to the pitcher, the same way K/9 and BB/9 are, and in a way that BABIP is not due to fielding.
I will add a few things.
1. It's better to look at K/PA, BB/PA, and HR/FB.
2. Park effects also contribute to a pitcher's HR/FB or HR/9.
3. BABIP does have a skill, but it is smaller than the three true outcomes. A pitcher's defense behind them could make a large impact on that. Just look at the Orioles from the late 1960s and 1970s.
Bobzilla
The key with BABIP is that we all recognize it contains 3 components, both in real life and HBD:
1) Skill of the pitcher (again- ignoring batter skill assuming similar quality batters faced)
2) White Noise due to fielding
3) White Noise due to random variance
The concept behind FIP/xFIP is that the White Noise on BABIP is greater than the value of the skill data. Those Orioles clubs are a great example of the truth behind this in real life. In HBD, I really think you have to live with the White Noise because there is less of it and more skill data contained in the BABIP info than in real life.
Regardng replacement level, one issue is salary demands. At the end of Spring Training, unsigned ML players fall into 2 basic categories: (a) guys who will sign for something close to the minimum (say, $600k or less against a min around $425k), and the occassional aging star who would rather sit on his couch than play unless offerred a really nice opportunity (Roy Oswalt may end up in this boat, as he seems intent on either getting $7.5m+ to play for a contender, or sitting at home and playing with his kids). There's not a lot in between.
In HBD, at the end of Spring Training, there's not a lot out there for near the minimum ($325k). But there is often a bunch of league-average-ish talent to be had in the $750k-2.5m range. There are also some above average Type A guys out there at reasonable prices who just don't get signed because of the comp pick, which doesn't happen in MLB for a variety of reasons. So what is a Replacement Player in HBD? How can we define that?
Along these lines, by the way, I should note that the salary difference between Sheldon and Ducey is large enough for Sheldon's owner to easily pay for those 81 innings at league average rates, I would think.
Another major replacement level issue is that I would suggest that overall League Average for SPs differs from League Average for RPs more than in real life. Good RP innings are fairly easy to find, and the best and most durable RPs can log nearly 200 innings. I think it would be pretty tough to look at replacement level without seperating SP and RP numbers.