Posted by JFerg on 2/21/2012 11:06:00 PM (view original):
Posted by MikeT23 on 2/21/2012 10:57:00 PM (view original):
There are certain numbers to use that we know work within the HBD world. And there are numbers that have been brought into the conversation that may or may not have any place within the HBD world.
This:
I'm sure NONE of these people build their HBD while concentrating solely on Control, Velocity and GB/FB even though those EXACT pitchers would have the best FIP.
Does ERA and IP count as stats that work within the HBD world? Does comparing a pitcher's actual earned runs allowed to what an average pitcher would have done in the same playing time work within the HBD world? Does using a replacement level work? Park effects? 'Cause that's all you need to make a case that the wrong guy won the award. I just want to know at what point does that cross the line and no longer work.
Bobzilla
No, replacement level works differently here, and replacement level is WAR, an Advanced Metric. Believe me, I'm a Saber-Geek. I understand the real-life magic of WAR. But the shape of the talent pool is different here in many ways (example- the way Free Agents go about making contract demands), and the formula for valuing WAR, if one were to create one for HBD, would certainly give less weight to IP than in real life, because the constraints on bullpens here are far more sequential (needing to have enough guys to have a few available for all 162 games) than cumulative (needing to have enough guys to eat all the innings you will need your bullpen to throw) compared to real life. Also, the talent acquisition markets are less efficient here because (a) some owners are really bad at this game in ways no ML GM is, (b) you don't get fired if you lose for a bunch of years in a rebuilding effort, so some owners take a long-term approach that you just don't/can't see in real life, and (c) prospect development is a lot more certain than in real life. As a consequence, opportunities to acquire league-average or better players at very reasonable prices abound.
In other words, those 81 IP are worth far less here than they would be in real life, for a wide variety of reasons.
Also, you have not once mentioned OAV as a stat (a stat which works against Ducey) in part because, in real life, you've been trained (correctly) to filter it out because it's heavily loaded with fielding-dependant white noise and random variation, via BABIP. Here, it's a whole different ballgame. There's still white noise in BABIP, of course, but there's a lot less of it, and a lot more actual data, such that ignoring it doesn't make the picture clearer, as it does in real life. It makes the picture cloudier, because you're losing vital data without which an informed opinion cannot be rendered.