btw, I just found a Chan Ho Park video: [youtube]http://www.youtube.com/watch?v=6GIEHPGj9sI&feature=player_embedded[/youtube]
They don't lie, but they can mislead people into making improper conclusions if you're not applying context. Think about the example I gave you before. If you have 20 starts against one of the best offenses in the league with catcher A, and 20 starts against middling offenses with catcher B behind the plate, which catcher do you think will have better results with the pitcher? So what do you take away from that? The numbers wouldn't be lying to you if you thought the pitcher threw better to Catcher B, but is that the right conclusion? And "what the reasons might be" is at the heart of this, because if the reasons really is that Burnett is worse off with Posada catching him because of some personality conflict, then you can start to seriously consider making a switch and have it be of the most benefit to the team's performance (after you consider other factors like downgrade in offense, etc.). If, however, there are other reasons that make up the bulk of the difference, then you don't make the switch. In this case, there isn't enough evidence to show much of anything. If you're taking the whole picture into account, it's still only about half a season's worth of starts with each catcher, and you also have no idea if it's specific to Burnett. Hell, CC threw three awesome starts in a row with Cervelli back there. Maybe he works better with Cervelli than Posada? And if you start to apply context, it appears that there's not much of a difference.
That's the most extreme example, and of course I'd agree in that case. That's not the case here. Posada has caught against both good and bad teams, and so have Molina and Cervelli. If you want to break it down start by start and figure out who on average has faced better offenses, that's fine, but the difference is not significant enough to account for the disparity in numbers.
I give up. You're continuing to miss the point. Yes, the difference is indeed significant enough to account for most of the disparity in numbers for Burnett and Posada specifically. It is. If I thought it'd be helpful, I'd break it down for you. It will fall well within the boundaries of variance.
