H1-H3 don't fully replicate all the conditions of being on national television against elite teams late in the season. They isolate those factors one by one, but perhaps in combination it's too much for him to handle.
if you believe that part of the reason we acquired Beckett & Schilling was their postseason dominance in much smaller, and about the same sample size respectively as Shields before they joined the Sox... and that these were good decisions b/c their continued dominance with us was "real" and not just continued luck, then that argument cuts both ways.
Same thing for Lester. If you think there is value to his strong postseason track record in only slightly larger sample size, then the same logic cuts both ways.
Also, I agree that there are good reasons to believe that a random string of 11 consecutive games within a season might not be a big enough sample size, due to serial autocorrelation caused by an extraneous factor (mechanics temporarily out of wack, injury, etc). But, if you take 11 games selected from three different seasons, then the cold/hot streak is much less of a statistical issue. I guess an analogy to a hitter would be that taking 60 consecutive plate appearances from three different seasons, each 2-3 years apart, should be a better statistical sample than 180 consecutive plate within one particular season.
I understand Shields is the #1 starter on a team playing in the WS. He was tremendous helping them get to the playoffs, but they're winning in the playoffs in spite of him, not because of him.
I'm not trying to totally crap on Shields. I like him. I think if he's our #2 on the right terms we'll be in good position to contend.