[d4cohn]

Meanwhile Connelly's S&P+ system dropped Stanford by 13(!) places. Seems turnover luck played a role -- it expected Stanford to be -.61 on the turnover margin while Stanford (obviously) won the turnover battle 3-0. It seems a bit puzzling because Stanford had 11 passes broken up, which is a high number; two picks doesn't at all seem inordinate within that context. But what do I know? 

ESPN FPI moved Stanford up to #10. 

Sagarin kept Stanford stable.

I've read that Connelly made some changes to S&P+ this year leading to crazy movement early on. 

[/quote]
It's weird -- Stanford drops by 13, SC drops by 11 ... SDSU rises by 12, and SC's first-game opponent rises by 10. Very hard to figure out what's going on here.
[/quote]

This is an indication, to me at least, that perhaps the changes to the metric weren't good ones. Granted, it is a tough balancing act between giving too much weight to your prior (I would put ESPN FPI in this camp; for example, it still has Michigan at No. 7, and LSU at No. 20, even though the eye test, let alone basic stats, would say that LSU and Michigan are at the very least comparable, if not LSU being better) versus making your metric too susceptible to fluctuations by introducing new data. It is not easy, and even the best statisticians produce a model or metric that fails. If anything, it indicates why one would hope that selection committees across the sports (football and otherwise), are using a variety of metrics (and truly considering all of them) as they profile each team.

[d4cohn]

Meanwhile Connelly's S&P+ system dropped Stanford by 13(!) places. Seems turnover luck played a role -- it expected Stanford to be -.61 on the turnover margin while Stanford (obviously) won the turnover battle 3-0. It seems a bit puzzling because Stanford had 11 passes broken up, which is a high number; two picks doesn't at all seem inordinate within that context. But what do I know? 

ESPN FPI moved Stanford up to #10. 

Sagarin kept Stanford stable.

I've read that Connelly made some changes to S&P+ this year leading to crazy movement early on. 

[/quote]
It's weird -- Stanford drops by 13, SC drops by 11 ... SDSU rises by 12, and SC's first-game opponent rises by 10. Very hard to figure out what's going on here.
[/quote]

This is an indication, to me at least, that perhaps the changes to the metric weren't good ones. Granted, it is a tough balancing act between giving too much weight to your prior (I would put ESPN FPI in this camp; for example, it still has Michigan at No. 7, and LSU at No. 20, even though the eye test, let alone basic stats, would say that LSU and Michigan are at the very least comparable, if not LSU being better) versus making your metric too susceptible to fluctuations by introducing new data. It is not easy, and even the best statisticians produce a model or metric that fails. If anything, it indicates why one would hope that selection committees across the sports (football and otherwise), are using a variety of metrics (and truly considering all of them) as they profile each team.