[BostonCard]

[quote author=Boston Card link=topic=12578.msg125614#msg125614 date=1440621024]
The same exercise for Stanford:

I'd have to think that is too much effort to expend based upon input numbers that have no information content at this time.

I cannot find an explanation from ESPN as to what goes into their FPI [perhaps just because their newish mobile-friendly site is quite difficult to manage], but whatever the inputs there clearly is no 2015 performance information included.
[/quote]

I agree that at this point any system, whether quasi-mathematical or not is really not much more than an educated guess.  A reasonable argument can be made that what I did was garbage in/garbage out, and a better model of things would be Bayesian, since the games occur sequentially, the results of one game give us some information about the next one, and thus are not completely  independent.

And yet, I still found the exercise somewhat useful even if the specific numbers are garbage.  First, the results correlate reasonably well with others, based on "gut feelings" and subjective analyses by both us and professionals; I think an 8 - 10 win season is certainly the most likely outcome for us.  Secondly, it gives us a sense of a range of what to expect.  It looks very unlikely that we would fail to get 6 wins; thus we have every reason to expect a bowl game.  Likewise, an 11-1 or 12-0 regular season is very unlikely but not impossible; couple that with a 50% probability of winning the Pac-12 championship if we do wind up 11-1, and I think making the playoffs is very unlikely but high enough to give me hope that there is a chance.

Josh

[BostonCard]

[quote author=Boston Card link=topic=12578.msg125614#msg125614 date=1440621024]
The same exercise for Stanford:

I'd have to think that is too much effort to expend based upon input numbers that have no information content at this time.

I cannot find an explanation from ESPN as to what goes into their FPI [perhaps just because their newish mobile-friendly site is quite difficult to manage], but whatever the inputs there clearly is no 2015 performance information included.
[/quote]

I agree that at this point any system, whether quasi-mathematical or not is really not much more than an educated guess.  A reasonable argument can be made that what I did was garbage in/garbage out, and a better model of things would be Bayesian, since the games occur sequentially, the results of one game give us some information about the next one, and thus are not completely  independent.

And yet, I still found the exercise somewhat useful even if the specific numbers are garbage.  First, the results correlate reasonably well with others, based on "gut feelings" and subjective analyses by both us and professionals; I think an 8 - 10 win season is certainly the most likely outcome for us.  Secondly, it gives us a sense of a range of what to expect.  It looks very unlikely that we would fail to get 6 wins; thus we have every reason to expect a bowl game.  Likewise, an 11-1 or 12-0 regular season is very unlikely but not impossible; couple that with a 50% probability of winning the Pac-12 championship if we do wind up 11-1, and I think making the playoffs is very unlikely but high enough to give me hope that there is a chance.

Josh