[BostonCard]

The new coaches poll is out, and features Stanford at #9 and Washington at #11.

Let's get this straight. Washington largely outplays #9 Auburn in a virtual home game for the Tigers 3,000 miles from Seattle, and the Huskies fall 5 spots after a close loss. 

First of all, with one game played, the functional difference between a team ranking #11 (as the coaches did) and #9 as you ranked them is insignificant.  Pretty much every ranking right now has a big margin of error, and a 2 position difference is just meaningless.  Even well into the season, a two position ranking difference is effectively noise beyond the second or third ranked team.  So, I don't know why you are upset at the coaches.

Secondly, given that there's a lot of uncertainty around every ranking, every data point is important.  Thus, a loss, even to an expected good team should be damaging to ranking.  Why?  Well, part of Washington's pre-season ranking at #6 is the recognition that it might be better than 6 and it might be worse than 6.  Losing to a presumed good team doesn't mean that a top-10 ranking is out of the cards, but it does suggest that unless the outcome of that game was an outlier, that it is unlikely that Washington is an elite (top-5 team).  On the other hand, we don't know if they would also lose a game to a somewhat weaker team, since they haven't played.  Nor do we know if Auburn is as good as we think it is (it beat Washington, so you can presume it's not bad, but you could easily see a situation where Auburn is, after a few losses, ranked a something like #15, and Washington, after a couple of losses is hanging around #20).  So, if you figure Washington might reasonably be as good as #6 and reasonably as bad as #20 (depending on what subsequent games show), a ranking at #11 might be a reasonable place to slot them while you wait for more data.  Or you could slot them at #9, after all, this early the two rankings are basically identical.

From a Bayesian perspective, you figure current ranking = prior ranking + game results.  In most cases, because the likely outcome happened (Stanford was a 14-point favorite and won, the only new thing we know is that the downside risk, that Stanford is a middling or worse team, is less likely, even though it was a lowish probability event to begin with).  With Washington, they lost a game they were just one point dogs, so that loss is pretty informative.

But here's the thing, we have a whole new slate of games this week to further update our priors.  Washington plays North Dakota, so unless Washington loses, that game will not be informative.  On the other hand, Stanford plays USC as a 4-ish point favorite.  If we win that game, like Auburn's win against USC it will establish that, unless USC was massively overrated or the game was a fluke, Stanford is unlikely to be worse than a good team.  Lose to USC, and, like UW, you can reasonably say that we are unlikely to be elite.

BC

[BostonCard]

The new coaches poll is out, and features Stanford at #9 and Washington at #11.

Let's get this straight. Washington largely outplays #9 Auburn in a virtual home game for the Tigers 3,000 miles from Seattle, and the Huskies fall 5 spots after a close loss. 

First of all, with one game played, the functional difference between a team ranking #11 (as the coaches did) and #9 as you ranked them is insignificant.  Pretty much every ranking right now has a big margin of error, and a 2 position difference is just meaningless.  Even well into the season, a two position ranking difference is effectively noise beyond the second or third ranked team.  So, I don't know why you are upset at the coaches.

Secondly, given that there's a lot of uncertainty around every ranking, every data point is important.  Thus, a loss, even to an expected good team should be damaging to ranking.  Why?  Well, part of Washington's pre-season ranking at #6 is the recognition that it might be better than 6 and it might be worse than 6.  Losing to a presumed good team doesn't mean that a top-10 ranking is out of the cards, but it does suggest that unless the outcome of that game was an outlier, that it is unlikely that Washington is an elite (top-5 team).  On the other hand, we don't know if they would also lose a game to a somewhat weaker team, since they haven't played.  Nor do we know if Auburn is as good as we think it is (it beat Washington, so you can presume it's not bad, but you could easily see a situation where Auburn is, after a few losses, ranked a something like #15, and Washington, after a couple of losses is hanging around #20).  So, if you figure Washington might reasonably be as good as #6 and reasonably as bad as #20 (depending on what subsequent games show), a ranking at #11 might be a reasonable place to slot them while you wait for more data.  Or you could slot them at #9, after all, this early the two rankings are basically identical.

From a Bayesian perspective, you figure current ranking = prior ranking + game results.  In most cases, because the likely outcome happened (Stanford was a 14-point favorite and won, the only new thing we know is that the downside risk, that Stanford is a middling or worse team, is less likely, even though it was a lowish probability event to begin with).  With Washington, they lost a game they were just one point dogs, so that loss is pretty informative.

But here's the thing, we have a whole new slate of games this week to further update our priors.  Washington plays North Dakota, so unless Washington loses, that game will not be informative.  On the other hand, Stanford plays USC as a 4-ish point favorite.  If we win that game, like Auburn's win against USC it will establish that, unless USC was massively overrated or the game was a fluke, Stanford is unlikely to be worse than a good team.  Lose to USC, and, like UW, you can reasonably say that we are unlikely to be elite.

BC