The "Seahawks Hangover"

Barnwell analyzes the "Seahawks Hangover" and its validity.
http://grantland.com/the-triangle/curing-the-seattle-hangover/

Was a valiant effort but I don't think it's quantifiable outside of opposing players giving an honest opinion on the topic. It's fun to talk about and laugh at but in the end, Barnwell has no idea. Only the players do.

The analysis looks oddly flawed.

First, an 0-10 process is indicative of a relationship. "Correlation does not mean causation" but it usually means there is some underlying relationship.

If the #s he put out, 6 in a row in 2013, 10 in a row in 2014 are correct then there IS a hangover. What it is becomes just semantics. (I did not check his #s btw).

Also, what would be interesting to track are the magnitude of the losses and how it differs from how the team normally performs against that team. You don't get every team getting to play the other twice, and the home vs away #s skew things but there should be enough data to paint a picture.

Finally, the impact could be felt by taking the above .500 teams and checking how they played, as well as identifying how many of these are "bad losses" where a stronger team lost to a substantially weaker team.

OR just do it the lazy way and count the # of blowouts vs a collection of random samplings of 10 teams.

His analysis went deep in a few areas, overlooked that the 10 (16?!?) in a row #s are too large a collection to just throw to chance - but then concluded there is no effect. Not sure the method or even the #s bear that out.

Barnwell tries to dismiss it, but it's not that easy to dismiss. After the 6 game hangover streak in 2013 is when the theory got legs. It's easy to imagine teams wanted to take these cocky new hotshots down a peg or two, and trying your best but still walking away with a loss has got to be demoralizing. That's not to mention the reasonable possibility that playing Seattle is physically more demanding.

Was the hypothesis legit? Well, the next 23 games contained a hangover losing streak of 10 games. The odds of a 10 games or greater hangover streak occurring within that 23 game segment by chance is around 2%. The probability that an unlikely phenomenon results from something other than chance is proportional to how plausible the other explanations are. Is the "beat down/letdown" theory plausible? I think so: It was proposed *before* the unlikely supporting evidence accumulated. The 10 game streak wasn't dug up later by sifting through mounds of raw data looking for patterns, which often results in spurious correlations.

It's not enough evidence to be proof, but it's too much to brush off.

I read the article earlier today and was pretty disappointed in the analysis. He should've looked at something like DVOA, which adjusts for opponent quality and is a better predictor of future success than simple win-loss record. And there's really no excuse for not at least comparing to a baseline performance, i.e., how did the teams do in all their other games? It's actually really surprising to me that he did neither of those things, since Barnwell is easily one of the more advanced-stats-obsessed writers out there. You can usually count on him to at least present all the relevant statistics. It's really a pretty disappointing effort on his part, and makes me think he was skeptical of the premise going in, and stopped halfway through when his initial analysis confirmed his bias.

His analysis of win-loss record also suggests that teams should regress to the mean the following season, yet the Seahawks did the opposite. That doesn't necessarily prove anything, but when you combine it with the anecdotal evidence from players who actually play the game, it suggests that the idea deserved much more careful scrutiny than it got.

There may not be anything to the Seattle hangover, but the definitive analysis has yet to be done.