In the immediate aftermath of the New England Patriots’ 28-24 Super Bowl victory over the Seattle Seahawks Sunday night, we gave a preliminary measurement of how exciting the game was using what Brian Burke of Advanced Football Analytics calls the “Excitement Index.” This index tracks the cumulative change in win probability throughout a game, the logic being that bigger swings in win probability indicate a more thrilling game.
By NumberFire’s live in-game probability model, the Excitement Index of Super Bowl XLIX was curiously modest, ranking just 12th all time. But, as we noted, win probability models can vary to a surprisingly large degree, so we wanted to recalculate the Excitement Index using data from Pro-Football-Reference.com, which hadn’t yet updated Sunday night. (We used Pro-Football-Reference in our original ranking of the most exciting Super Bowls last week.)
The difference between the two sources is bigger than you might think. If we run the Pro-Football-Reference numbers, Sunday’s game comes in at No. 3 all time, a much higher — and, in our subjective view, more deserving — placement than we’d originally calculated.
In addition to the vagaries of competing win probability metrics, it’s also worth noting that the Excitement Index has obvious limitations. Similar to the way the coastline paradox makes it difficult to pin down the true length of a landmass’s coastline, a bunch of incremental changes to win probability can add up quickly for a game’s Excitement Index even if the overall trend of a game is in the same direction.
The Excitement Index (as originally defined by Burke) also counts a swing in win probability from, say, the first quarter the same as one from the fourth quarter. In terms of leverage index, this is entirely appropriate — but it may not track as well with the subjective feeling of excitement we tend to experience as fans, where late-game moments are given much more weight.
To try to capture some of that feeling, we also calculated an alternative version of the Excitement Index that puts more weight on the end of the game. Here’s how it works: A play at the very end of the game receives a weight of 2, halftime receives a weight of 1, and the opening kickoff gets a weight of 0. It’s a bit ad hoc, we know, but it seems to produce ratings that match the perceived excitement of Super Bowls better than an unweighted sum of win-probability changes.
Using our alternative Excitement metric, Super Bowl XLIX is second all time. (See the table below for the updated ranking by this method.)
Then again, as we wrote last night, no index can ever really put a number on the elation felt by Patriots fans — and the corresponding despair of those rooting for the Seahawks — at the end of a game like Sunday’s.