We are just over 80% of the way through the NFL regular season. What if we looked at power rankings based on each team’s likelihood of winning each week, based on 80% of each game–roughly the first 48 minutes of a 60 minute game? It’s an arbitrary number, as all of them really are, but I chose it since it meant we were looking at a sizeable chunk of the game already played.
To calculate it, I used Advanced NFL Stats’ game probabilities and found the point in time closest to the 12 minute mark of the fourth quarter. Teams were then given credit for a fraction of a win equal to their game winning probabiity at that time. For example, there were 41 occasions (not including Thursday Night’s game involving Cincinnati and Philadelphia) where a team had greater than or equal to a 99% chance of winning at the 12 minute mark (and all 41 won). They were given credit for 0.99 wins. If a team had a 10% chance of winning, like Dallas did last week when they came from 9 points down against Cincinnati, they would get credit for 0.1 wins. Cincinnati would get credit for 0.9 wins in that scenario.
Here are the “80% win probability win totals” through week 13, listed along with the actual win totals, and the pythagorean win expectation based on points for and points against for the year:
While New England trails Houston and Atlanta in the win column, they rank way out in front here, as New England has been given at least an 80% chance of winning every game except for Arizona game (and how weird was that result?). The Patriots had a 9 point lead against Baltimore and a 10 point lead against Seattle, both games they lost by a point.
The two largest positive differences, not suprisingly at all, are Indianapolis and Atlanta. Indianapolis’ expected win total of 5.6 is in line with their pythagorean expectation of 5.3 wins, and also near where many play by play rankings have them. The Colts have basically won every game that was still in doubt with 12 minutes remaining, even if a comeback by them was unlikely. Indianapolis has played eight games where the favorite’s win probability was less than 90% at the 12 minute mark of the fourth quarter. Despite being the underdog by trailing in five of those eight games, the Colts are 7-1.
At the opposite end is the Chargers. San Diego and Carolina are the two teams who have underachieved the most based on what we should expect knowing how the first 48 minutes went.
One of the reasons why I did this is to see the differences between looking at a win probability model and the pythagorean model. We know that not all close games are equal, and singular blowouts can skew the pythagorean results. For example, a team that is up by 14 but wins by 7 when the opponent scores a late touchdown is probably more dominant than one who wins a tossup game.
We also know that pythagorean win totals are more predictive of the following season than just looking at the actual wins. I would be interested in looking further to see if teasing it out even further and looking at win probabilities through certain points were more predictive than pythagorean results.
In looking at teams where the pythagorean and win probability models differ, we see teams like Chicago and Seattle are basically in line with their win totals when we look at win probabilities at the 80% mark, but have higher pythagorean totals. This is to be expected, where Seattle just won 58-0, and Chicago had some blowouts early. Seattle, for example, would have had roughly the same chances of winning that game if they were up 29-0 rather than 58-0, and thus the effects of a blowout are muted.
[photo via USA Today Sports Images]