This past weekend the Raiders took the Broncos out to the woodshed to the tune of a 59-14 victory. They scored through the air and on the ground; they scored on an interception return and with a field goal. They had twice as many yards gained, nearly three times as many first downs, and forced three turnovers. About the only thing they did wrong was get 10 penalties. The model rewarded Oakland by moving them up 11 spots from 25th in the league to 14th:
The only other big movers were teams dropping in the standings; Dallas, Denver, and Philly all lost ground, although Dallas and Philly dropped in the middle of the pack where teams are closer together. Dallas should be more worried about losing Tony Romo for over a month; the Cowboys do not run well and their defense isn’t going to help them out, as evidenced by giving up almost 60 points despite getting 5 turnovers. If they can’t throw the ball, they are going to be in deep trouble.
Here are the predicted wins for the season.
I made some changes in the presentation. The first two columns are still games actually won and then expected wins, which is a sum of each team’s win probability in each game they play from weeks 2 to 17, plus one if they won in week 1 (which isn’t predicted). These probabilities are updated each week, so Oakland for example is now expected to win almost 2 more games than they had last week. The next column is new; it’s the sum of the win probabilities for games already played, plus one if they won in week 1. You can view this as a measure of how a team should have done so far given what we know now. You can also compare it to the actual win column to get a sense of how lucky or unlucky a team has been so far. By ‘lucky’ I don’t mean rabbits feet or voodoo curses, but things that are generally not predictive of future performance but can affect single games, like special teams plays, turnovers at critical places on the field, etc. At the top of the list you can see Pittsburgh, who is essentially performing as expected. But you can scan down to a team like the Jets, who the model thinks should be about average but are in reality tied for the best record in the league. The Jets have recovered more fumbles than one would expect so far and have had pretty decent special teams play, and so have won a bit more than the model thinks they should.
The final column is also new and is the total win probability for future games for each team. You’ll notice that the last two columns add up to the second one, since the expected wins in future games and previous games adds up to the expected season wins. This might be the most interesting column (depending on how you feel about luck) since it tells us how teams should do going forward given their schedule. As I’ve said before, power rankings are nice but aren’t useful if teams have different strength of schedule in their remaining games. The Jets actually jump out again; they appear to have the good fortune of being lucky early and having a nice schedule late. They could challenge the Steelers for best record in the league.
Here are the updated playoff predictions based on current wins plus predicted future wins: In the NFC, the division winners should be the Giants, Green Bay by a nose, Atlanta, and Seattle. The favorites for a wild card are New Orleans, Chicago, and Tampa Bay, but teams are pretty grouped together. The AFC division winners are the Jets, Pittsburgh, Tennessee by a nose over Houston and Indy, and Kansas City. The leaders for the wild card spots are New England followed by Baltimore, Houston, and Indy.