Super Bowl Prediction: Slight Edge to Seahawks
Football fans should be in for a treat Sunday, as Denver and Seattle ended the season ranked No. 1 and 2 in the Advanced NFL Stats efficiency model by virtue of each team’s dominance on one side of the ball. For the Broncos, that side is obviously offense, led by quarterback Peyton Manning. And for the Seahawks, it is defense, led by their smothering secondary.
In place of the game probabilities table in the graphic accompanying this article, there is a small table listing each team’s core statistics in the model along with the league average for each category. YPA refers to yards per attempt, a measure of passing efficiency. Run SR refers to rushing success rate, which is the proportion of runs that improve a team’s chance of sustaining a drive and scoring. Int and Fum % refer to interception and fumble rates per snap. D refers to the defense’s version for each statistic. PenRate represents a team’s penalty yards per snap.
Right away, the table boils down the most significant similarities and differences between the teams. They appear roughly equal in the running game and offensive turnovers. They are both penalized slightly more often than average, although Seattle more so.
<snip>
Seattle’s playoff run was tougher, too. New Orleans and San Francisco were measurably better teams than San Diego and New England. Accounting for strength of schedule, Seattle is the slightly stronger team and should be favored to win Super Bowl XLVIII with an edge of 52 percent to 48 percent.
http://www.nytimes.com/2014/01/31/sports/football/super-bowl-prediction-slight-edge-to-seahawks.html?_r=1
His prediction models are actually very good, for an understanding on how they work I refer you to this
<snip>
The most predictive stat is team net passing yards per attempt. “Net” means that sack yards are factored in to the average. Both offensive and defensive pass efficiency are of equal consequence, but offensive efficiency is more consistent. If you just looked at those two stats alone, you’d have a fairly good prediction model for the N.F.L.
Turnovers matter a great deal in determining winners, but turnover rates aren’t as predictive of future outcomes because they don’t tend to be consistent for teams from week to week. I use offensive and defensive interception rates and offensive fumble rates in the model, but they are not weighted as heavily as you might think.
Running obviously matters, but not in the same way as passing. The model uses running “success rate” instead of yards per carry to represent each team’s running ability. Success rate is the percentage of run plays in which a team’s down-distance-yard-line situation is improved. This is better than yards per carry because of how many run plays are designed for short gains. A 2-yard gain on third-and-1 or fourth-and-1 would hurt a team’s yards per carry average, but success rate would recognize it as a positive play.
The model also includes team penalty yards per play. This is a surprisingly consistent team stat, and it can certainly have a significant effect on game outcomes.
http://fifthdown.blogs.nytimes.com/2012/09/27/n-f-l-week-4-game-probabilities-are-back/
= new reply since forum marked as read
