The game probability for Super Bowl XLVIII is now available at the New York Times. This week I take a look at why the efficiency model may differ from some other models and the public consensus.
...Some other analytical models, along with the consensus odds, give Denver a small advantage. I suspect the disagreement can be attributed to two factors. First, I doubt the strength-of-schedule effect is fully appreciated by wider audiences. And second, recent outcomes are often overweighted both by quantitative models (by design) and by fans and analysts (often unwittingly). Focusing on recent games would tend to favor Denver, which has appeared to be winning in easier fashion lately...
Game probabilities for the conference round are available at the New York Times. This week I explain why the Patriots' recent success running the ball is not so recent at all.
But to followers of advanced metrics, that is far from a surprise, as the Patriots have been mercilessly efficient on the ground for the past several years. Using success rate to measure running effectiveness provides a much more accurate picture of how the Patriots win with the run than traditional measures like total yards or even yards per carry...
...What the Patriots have figured out is how to use the running game effectively. They do not use it as a bludgeon on most first and second downs, as other teams do...
...Running backs like Adrian Peterson or Chris Johnson who are known for their occasional 80-yard breakaway touchdowns might be first-round draft picks in your fantasy league, but in the real world, I’ll take the Patriots’ running game any day.
Game probabilities for the division round are available at the New York Times. This week I take a look at the role of luck in the composition of the playoff field.
...the luckiest may be the Indianapolis Colts. Not only did they survive a 28-point second-half deficit in the wild-card round, but they were also the team that most overperformed statistical expectations in the regular season.
Based on the core efficiencies of passing, running, turnovers and penalties, plus the Colts’ strength of schedule, my efficiency model expected them to have between seven and eight wins. But they finished with 11, which was 3.2 more than expected.
All but 2 of the 12 playoff teams were above average in terms of beating their statistical expectations, and all but one of the surviving eight beat expectations. Although you may at first think this pattern points to an obvious flaw in the efficiency model, it turns out this is exactly what the field of statistics predicts...
Game probabilities for the wildcard round are available at the New York Times. This week I take a look at the matchup between KC and IND.
When the Colts host the Chiefs on Saturday in the first of this weekend’s four wild-card games, each team will feel as if it is looking into the mirror. The two teams are nearly statistically identical and tend to play disciplined, low-risk, ball-control football while letting their opponents make mistakes...
The two opponents are even more similar on defense. Kansas City’s and Indianapolis’s net yards per pass attempt allowed are 6.4 and 6.2, both close to the average of 6.2. They both allow 4.5 yards per carry, a bit more than the league average of 4.1. Both teams have grabbed more than their share of interceptions, as the Chiefs intercept the ball on 3.3 percent of all pass plays and the Colts on 2.8 percent...
Game probabilities for week 17 are available at the New York Times. This week I take a look at the narratives surrounding Tony Romo.
...However, the part of the story line that’s true is that Romo is a boom-and-bust performer. He can both lead his team to a dramatic victory, as he did last week in Washington, or make bone-headed plays to let victory slip from his team’s hands, as he did the week before in Green Bay.
Of all the quarterbacks with over a full season worth of pass attempts since 1999, Romo ranks 15th out of 152 qualified players in the game-by-game variance of WPA, which measures how far apart the peaks and valleys in a statistic tend to be...
Game probabilities for week 16 are available at the New York Times. This week I take a look at how to measure the playoff impact of individual games for both conferences.
...As the playoff field narrows with two games remaining, we can use what’s known as a Monte Carlo simulation to play out the remainder of the season several thousand times, weighting the game outcomes according to the game probabilities produced by the Advanced NFL Stats efficiency model...
Game probabilities for week 15 are available at the New York Times. This week I note the high playoff leverage of the NE-MIA game.
Game probabilities for week 14 are available at the New York Times. This week I look at the Kansas City Chief's unusual season.
...The efficiency model never thought the Chiefs were a dominant contender. At their peak after Week 7, they were ranked seventh over all. Since then, they’ve slipped to 22nd.
Admittedly, two of the Chiefs’ three losses came against a very good Broncos team, but the efficiency model accounts for opponent strength. In fact, despite being outplayed in their most recent loss, they rose one notch in the model’s efficiency rankings....
Game probabilities for week 13 are available at the New York Times. This week I look at Aaron Rodgers's value to the Packers.
...The model used here to produce the game probabilities is a “parametric” model, meaning its inputs are parameters that can be varied easily. That’s obviously a good thing, as team statistics certainly vary from week to week. It’s also handy because those parameters can be manually adjusted to account for injuries and just to play what-if...
By replacing Rodgers’s numbers with Flynn’s, and holding all other factors as is, we can get a rough idea of how the game probability should change. Unsurprisingly, the numbers would now heavily favor Detroit...
Game probabilities for week 12 are available at the New York Times. This week I look at the Jets' strange Pythagorean record.
...This season the Jets are doing their best to defy James, Pythagoras and their formulas. They have scored 183 points and allowed 268 on their way to a 5-5 record. That’s a net point difference of -85, which is second worst in the league, topping only the Jaguars. According to the Pythagorean expectation, the Jets should have about a .265 winning percentage, good for a 3-7 record or perhaps 2-8...
Game probabilities for week 11 are available at the New York Times. This week I do a back of the envelope analysis of how likely it would be for a team like the Chiefs to finish the regular season undefeated..
...But we know N.F.L. games are not coin flips, so how good would a team have to be to have a 50/50 shot at going undefeated through 16 games? If a team were so good that it had a 90 percent chance of winning any one game, it would only have a 19 percent chance of going 16-0. In fact, a team would need a 96 percent chance of winning any one game before it had better than even odds of going undefeated. So it’s a rare thing for a reason...
Game probabilities for week 10 are available at the New York Times. This week I take a quick look the changing of the Guard in the AFC North.
The model has not been a fan of the Ravens, even during last season prior to their playoff run to a Super Bowl victory. This season, the model says Baltimore's "true" winning percentage is 0.333, which isn't far off from their actual 0.375 mark (3-5), particularly considering they've had a slightly soft schedule. The Ravens' passing efficiency differential is -0.5 yards per pass play, which is 22nd in the league. But their biggest problem is a costly inability to run the ball. The Ravens offense is last in yards per carry, and is a distant last in run success rate...
Game probabilities for week 9 are up at the New York Times. This week I take a quick look at much things have changed in New England and Pittsburgh.
Pittsburgh's defense ranks 26th in the league in yards allowed per play, and its offense is struggling to run the ball. The Steelers rank 29th in the league in running Success Rate--the proportion of running plays that result in an improved scoring potential. Only the Giants, Jaguars, and Ravens rank behind the Steelers in running success...
Game probabilities for week 8 are up at the New York Times. This week I take a look at how lopsided this weekend's matchups are.
In terms of probabilities for the favorite and underdog, a game is typically about a 62-38 percentage affair. When the underdog wins almost 4 out of 10 games, it can make for an exciting Sunday afternoon. But this Sunday probably won’t be one of them...
On the other hand, when there are so many mismatches, there is bound to be a big upset. Statistically, it’s very unlikely that all those favorites will win...
Game probabilities for week 7 are up at the New York Times. This week I take a look at the Patriots Jets matchup and some surprising numbers.
A purely statistical approach is (deliberately) ignorant of things like the identity of the quarterbacks. It doesn’t know about Super Bowl rings or supermodel wives. It doesn’t know who is a rookie and who is a certainty for the Hall of Fame. It doesn’t know that Patriots linebacker Jerod Mayo is out for the regular season with a chest injury.
Game probabilities for week 5 are up at the New York Times. This week I discuss the how teams with great offenses can mask the true strength of their defense.
Since the 2000 season, the 32 teams with the best offenses allowed 2.1 more Expect Points per game than the average team over the same period — a significant difference. This translates to about a 4 percent chance of winning a game when matched against a roughly equal opponent, and slightly less when matched against a lesser opponent.By the way, I gooned up the table on the post, leaving out Monday night's game. It's .57 - .43 IND over SD.