Chiefs Advanced NFL Stats: Chiefs-Giants closer than you'd think

**DaFace** · 09-26-2013, 07:07 PM

Not that this is surprising, but interesting to see a statistical analysis of this week's game.

http://www.nytimes.com/2013/09/26/sp...or-giants.html

Week 4 Game Probabilities: Better Outlook for Giants
By BRIAN BURKE

Weekly game probabilities return for 2013. In this space, we will reveal the outcome probabilities for each week’s N.F.L. games while examining interesting aspects of game prediction and football analytics.

This Sunday, the 0-3 Giants visit the 3-0 Chiefs. On the basis of each team’s record, the game looks one-sided, but a closer look reveals a much more even matchup.

Both the Chiefs and the Giants have turnovers to thank and to blame for their records through Week 3. The Chiefs haven’t turned the ball over once this season. They haven’t even fumbled, much less lost a fumble. On defense, the Chiefs have intercepted 3.5 percent of opponent pass attempts, significantly higher than the league average of 2.5 percent.

The Giants are at the other end of the spectrum. Quarterback Eli Manning has been intercepted on an amazingly frequent 7.6 percent of his pass attempts. And the Giant defense has a below-average 1.7 percent interception rate.

In retrospect, turnovers explain a great deal of a team’s fortunes. But prospectively, team turnover statistics don’t predict game outcomes as much as you might think. The reason behind this distinction is something called auto-correlation. Put simply, turnovers are very random. Only a small portion of a team’s past turnover rate carries forward to be predictive.

Consider offensive interception rate. In a sample of several recent seasons, interception rate correlates with a team’s win total at -0.46. Correlations are a measure of how connected two variables appear to be, calculated on a scale of -1 to 1. A correlation of 1 is perfect lock-step connection, and a correlation of zero means there is very likely no connection. A negative number means the two variables are inversely related; as one goes up, the other goes down. In this case, as offensive interception rate increases, win totals decrease as we would expect.

But if we divide the season into halves — all even numbered games and all odd number games — we see that team interception rate isn’t consistent. In fact, the correlation is about 0.1, meaning that a team’s past interception rate doesn’t tell you much about how many interceptions it may have from game to game. To put that in perspective, the most consistent aspect of team performance is offensive pass efficiency, which correlates from week to week at 0.6.

And that’s why the Chiefs-Giants game isn’t as lopsided as one might think. The game probabilities for Week 4 are at right. (* Denotes the neutral site game in London, where home field advantage was removed from the probability model.)

09-26-2013, 07:07 PM	Topic Starter
DaFace Kind of a mod Join Date: Aug 2005 Location: Donkey Land Casino cash: $1526899	Advanced NFL Stats: Chiefs-Giants closer than you'd think Not that this is surprising, but interesting to see a statistical analysis of this week's game. http://www.nytimes.com/2013/09/26/sp...or-giants.html Week 4 Game Probabilities: Better Outlook for Giants By BRIAN BURKE Weekly game probabilities return for 2013. In this space, we will reveal the outcome probabilities for each week’s N.F.L. games while examining interesting aspects of game prediction and football analytics. This Sunday, the 0-3 Giants visit the 3-0 Chiefs. On the basis of each team’s record, the game looks one-sided, but a closer look reveals a much more even matchup. Both the Chiefs and the Giants have turnovers to thank and to blame for their records through Week 3. The Chiefs haven’t turned the ball over once this season. They haven’t even fumbled, much less lost a fumble. On defense, the Chiefs have intercepted 3.5 percent of opponent pass attempts, significantly higher than the league average of 2.5 percent. The Giants are at the other end of the spectrum. Quarterback Eli Manning has been intercepted on an amazingly frequent 7.6 percent of his pass attempts. And the Giant defense has a below-average 1.7 percent interception rate. In retrospect, turnovers explain a great deal of a team’s fortunes. But prospectively, team turnover statistics don’t predict game outcomes as much as you might think. The reason behind this distinction is something called auto-correlation. Put simply, turnovers are very random. Only a small portion of a team’s past turnover rate carries forward to be predictive. Consider offensive interception rate. In a sample of several recent seasons, interception rate correlates with a team’s win total at -0.46. Correlations are a measure of how connected two variables appear to be, calculated on a scale of -1 to 1. A correlation of 1 is perfect lock-step connection, and a correlation of zero means there is very likely no connection. A negative number means the two variables are inversely related; as one goes up, the other goes down. In this case, as offensive interception rate increases, win totals decrease as we would expect. But if we divide the season into halves — all even numbered games and all odd number games — we see that team interception rate isn’t consistent. In fact, the correlation is about 0.1, meaning that a team’s past interception rate doesn’t tell you much about how many interceptions it may have from game to game. To put that in perspective, the most consistent aspect of team performance is offensive pass efficiency, which correlates from week to week at 0.6. And that’s why the Chiefs-Giants game isn’t as lopsided as one might think. The game probabilities for Week 4 are at right. (* Denotes the neutral site game in London, where home field advantage was removed from the probability model.)
Posts: 52,038