Recently I’ve been working on a better way to compile team statistics and to project team wins. A very good and easy way is use what I call ‘QB Rating Differential.’ This takes a team’s offensive QB Rating and subtracts their defensive QB rating allowed.
For you non-stat buff people out there, a good way to figure out the ‘power’ behind a statistic is to find what’s called the ‘correlation coefficient.’ This is a number that tells how on thing correlates to another thing. And it is a number that goes from -1.0 to +1.0.
The closer the number is to +1.0, the stronger of a direct correlation is. So for instance, if you owned a store and wanted to see the correlation between iced tea sold and temperature, you may come up with a correlation coefficient of +0.7 which is a pretty strong correlation….meaning, the higher the temperature outdoors, the more likely you are to sell more iced tea. The closer the number is to -1.0, the stronger the indirect correlation is. And the closer to 0 means there is no correlation.
The cool thing about QBRD is that it’s easy to figure out and it has a very strong correlation to winning and losing in the NFL. Over the past 6 seasons, the correlation coefficient of QBRD to wins is at +0.8. Furthermore, QBRD can be easy to understand because typically a team’s QBRD that is close to zero usually means they’ll win about 8 games.
When we have the correlation coefficient of any statistic, we can then make projections/predictions. And the closer that coefficient is to +1.0, the more accurate our projections/predictions will be.
And while QBRD has a very strong correlation coefficient and generally does a very good job of projecting wins, there is always about 4 teams each year that are ‘way off’ from the projection. Usually 1-2 teams will ‘overachieve’ (have much more wins than their QBRD suggests they should have) and 1-2 teams that ‘underachieve.’ (have much less wins than their QBRD suggests).
So I wanted to come up with a way to use some other stats so I can get an even stronger correlation coefficient.
So I took what I believe to be important factors in the game and also sought out to see what teams that that greatly ‘overachieved’ and ‘underachieved’ to see what they did that allowed them to skew so far from what their QBRD was. 2008 was the season where we saw some big ‘overachievers’ and ‘underachievers’ in the ’08 Carolina Panthers (overachieved), ’08 Denver Broncos (overachieved), and the ’08 Green Bay Packers (underachieved).
Obviously, the first statistic I use is QBRD. It’s correlation is strong, so those 3-4 teams a year that skew from the QBRD don’t really ‘spoil the pot.’
The 2nd statistic I came up with was this formula: (Points Scored / Total Yards Gained) * 1000
I found that points scored / total yards gained has a very strong correlation of +0.75. I needed to multiply it by 1000 in order to show the significant effect this stat has. This statistic I think shows a few things:
1. A teams ability to move the ball. Theoretically you may only move the ball 20 yards a drive and score 40 points, but that would require tremendous field position. So in order to score, you just have to move the ball, period.
2. A teams ability to ‘finish off’ the drive. Not only how well they do it by scoring TD’s over FG’s, but also making those field goals.
3. A team’s ability to win the field position battle. Not only in special teams, but getting turnovers.
Now, I tried to do something similar for defense (total yards allowed / total points allowed), but the correlation wasn’t nearly as strong (about +0.5) so I disregarded it.
The 3rd stat is the team’s Points Allowed Ranking. This I found this interesting because from 1984 to 2005, only 1 team NO team had *made* the Super Bowl and finished higher than 8th in points allowed that season.
After ’05, we then had the ’06 Colts, the ’07 G-Men, the ’08 Cardinals and the ’09 Saints all make the Super Bowl and finish outside the top 8 in points allowed. The Colts defense had just regained Bob Sanders before the playoffs and shutdown opposing offenses with Sanders playing. The G-Men’s defense was starting to come along and led the league in sacks. The Cardinals were very unique to make the SB, although they played in a super weak division and had the advantage of homefield. And the ’09 Saints had a great offense and created a massive amount of turnovers.
Perhaps the correlation is going to fall in time and the league becomes more of a scoring league, but the correlation between winning and points allowed is still pretty good (around +0.65) so I kept it in there.
Lastly, the statistic I used was FootballOutsiders.com’s Adjusted Sacks *allowed* rate. This is a stat that not only factors sacks allowed, but how many pass attempts you had, the down and distance (3rd and long being an easier situation to give up a sack) and the quality of pass rush the team is facing.
In the past I had used defensive sacks, but the correlation between defensive sacks and winning or losing its steam. Probably because of the more shotgun happy offenses and the more blitz happy defenses. The correlation is decent (+0.61) but this was usually the missing link between all of the overachievers and underachievers.
The ’08 Panthers and Broncos whom ‘overachieved’ did very well that year in Adjusted Sacks Allowed and the ’08 Packers who underachieved did very poorly in Adjusted Sacks Allowed.
My guess is that the reason for this is that not only does it prevent the QB from getting rattled, but also the likelihood of the QB getting injured is influenced by this as well. Lastly, it plays a big difference in both field position and down and distance. Get sacked on 1st down and then you’re left with a 2nd and 17 and the defense knows you’re almost likely to throw the next 2 downs so you can pick up a first down.
With that, I made a formula to weight these stats and over a 6 season span I got a correlation of +0.9. Since I can’t really figure in strength of schedule, I can’t imagine it getting much higher than that. But with:
QBRD
Points Per Yards Gained
Turnover Margin
Points Allowed Ranking
Adjusted Sacks Allowed
I think I’ve attacked key elements of the game in the passing offense, passing defense, pass protection, ability to score points, turnovers, FG kicking, field position, etc.
So with that, I’ve used a linear regression (a formula that can be used to project something) to project the wins based on current performance. Obviously, this will fluctuate, I really haven’t come up with a name for my formula of projecting wins other than Team Statistical Health Index. But here’s the projections going into Week 4.
1…Indianapolis Colts……………11.77
2…New York Jets……………11.67
3…Green Bay Packers……………11.52
4…Pittsburgh Steelers……………11.14
5…Tennessee Titans……………10.91
6…Atlanta Falcons……………10.82
7…Philadelphia Eagles……………10.51
8…Kansas City Chiefs……………10.03
9…Chicago Bears……………10.02
10…New England Patriots……………9.90
11…New Orleans Saints……………9.79
12…Cincinnati Bengals……………9.39
13…San Diego Chargers……………9.18
14…Seattle Seahawks……………9.17
15…Tampa Bay Buccaneers……………8.71
16…Miami Dolphins……………8.25
17…St. Louis Rams……………8.19
18…Dallas Cowboys……………8.15
19…Baltimore Ravens……………7.86
20…Denver Broncos……………7.06
21…Washington Redskins……………6.96
22…Cleveland Browns……………6.71
23…Detroit Lions……………5.74
24…Arizona Cardinals……………5.37
25…Minnesota Vikings……………5.30
26…New York Giants……………5.21
27…Houston Texans……………4.33
28…Oakland Raiders……………3.93
29…Buffalo Bills……………3.06
30…Carolina Panthers……………2.54
31…Jacksonville Jaguars……………2.22
32…San Francisco 49ers……………1.38
3JACK
No comments:
Post a Comment