Tuesday, December 30, 2008

Characteristics of Winning Teams

somehow, brian burke and i ended up independently writing about establishing the run on the same day. he looked at whether or not running the ball early/more really does tend to wear down a defense using the premise that if a defense really was wearing down, you'd see yards per carry increase as the number of runs increased -- that the YPC would be higher for the 30th carry than the 1st carry. turns out that conventional football wisdom may be wrong about this as well.

in fact, if you look at his graph, it actually looks like it has a slight downward slope -- that the early runs tend to be more effective than the late runs. though that could be impacted by other factors, e.g. teams that have a 40th carry are probably protecting a big lead forcing the other team to stack the line trying to get the ball back.

so, i wanted to take a look at what team stats correlated strongly with winning this season -- it's been done before, but i may be the first to do it with this year's numbers. here is what i came up with (correlating the following stats for each team with that team's number of wins this season):

Stat Corr
PF 0.692
PA -0.784
Net Pts 0.909
TO 0.430

so this basically makes sense, points scored correlates pretty well with the number of wins, but not as strongly as points against (defense wins championships, you know). though neither of them as strongly point differential, validating what we already know -- that you need both offense and defense to win. interestingly, turnover differential isn't that strong of a predictor.

here are some of the individual offensive and defensive stats (from nfl.com)

Category Stat Corr Category Stat Corr
TotOff Yds/G 0.561 TotDef Yds/G -0.726
TotOff Yds/P 0.477 TotDef Yds/P -0.682
TotOff 3rd Pct 0.552 TotDef 3rd Pct -0.480
TotOff Pen 0.004 TotDef Pen 0.192
TotOff Pen Yds 0.029 TotDef Pen Yds 0.178
TotOff RunPct 0.373 TotDef RunPct -0.715
OffPass CmpPct 0.488 DefPass CmpPct -0.405
OffPass Att/G -0.087 DefPass Att/G 0.426
OffPass YPA 0.483 DefPass YPA -0.689
OffPass Yds/G 0.297 DefPass Yds/G -0.424
OffPass 1st% 0.652 DefPass 1st% -0.580
OffPass 20+ 0.169 DefPass 20+ -0.378
OffPass Sck -0.473 DefPass Sck 0.562
OffPass QBRate 0.561 DefPass QBRate -0.594
OffRush Att/G 0.547 DefRush Att/G -0.775
OffRush Avg 0.149 DefRush Avg -0.532
OffRush Yds/G 0.484 DefRush Yds/G -0.722
OffRush 1st% 0.201 DefRush 1st% -0.509
OffRush 20+ 0.301 DefRush 20+ -0.591

some interesting stuff here. for the 2008 season, there were only five offensive team offensive stats that correlated fairly strongly with wins, they were (in order):

- percent of passes resulting in a first down
- total yards per game
- team QB rating
- third down conversion rate
- number of rushing attempts per game

of course, this doesn't tell us which of these factors are causative and which merely correlate, but here's how i interpret what this is telling us -- winning teams move the ball up and down the field via efficient passing games to score points and then protect them by running the ball.

there were lots of decent to good defensive predictors of success, the top 10 were (in order):

- rushing attempts per game against
- total yards against per game
- rushing yards against per game
- run/pass play mix against (more runs against = fewer wins)
- yards per pass against
- yards per play against
- QB rating against
- big rushing plays against (20+yards)
- percent of passes resulting in first down
- yards per rush against

a lot of these defensive stats overlap, but it's clear that stopping the run correlates strongly with winning. the question is, does stopping the run cause winning or does winning cause other teams to not run? it's not clear cut, but yards per rush against (i.e. defensive success against each running play) is a much weaker correlation than rushing attempts per game and total rushing yardage against which may indicate that that forcing teams to pass is slightly more important than stopping the run.

a few surprising things that conventional wisdom says are important, but don't in fact seem to correlate strongly with wins are:

- number of penalties (for or against)
- penalty yards (for or against)
- offensive rushing yards per carry



Anonymous Phil said...

Great analysis, however 1 season is probably too small a sample size to get any meaningful results. For instance, all the offensive #s vs. wins are skewed down by NO being #1 in offense (8-8) followed by ARI (9-7) and DEN and HOU which were both 8-8 as well.

On a totally different topic, when did the Eagles offense throw out the slant pass? You also hardly ever see them run the screen pass anymore. If you think about it, this offense is really quite a bit different than it was 5 years ago.

7:42 AM EST  
Anonymous Phil said...

And they were average teams because they all had epicly bad defenses. And average defense for any of them would have gotten those teams to 11-5 or better.

7:46 AM EST  
Blogger The Mean Guy said...

Great analysis, however 1 season is probably too small a sample size to get any meaningful results.

agreed, my plan is to try to do this going back for the last 5 seasons and see what that looks like.

8:09 AM EST  
Anonymous Anonymous said...

I'm surprised that running efficiently didn't correlate more, and that big passing plays also weren't as important as moving the chains through the air.

Really interesting and good work Meanguy, I have to think more about this after the New Year's hangover goes away.

- Behan

11:13 AM EST  

Post a Comment

<< Home