Characteristics of Winning Teams (2004-2008)
following up on the previous post, here are the regressions for team stats over the last five seasons. whereas last time, i listed tried to eliminate duplicates (e.g. attempts and attempts per game) and limit the list to only the variables i thought were important (e.g. i removed 4th down conversions and pcts). this time, i thought it would be best to just post all the data so everyone can make their own assessments.
i've done a little more formatting than i typically do on this minimalist blog due to the increased amount of data.
so just to recap, the following tables display the correlation coefficient between all of the basic team stats from nfl.com over the last five seasons and the number of wins earned by the team that generated those stats.
cells highlighted in yellow represent a correlation coefficient between .5 and .65.
cells highlighted in orange represent a correlation coefficient between .65 and .8.
cells highlighted in red represent a correlation coefficient greater than .8.
not sure exactly what this tells us, but here are some initial observations:
i've done a little more formatting than i typically do on this minimalist blog due to the increased amount of data.
so just to recap, the following tables display the correlation coefficient between all of the basic team stats from nfl.com over the last five seasons and the number of wins earned by the team that generated those stats.
cells highlighted in yellow represent a correlation coefficient between .5 and .65.
cells highlighted in orange represent a correlation coefficient between .65 and .8.
cells highlighted in red represent a correlation coefficient greater than .8.
not sure exactly what this tells us, but here are some initial observations:
- clearly the strongest correlation for defense is to not get run on. not to "stop the run" necessarily (though obviously forcing 3 and outs would limit the number of rushes against), but the number and percentage of carries against were the strongest relationships on either side of ball.
- against the rush, the aggregate stats (attempts, total yards, number of 1st downs) appear to correlate much stronger than the efficiency stats (avg/YPC, 1st down pct)
- against the pass, the reverse is true as the efficiency stats (avg/YPA, QB rating) show much stronger correlation than the aggregate stats (completions, yards, 1st downs)
- on offense, we see a similar pattern with aggregate stats mattering for rushing (attempts, number of TDs, number of 1st downs) and efficiency stats mattering for passing (QB rating, avg/YPA, 1st down pct)
- i continue to be shocked how little YPC seems to correlate to the win totals of the teams
- turnovers showed a much stronger correlation with the broader data set
here are a couple of scatterplots to show you how recent eagles teams compare against the league over the last five seasons.
first wins vs. point differential for the whole league over the last five seasons. the last five eagles teams are shown as large red points. this is a pretty obvious relationship, but it's interesting to see just how much this year's team has underperformed relative to their overall point differential (point differential has a .917 correlation with wins).
next is one of the stronger offensive stats -- percentage of passes that result in a first down. this had a correlation coefficient of .601.
Labels: football
posted by The Mean Guy at 5:02 PM
3 Comments:
YPC doesn't mean much because the lower percentage pass plays usually pick up more yardage than the higher percentage ones (i.e. screens). Better to hit 1 in 5 for 50 yards than 4 in 5 for 4 yards each. Think that's why YPA works better.
Maybe a theory: efficient passing allows more conservative playcalling, meaning more runs in the mix to control clock and keep the defense off balance. This theory seems to fit Burke's analysis that more conservative playcalling leads to wins.
This also suggests that running to get into easier passing situations is important to establish an efficient passing game, and in turn to winning. Teams should play first and second somewhat conservatively against base defenses.
Any thoughts on why the Eagles under perform compared to their point differential? I think they over performed in the early Reid years, but I'd have to double check FO.
- Behan
Watching Atlanta, they seem to have the type of offense we're talking about. Number four in yards per completion, few sacks or picks. They're number nine in yards per carry, but second in rushing attempts. Their running game is only thirteenth in DVOA, passing game fourth. So they are an efficient passing team that gets lots of rushing attempts even though they're not as stand out in rushing efficiency.
Their balance is really skewed to the run, but they could get scary as they trust Ryan more and open up the offense in future seasons.
- Behan
Post a Comment
<< Home