wiz, i am going to make sure david lee
is playing b4 i lock this sucker in.
is playing b4 i lock this sucker in.
-
Comment -
In it for marquette and purdue...let's get itComment -
hey bean i like over 48.5 for mid tenny game 3xLETS GET ITComment -
great info must read!
By Zach Fein
Eaglesblog.net Writer
Published: May 13, 2009
Comment
In football, there are some truths that are self-evident.
Defense wins championships. A robust rushing game is key for all successful teams. You can’t win a Super Bowl without a fierce passing attack.
Given the frequency with which these notions are maintained, there shouldn’t be any disagreement with the statements. Every Tom, Dick, and Harry believes you need a franchise quarterback to win a championship. Everyone also thinks that you should only go for it on fourth down with one or two yards to go (hint: that’s wrong).
What if those self-evident truths weren’t self-evident truths? What if they were simply nonsensical lies? Has the general public been deceived over these so-called doctrines?
I went mining for answers—and I’ve found them.
THE DATA
Data were taken manually from Pro-Football-Reference’s year pages. (Why they don’t offer a downloadable team database is beyond me.)
I chose statistics for both sides of the ball in the following categories: a team summary, passing, rushing, kick and punt returns, and kicking and punting; in total, there were 180 different statistics I used in the study—about 50 of which were differences between offense and defense (such as the difference between passing yards and passing yards allowed). The sample went back 15 years—from 1994 to 2008—giving me 465 team seasons with which to work.
THE METHOD
To determine each stat’s relative value, I used to methods of choice: correlation and regression.
Correlation shows how relationship between two variables in a number between negative-one and one. The more related two variables are, and the more the graph between the two looks like a perfect line, the closer to one or negative-one the correlation is; a positive number represents a positive relationship—that is, as one stat goes up, so does the other (such as points scored and wins—and a negative number represents a negative relationship—when one variable goes up, the other goes down (such as points allowed and wins).
This chart shows several relationships and the resulting correlation.
Note: The correlations of an offensive stat and its defensive counterpart—such as offensive passing yards and passing yards allowed—are nearly the same, and you get the same conclusion whether you look at the offensive or defensive stat. For example, in the final table in the “Passing and Rushing” section, I show the correlations of five passing statistics; on the defensive side, the correlations of those five stats are about one-tenth less as a whole, but the order of the stats based on correlation with wins is the exact same.
Regression can be used to estimate a dependent variable based on one or more independent variables. In other words, if I ran a regression based on wins, points scored, and points allowed, the resulting regression equation will be something like this (numbers as illustration only): Wins = 0.13 x points — 0.12 x points allowed + 7.5.
What regression does is take a set of data and forms a line that creates the smallest difference between actual wins and estimated wins (based on points scored and allowed).
However, if I include a stat like touchdown percentage in a regression with yards per attempt, the equation may look like this: Wins = 50 x TD% + 2 x YPA — 6.5. That may look like touchdown percentage is 25 times more important than yards per attempt, but because the scales of the two stats are much different, we must take one more approach to see which is more important to wins.
Standardization. We standardize by computing how many standard deviations a variable is from the average value. That sounds like a lot, but by doing so, we put both on the exact same and thus comparable scale in the regression. Generally, a 3,900-yard passer with 28 touchdowns is one standard deviation above the average in both stats, while 4,400 yards and 35 touchdowns are both about two standard deviations from the average value.
PASSING AND RUSHING
Running the ball wins games more often than passing.
That statement is true; since 1995, in fact, teams have won 84 percent of the time when one running back has 30 attempts, but they’ve won just 28 percent when their quarterback has 40 attempts. This is even backed up by the correlations of pass and rush attempts to winning.
Stat Win Correlation Rush Attempts 0.55 Pass Attempts -0.16 The problem lies, however, in the analysis. Running does not lead to wins, and passing does not lead to losses; it’s the other way around.
When a team is up by 10 points in the fourth quarter, you bet they’ll rush the ball to run out the clock. Similarly, a team down 10 will in no way make an attempt to run. They’d pass as much as they can in order to start a comeback.
So when analysts see that Arizona is 9-1 when rushing the ball 19 or more times, and 0-6 in all other games, they’ll point out that the Cardinals just need to run the ball to win—even though they may actually need to pass first to get a 10-point lead, when they can then run the ball.
There’s a similar problem when looking at pure yards. Teams have won 74 percent of the time when they have at least one 100-yard rusher, and only 53 percent of the time when they have a 300-yard passer—but that doesn’t mean that teams should go out and try as hard as they can to have a 100-yard rusher and not 300 passing yards.
That’s why it’s best to look at efficiency stats, such as yards per attempt. When teams have at least one running back with 10 attempts and five yards per attempt, they win 65 percent of the time; with a quarterback with 25 attempts and eight passing yards per attempt, they win 70 percent of the time.
The table below shows this, including correlations with points scored.
Stat Win Correlation Points Correlation Pass Attempts -0.16 0.09 Pass Yards 0.31 0.64 Pass Yards/Att 0.58 0.80 Rush Attempts 0.55 0.30 Rush Yards 0.45 0.41 Rush Yards/Att 0.18 0.35 By dividing rushing yards by attempts, the correlation/causation issue—more of any rushing stat equals more wins—is gone. There’s also clear evidence that pass efficiency is more important than either pass yards or pass attempts.
There are, however, other passing statistics that correlate better with wins and points than yards per attempt, and all take the standard yards per attempt and add on to it.
There’s net yards per attempt, which subtracts sack yards from pass yards and adds sacks and to pass attempts; the popular quarterback rating, with its arbitrary, complex formula; adjusted yards per attempt, which adds 20 yards for each touchdown pass and subtracts 45 for each interception to passing yards; and adjusted net yards, which is adjusted yards per attempt with sacks and sack yards added in.
Stat Win Correlation Points Correlation Adj Net Yards/Att 0.65 0.84 Adj Yards/Att 0.64 0.83 QB Rating 0.62 0.81 Net Yards/Att 0.61 0.81 Pass Yards/Att 0.58 0.80 TOUCHDOWNS AND TURNOVERS
Touchdowns are an odd thing.
At first, it appears that passing touchdowns are the result of luck and surrounding talent. A quarterback gets credit for a touchdown if the wide receiver runs 70 yards to the end zone on an eight-yard slant route; a quarterback loses a touchdown if the wide receiver gets pushed out at the two-yard line on a perfect, 40-yard throw.
Despite that, touchdowns and passing yards have a 0.73 correlation—and that number rises slightly if you subtract sack yards and 45 yards per interception from passing yards.
Clearly, passing touchdowns are affected by pure passing production as much as—if not more than—luck and surrounding talent.
Rushing scores, on the other hand, seem to be the result of solely running back talent. There are some situations in which a team can set up a three-play, 70-yard drive and get down to the two-yard line for the running back to punch it in. But the majority of the time, the running game keeps the drive alive or sets up the score.
Yet, the correlation between rushing touchdowns and yards is just 0.61, showing that there’s more to rushing touchdowns than just rushing yards, at least compared to passing.
And then, something weird happens—our intuitions were correct after all.
Stat Win Correlation Points Correlation Pass Yards/Att 0.58 0.80 Pass Touchdowns/Att 0.55 0.78 Rush Touchdowns/Att 0.41 0.61 Rush Yards/Att 0.18 0.35 What this shows is that rushing touchdowns are worth more than the yards it takes to get them, and vice versa for passing. The complaints about quarterback rating that argue about its inclusion of passing touchdowns are well-founded, to say the least.
One thing quarterback rating fails to take into account is fumbles. Fumbles may arguably be worse than interceptions, because a defense starts out with the ball after a fumble about 15-20 yards ahead of where it would have started after an interception. Most fumbles occur near the line of scrimmage, while interceptions happen 20 yards downfield, depending on the pass.
It should be of no shock, then, that fumbles correlate better with wins than do interceptions. There will be a shock, though, when it comes to how many wins each turnover is worth—that will come in Part II.
(I define fumble rate as fumbles divided by fumble chances, or completions plus rush attempts plus sacks. Fumble rate did have a higher correlation with wins if I divided fumbles just by rush attempts, but although leaving out completions and sacks results in a higher correlation, it doesn’t provide a real estimate of fumble rate.)
Stat Win Correlation Fumble Rate -0.45 Fumble Lost Rate -0.42 Interceptions/Att -0.40 It’s also beneficial to use fumbles instead of fumbles lost. Many studies, including some by Football Outsiders, have said that forcing a fumble takes skill, but recovering it is based solely on luck; my data says the same.
The year-to-year correlation of fumbles is 0.29; for fumbles lost, it is 0.13; and for fumble lost percentage, it is 0.01.
DEFENSE AND SPECIAL TEAMS
I said in describing the method to the madness, “The correlations of an offensive stat and its defensive counterpart—such as offensive passing yards and passing yards allowed—are nearly the same, and you get the same conclusion whether you look at the offensive or defensive stat.”
I haven’t mentioned any defensive stat up until now simply because there’s no point—for every offensive stat I took from Pro-Football-Reference, I also have its defensive equivalent.
I decided to look instead at tackles (total, solo, and assisted) to see if they had any bearing on wins. Since my final goal in this is to evaluate players on both sides of the ball, I had to use a common defensive statistic.
I don’t think I will anymore.
Stat Win Correlation Points Allowed Correlation Solo Tackles -0.32 0.35 Total Tackles -0.21 0.17 Assisted Tackles -0.00 -0.07 You read that right—as tackles go up, wins go down and points allowed goes up. I found that a tackle was worth about negative-0.06 wins (not by itself, of course, but in a regression with several other stats), which means that if I were to apply this to individual players, Patrick Willis’ 141 tackles would have been worth almost negative-8.46 alone last year.
As for special teams, there aren’t many stats that correlate well with wins, except for useless stats such as extra points or kickoffs (which are both about equal to touchdowns). The table below shows some special teams stats along with their correlation with winning.
Stat Win Correlation Field Goal % 0.24 Punt Average -0.14 Punt Return AVG 0.13 Kick Return AVG 0.11 Kick Return TD 0.05 Punt Return TD 0.04 A Devin Hester or Josh Cribbs certainly can help your field position when needed most, but the fact is, there’s so few punt or kickoff returns to make the effects any worthwhile. As well, the difference between one or two return touchdowns per years is meaningless, considering it takes about 34 points to add on an extra win (that’s explained in the next article).
In Part II of this series, I’ll use regression to estimate wins based on stats and apply that to the player level
Comment -
good infoComment -
Running the ball wins games more often than passing.
That statement is true; since 1995, in fact, teams have won 84 percent of the time when one running back has 30 attempts, but they’ve won just 28 percent when their quarterback has 40 attempts.Comment -
purdue/psu o/u 126 over 2x
mich/wisc o/u 116.5 over 2x
purdue -6.5 1x
bucks +12 2xComment -
celtics -2 1x just cause ill be there
BOL fellasComment -
hey ayo where you from ?Comment -
really like all your plays buddy glComment -
nba
bucks +12 1x
no celtics i bet the spurs +3 1x
celtics/spurs o/u 190 under 1x
bulls -5 1x
nhl
atl thashers 1x ml
college hoops
fordham+17 1x
purdue -6 1x
charlotte +14 1x
uab/duke o/u 142 over 1x
william+ mary +7 1x
dayton +1 1x
parley celtics under 190 spurs +3 and the bucks +12 .2x
Comment -
bean i love lasalle tonight getting 2 against an awful george washington team.... and for everyone to know red hot dayton is playing st louis who is not playing very well and will b without there two best players tonight bc of pending sexual assault chargesComment -
also gonna go memphis tonightComment -
ya on dayton thanks i like laslle no money to bet it .good luck guysComment -
be back for the games laterComment -
nba
utah -5 1x
college hoops
n texas -1 1x
unlv -4 1x
unlv over 148 1x
memphis +7 1xComment -
nba
suns +5 1x
denver ml 2x
jazz -51x
Comment -
see ya tomorrow good luckComment -
wiz, cash that golden state warriors +6.5
and the +250 on the ml.
Comment -
cash purdue and the over
Wiz you shoulda seen the garden tonight it was bumpin!!, GREAT game to be at and GREAT win for the celts... Good to be in boston for a night i live in philly now but grew up in CT.
See ya tommorowComment -
wiz, liking blues and kings in
a nhl parlay
Comment -
wiz, cover alert, liking the seahawks
+ 10.5, saints will be down to their 3rd string
running back, both 1 and 2 will not play.
I like the points here.
Comment -
what up crew lets get it
LETS GET ITComment -
hey guys ,ayo looked like the foundation was about to cave in at the garden !ray ray missing the free throws wow two in a row!Comment -
thunder -1.5 3x
kings +4.5 3xLETS GET ITComment -
...Comment -
ya sonny i like the kings but are weary of the thunder and the way they have been playing lately !the kings are finally playing the right peeps off there bench!Comment -
hey in the nfl i like the jets and the over 44.5 and the pts and sea are looking interesting and 44.5 i like the under!what do you guys think .Comment -
but the fact that gray bracket is back for indy defense is huge!He's the missing link to that defense and the general who calls there plays !THat was a huge blow with him going down and bob sanders!plus the fact that the jets have lost 3 out of the last 4!Comment -
ha couldnt believe ray missed BOTH, worked out in the end but close call.
xavier +9 1x
cofc +2.5 1xComment -
no team that won the super bowl the year before has gotten past the division championship game the next year!Comment -
hey how is xavier this year?ayo?Comment -
ya sonny going to look at that over right now buddy!for the bowl gameComment
