Followup to the Elway thread

[Samdari]

Giddy with excitement over an honest to goodness football discussion, I decided to waste my day analyzing a statement made in the thread about Elway's passer rating.

In that thread andy m quotes a book called "the hidden game of football" in which it is claimed that a high yards per attempt stat correlated highly with winning. I decided to put that to the test. I have some interesting data worked out, but first, I'd like to hear guesses (DOH!, I meant to say informed opinions) from the peanut gallery on the following topic:

Which stat most closely correlates with winning (other than points scored, hopefully to head off some jokes). In other words, the team with the highest {BLANK} wins the highest percentage of games?

[Craptacular]

Where's Gene Rayburn?

[Craptacular]

Interceptions by the defense

[TroyF]

My thoughts would be, in order:

Yards per pass attempt.
Yards per play.
TD conversion inside the Red Zone.

TroyF

[dawgfan]

I'll take a shot: turnover ratio (takeaways vs. turnovers)

[mrsimperless]

YPA + (YPC * 3)

[Daimyo]

I would guess passing yards per attempt and turnover ratio are the big ones.

[revrew]

Considering how successful Oakland and Dallas have been through the years, I might suggest the key factor is how hot the cheerleaders are. That might also explain why the Bears' glory days are so long ago. They haven't had cheerleaders. Of course what happened in '85? Then again, McMahon and Payton were so pretty...

But why haven't we seen more victories in Philly?

Shoot. There goes a perfectly good theory.

[Samdari]

Originally posted by TroyF

TD conversion inside the Red Zone.

TroyF

I did not have access to this data.

I had rushes, Ruyds, fum, RushTD, PA, comp, PAyds, int, PATD, sacks and score.

I am not going to post my results until tomorrow noonish, to give people a chance to respnd/argue first.

[pjstp20]

That's easy, I'd say rushing yards. Although as a Miami Dolphin fan I guess I shouldn't be so sure.

[tucker342]

turnover ratio is probably the major one

[cuervo72]

I say rushing attempts.

[QuikSand]

I think rushing attempts is a possible sleeper in that list, too. I might have gone with turnover ratio, but it doesn't seem to be exactly there (in the list posted above), and unless it's a margin or ratio (not just # of ints), I wouldn't think it would be enough to be the biggest correlation.

[vtbub]

turnover ratio, yards per attempt, 3rd down conversions?

[Draft Dodger]

my vote is rushing attempts

[kcchief19]

As usually, I have to go with QS. From the listed stats, rushing yards is probably the conventional wisdom. Sacks allowed and interceptions could be high on the list, but I think rushing attempts could be the suprise indicator.

[Daimyo]

Rushing yards makes a ton of sense for this.

[TroyF]

I don't need to look at stats to know that rushing attempts were fairly meaningless last year. Neither Tampa nor the Raiders ran the ball that often or did it with any sort of effectiveness on a consistent basis.

We can also safely assume that in 2000, neither the Ravens nor the Giants were all that solid in yards per pass attempt.

After thinking it over, I'd say rushing attempts would be the best indicator for one main reason: The best teams have the lead in the fourth quarter and run the ball more. The bad teams are throwing the ball late.

I'm not sure it really would indicate that they run the ball BETTER, more that they have a chance to stick to the run and use it to eat up clock.

I wonder what it would be if we took yards per carry. . .

good stuff, I love looking at things like this.

TroyF

[vex]

Rhett Bomar

[Chief Rum]

I'll say margin of victory.

Chief Rum

[Samdari]

Congrats to everyone but Chief Rum for holding off on the lame joke.

I am now scrambling to compute win% by some of the measures you guys have mentioned that I did not even think of. I'd love to look at a TD/FG ratio kind of thing to see how try TroyF's red zone effectiveness suggestion worked out, but don't think the files I downloaded have enough information.

[Samdari]

Originally posted by vtbub
3rd down conversions?

Don't have this either, but it would be interesting.

[henry296]

Samdari,

I assume you are doing this on a game by game basis and not yearly records. Also, over what period of time is your study?

I think it is Turnovers margin. At least in the playoffs the teams with less turnovers have win like 80% of the time.

Todd

[Gene Rayburn]

That football team was so good. (How good were they?)

They were so good, that during every game they would ______(Blank) their opponents on the field!

[Darkiller]

I take a shot :

1 - QB rating
Usually, when a QB ends the game with a rating = or >105-110, his team is the winner.

2 - sacks
It's difficult to imagine a defense sacking the QB = or > to 5-6 times and not being the dominant force all day long.
I guess if a defensive unit can get to the QB 5 or 6 times in a game, then they put pressure on him all day and he makes others mistakes (can lead to bad throws being intercepted -fear of the pass rush-, fumbles etc..) this lead to good field position when the opposite offense comes back on the field.

I would say a sack has a more psychological effect than an interception because on a sack : the entire Offensive line feels guilty, the QB is scared that it might happen again...while the interception is more of a one-shot mistake from one individual (usually the QB himself) but he can shake that quickly...not the same when he's sacked...fear comes into the mix.

3- Rushing attemps
When a team pounds the ball on the ground it means it has confidence in running the clock, in advancing through short gains...means this team feels good enough to win...so they're often the superior team (or believe so).

[albionmoonlight]

I also agree that rushing attempts may have more to do with being up by 14 in the fourth quarter than with using rushing attempts to get up by 14 in the first place.

I'll guess yards per pass attempt. My secondary guess is low fumbles and interceptions. If you do not turn the ball over, you win (or so I have been told).

The great thing about sports debates is that--whatever the answer is--we will immediately be able to think of a team that won the Super Bowl some year that flies in the face of the trend.

[Samdari]

Originally posted by Darkiller
1 - QB rating
Usually, when a QB ends the game with a rating = or >105-110, his team is the winner.

DOH! You are killing me. I redid the analysis to evaluate some previous guesses, but not this one. I have enough raw data to compute this figure, but not enough time to program in the formula today. I will play around with this when I revisit the program to analyze the playoff games (my results so far are only for the regular season).

I will leave this open for people's answers for another hour before showing some results. I appreciate everyone who has chimed in so far (especially my hero, Gene Rayburn).

[albionmoonlight]

Of course--OB rating is a combination of raw stats--so in some ways it is not quite on par with the others. It seems to have a bit of an unfair advantage cooked into it because it gets to take lots of stuff into account. Still, it would be interesting to see where it falls on the list (says the guy who is not doing any of the work.)

[Samdari]

And the envelope please.

First, a description of what I did. I took data from a file containing raw data (fields listed above) for each regular season game since 1993. Then I computed how many games each team that "won" a certain measure also won the game. The measures computed were:
Yards per pass attempt
Rushing Attempts
Turnovers
Sacks
Rushing Yards
Passing Yards

And in respnse to some guesses, I went back and computed wins based on the following:
Yards per play
Interceptions
Yards per Carry
Yards per Attempt + 3*(Yards per Carry) - hereby known as "formula"

There was one sticking point for me, and that was stats which normally have low integral values, and thus tie often, i.e. sacks, ints and turnovers. In order to preserve the, "when a team WINS this stat, they win the game flavor" of the numbers, games in which a stat tied were ignored when computing the win % for that stat. For example, in 454 pf 2408 games, the teams had equal turnovers. Of the 1954 games which one team had fewer turnovers, that team won 77.69% of the time.

For Gene - we polled a studio audience of 2408 regular season games. The most popular answers were:

Rushing Attempts - 79.86% of the time, the team with the better number won
Turnovers - 77.68%
Interceptions - 75.35%
Rushing Yards - 72.30%
Yards Per Pas Att - 70.93%
Sacks - 70.50%
Yards Per Play - 67.57%
Formula - 61.84%
Passing Yards - 54.53%
Yards per Rush - 51.95%

After seeing the commentary, I would love to have 3rd down conversions, and some sort of red zone effectiveness measure included. I strongly suspect that success in either of those measures would highly correlate with winning. I am dying to see those results, but leave that as an exercise for the reader.

Note that this includes ONLY the regular season games. I do in fact have the RAW data for all playoff games in the same time period, but did not include them in my computations. I may indeed go back and do so and present the results later, as I am curious to see if any trends change in the playoffs.

[/End Analysis]

[Sam's Conclusions/Commentary]
I have long thought that rushing attempts was the key stat. I expected turnovers to be second as well. There is certainly an amount of chicken/egg debate to had about the RA figure, but you cannot deny the actual correlation with winning. I am also not surprised to see that passing yards is so relatively unimportant, as the low win% of QBs who throw for 300 yards is oft-discussed.

The shocker there for me is the Yards per Rush. I expected that to be up there with rushing attempts and turnovers. I reran that about 10 times, checked my formula and such, and it appears true. The lesson here appears to be that it is more important to be committed to running the ball than to be good at it?

I do not mean to pick on TroyF here, as I value his opinion, and he has been extraordinarily helpful in the fantasy football league he pointed me to, but I also evaluated a few of his statements. Again, I do this not to pick on him, but he made the most declarations, so he has the most to evaluate.

"I don't need to look at stats to know that rushing attempts were fairly meaningless last year."
77.73% of the teams with more rushing attempts won games last year, and 8/11 (72.73%) in the playoffs. This is the second lowest number in the 9 years, with 1993 coming in (insignifacantly) lower at 77.23%. There is very little year to year variation in this number. Between 77-83% of teams with the most rushing attempts won games in any given year. The consistency of this figure is staggering to me.

The true shocker of 2002 is that the Raiders played 19 games, and 15 of them were won by the team with the most rushing attempts. Clearly this is evidence supporting the 'egg' side of the debate, as the 2002 Raiders could hardly be described as committed to the run. Also notable is that the higher yards per pass attempt team won 18 of those 19 games.

The Buccaneers, also thought to be a passing team, fell in line with both trends. They played 19 games, 16 won by the team with more rushing attempts, and 17 won by the team with the highest yards per pass attempt.

I suppose that it IS meaningful that for both teams, yards per pass attempt were stronger indicators than rushing attempts, but I will leave it to everyone else to explain to me exactly how and why.

"We can also safely assume that in 2000, neither the Ravens nor the Giants were all that solid in yards per pass attempt."

I did not compute how good they were relative to the league average in these stats, but, I can tell you they were better than their opponents.

In 2000, the Ravens played 20 games. 14 were won by the team with the higher yards per pass attempt. 16 were won by the rushing attempt leader. Both numbers are shockingly close to the 9 year leaguewide average.

The Giants that year played 19 games. 16 won by the team with higher yds per pass, all 19 by the higher rush attempt team.

Troy's statement may indeed be true, both teams may have simply been average at yards per pass attempt themselves and great at stopping the other guys, but they both often won that statistic.

And if last year's Raiders don't fly in the face of the strongest trend (rushing attempts) albionmoonlight, I simply cannot imagine who might have. I now wish I had some older data, to look at the Warren Moon Houston teams.

To be investigated on a future pass: QB rating, run/pass ratio, and playoff data.

[clintl]

I wonder, though, whether rushing attempts is a cause of success, or an effect of success. As has been stated, teams with leads in the second half resort to the run to eat up the clock. Thus, rushing attempts may not really be much of a factor in gaining and building leads, and are being used by good teams to protect leads they have already established. Whereas, teams that fall behind early are forced to resort to the pass, depressing their rushing attempt totals.

[Draft Dodger]

well, I was right, but here's the kicker.

I'm pretty sure that more rushing attempts do indicate which team won, but it's an effect, not the cause.

the team with the lead is going to rush the ball more. the team without the lead is going to pass.

so, in short, I think it's safe to say that:
"teams that win games are going to have more rushing attempts"

I DON'T think it proves the converse that:
"teams that attempt more rushes in the game will win games"

I'm sure some of the smarter members could have worded that better.

[henry296]

DD,

In statistics the saying is Correlation does not Equal Causation.

Todd

[Samdari]

Originally posted by henry296
DD,

In statistics the saying is Correlation does not Equal Causation.

Todd

Thanks Todd, I was going to post a very long winded message saying exactly that, but you summed it up far more eloquently than I could have.

[Travis]

Which would also explain the apparent drop in importance for yards per rush, as those teams rushing more at the end of the game tend not to be picking up big yards so much as just preventing turnovers and looking for a safe yard or two instead of trying to spring one.

[Travis]

Dola, I think in the end, especially after realizing the cause/effect angle of this, it really does show the overall importance of not turning the ball over as turnovers and interceptions suddenly become the top two factors percentage wise towards winning.

[TroyF]

Sam,

A couple of things here:

1) The Bucs and Raiders finished tied for 23rd in rushing attempts last year. 414 each. Tampa was ran on 410 times. Oakland only 384.

2) The Ravens finished with a putrid 6.2 yards per pass attempt while the Giants merely finished middle of the pack. Both excelled and were in the top 8 in yards per pass attempt defense.

What does this tell us? Well, for one, there is a stat we didn't look at which may help us out a little more: Time of Possession. Or maybe plays ran per game. (both combine how solid your defensive presence is along with how good you control the ball)

TroyF

[QuikSand]

While I agree with most of the above, let's not forget on other component about the presumed strong correlation between rushing attempts and winngin games. Rushing attempts are a subset of total offensiev plays - the team decides what share are rushes, but teams with more offensive plays get more opportunities to rush the ball, of course.

Having more offensive plays is presumably pretty strongly correlated with lots of things that are themselves indicative of success. Good yards per play translates to first downs, which gives you more plays to call. Good defense gets your offense the ball back, and more offensive plays. Causing turnovers, same. All these things are elements of winning football, and they correspond to more offensive plays (some of which are going to be rushing plays).


If we really wanted to understand all this better (again, coming from a guy not doing any of the work), maybe we'd need to look at the correlations for a few more stats:

Total Offensive Plays
Total Passing Plays
% of rushing plays called


This might help to shed some light on the Bucs of 2002 - maybe their good defense just helped them run more plays than most teams, even if theur run/pass split wasn't run-inclined.

And also, it suggests that perhaps run *defense* wins games, too. A team good enough to shut down the opponent's running game will have more rushing plays than the opponents, who resort to the inherently more risky passing game.


Lots of angles to this, some of them more subtle than "leading team eats the clock."

[Samdari]

Well, in the case of all four teams, why they had more success when beating their opponents in those stats appears to be that they were better at stopping their opponents from doing something than they were at doing it themselves. The statistics I computed make no measure of how doing these well over a season translates to wins, but how well doing something better than your opponent in a given game translates to wins.

Again, I am only the messenger. And I maintain that the statement "Rushing attempts did not matter last year" does not hold up under examination.

No matter where they finished in total attempts, the Raiders had more success when they outrushed (by attempts) their opponents. They had more rushes in 9 of 19 games last year, and won all 9. In games where their opponents had more rushes, they were 4-6. The Buccaneers had more rushes than their opponents in 12 of their 19 games, and again won all 12, leaving them 3-4 when their opponents had more rushes. The Bucs ran far more than their opponents in every playoff game as well, averaging more rushes per game (37.3) than anyone had during the season.

You can argue the cause all you want, the fact remains that both teams were most successful when they ran the ball more than their opponents.

EDIT: removed a paragraph which was a poorly worded version of QS's point on defense above.

[Fritz]

I would suggest looking at (Yard Per Play - Allowed Yards Per Play)

[cuervo72]

After a year of reading TMQ hammer the point of having more rushing attempts than your opponent being an indicator of success, I figured it might be a good guess

[Samdari]

Originally posted by Fritz
I would suggest looking at (Yard Per Play - Allowed Yards Per Play)

My yards per play essentially does that. I compare yards per play vs opponents yards per play on a game by game basis. Team A's yards per play is exactly team B's yards per play allowed. If a team's yards per play is higher than an opponent's, then by definition (yds per play - yds allowed per play) will be too.

[Samdari]

Dola

Originally posted by cuervo72
After a year of reading TMQ hammer the point of having more rushing attempts than your opponent being an indicator of success, I figured it might be a good guess

Do you remember his Super Bowl prediction that whichever team surprised the other with a running game would win? The Bucs ran the ball 42 times that day.

[cuervo72]

Originally posted by Samdari
Dola


Do you remember his Super Bowl prediction that whichever team surprised the other with a running game would win? The Bucs ran the ball 42 times that day.

I'd actually forgotten that....football season seems so long ago

[TroyF]

Originally posted by Samdari
Well, in the case of all four teams, why they had more success when beating their opponents in those stats appears to be that they were better at stopping their opponents from doing something than they were at doing it themselves. The statistics I computed make no measure of how doing these well over a season translates to wins, but how well doing something better than your opponent in a given game translates to wins.

Again, I am only the messenger. And I maintain that the statement "Rushing attempts did not matter last year" does not hold up under examination.

No matter where they finished in total attempts, the Raiders had more success when they outrushed (by attempts) their opponents. They had more rushes in 9 of 19 games last year, and won all 9. In games where their opponents had more rushes, they were 4-6. The Buccaneers had more rushes than their opponents in 12 of their 19 games, and again won all 12, leaving them 3-4 when their opponents had more rushes. The Bucs ran far more than their opponents in every playoff game as well, averaging more rushes per game (37.3) than anyone had during the season.

You can argue the cause all you want, the fact remains that both teams were most successful when they ran the ball more than their opponents.

EDIT: removed a paragraph which was a poorly worded version of QS's point on defense above.

Sam,

Not arguing this point at all. Just stating that the teams didn't rush the ball nearly as much as the league average. . . therefore it becomes more complex than just stating Team A ran more than Team B, thus they win.

I think what it shows is that if you don't have a good ground game, you'd better damned well have a dominating type of defense to make up for it. I'm just trying to find any outside reason for this. The big question in my mind is this: Is more rushing attempts a PLAN for all successful teams or is it a RESULT of successful teams.

I think the two are very different things. Lets use an example of one game from Oakland last year. The Buffalo game. The final stats show the Raiders with 27 rushing attempts to the Bills 18.

What it doesn't show is that Rich Gannon ran the ball 11 times in the game. (3 coming on the last series of the game with kneels) It shows that Bledsoe ran the ball twice. Thus, the 9 rush difference exists.

Which team made a calculated decision to run the ball? (doesn't look like either especially cared about the run) In other Raider games I looked at over the year, blowouts early led to big ground totals discrepencies. (The Raiders had huge leads against Seattle, Denver, Tennessee, and Arizona among others. They gained those leads due to the passing game and simply ran out the clock for 30 minutes.) In the Raiders case, I think we have a case of the rushing attempts being the result of having a good team, rather than a planned out form of attack. Had some of those blowouts been close, I have no doubt the Raiders would have aired it out more and had the ground numbers be a lot closer.

Again, not knocking the numbers at all, just trying to add to the discussion about some meaning behind them.

TroyF

[vtbub]

Still would think that teams would have more success the longer they keep their defense off the field.

Teams that can convert on third down, especially on third and longs, must win more.

[Vince]

So from a lay bystander (who doesn't hope to compete with the football knowledge being brandished here, but will try nontheless ), it seems that the argument about the 'chicken or the egg' of rushing attempts lies in the fact that the rushing attemtps garnered at the end of the game which are intended to gain 2-3 safe yards, but run off 40 seconds of clock skew the numbers towards this being an effect of success rather than a cause, because that wasn't part of the original gameplan. Is that a correct assumption (it sure is a long enough sentence)? If so, I don't know if the 'debate' is an answerable one.

Basically, my opinion is that there are two possibilities.

• Winning team is ahead for most, if not all of the game.
• Winning team is behind or game is tied until late in the game.

In case one, the winning team is going to try to 'run out the clock' the majority of the time, gaining them rushing attempts that skew statistics like these we are looking at. These rushing attempts are not a cause of winning, they are an effect. Because the team is winning, they can afford to just run the ball, and drain the clock.

In case two, however, the team that is behind is less likely to run the ball because of clock management. Passing is a better option, and rightfully so. If a team that wins in this manner still has a greater number of rushing attempts than their opponents, is it fair to assume that the number of rushing attempts played a great role in their win?

I'd venture to guess that the majority, if not all of the 21% of games when the winner had less rushing attempts than the loser fell under case two from above, and I'd make the same guess that case one applied to the majority of the other 79% of games. Just based on the sheer numbers, it's hard to argue with 79%. However, I feel that it is nearly impossible to differentiate between 'meaningful' rushes in the game, and those 'meaningless' clock wasters at the end of the game. I'm also not sure it's entirely fair to label those plays as 'meaningless;' after all, the team earned the right to call those plays by being ahead, did they not?

In all, I believe that many of the instances in the 79% group had inflated rushing attempts numbers because the winning team acquired the lead, without necessarily having an emphasis on rushing, and then they ran out the clock. Therefore I believe that this is an effect of success, rather than a cause of it.

Sam, is there any way to find a distribution of the 'margin of victory' in the rushing attempts category? I'd almost suggest that any games in which the rushing attempts were within, say 3 or 4 attempts could be thrown out because of this reason, giving a better indication of the importance of rushing attempts.

[Samdari]

Originally posted by TroyF
Sam,

I think what it shows is that if you don't have a good ground game, you'd better damned well have a dominating type of defense to make up for it.

I think the two are very different things. Lets use an example of one game from Oakland last year. The Buffalo game. The final stats show the Raiders with 27 rushing attempts to the Bills 18.

What it doesn't show is that Rich Gannon ran the ball 11 times in the game. (3 coming on the last series of the game with kneels) It shows that Bledsoe ran the ball twice. Thus, the 9 rush difference exists.
TroyF

Those three kneeldowns are probably the strongest indicator I would guess that the team that has the most kneeldowns wins over 90% of the games! That information is, alas, not in my data set.

I actually think stopping the run may be the best way to consistently win football games. Unfortunately, I cannot really think of a stat to prove this better than rushing attempts, and I certainly admit that using it alone has flaws. Your example above is a great illustrator of how it can be misleading.

I also think, however, that being persistent about running the ball does lead to success. As Troy so eloquently illustrates above, that is difficult to prove using simple statistics. Sometimes the rushing attempt difference is due to wanting to run the ball, sometimes it is due to completely shutting down the opponents, and sometimes it is from getting huge leads early due to a strong passing offense (see Raiders, Rams). Some sort of contextual clues are helpful, but finding and evaluating them is beyond me.