So far this week, the team that has run more time than its opponent is 3-0 (Ravens, Lions, Cowboys).
Interresting stats, follow up on offense balance.
- Thread starter NBP81
- Start date
2413fanphins
Active Roster
of course, we have to account that green bay, and oakland both suck. so does pittsburgh as far as that goes.
I'd say in the case of those three games, it's just the better overall team won. the three losing teams are a combined 12 games under .500
nyashfan
For Earth Below
Shouright & ATL_PHIN_FAN,
I'm glad this point was raised. My gut reaction is that rushing attempts differential is correlated with time of possession; however, my gut reaction is that time of possession is uncorrelated with win percentage. Why? There are too many instances in which games are decided by huge plays. Not just long passing plays, but especially special teams return touchdowns and defensive touchdowns. These occurrences will reduce dramatically such a team's time of possession advantage while dramatically increasing a team's rushing attempt differential and win percentage.
This is another reason why I doubt the causal relationship between rushing attempt differential and win percentage. To do so, you would have to eliminate games in which these unanticipated game-changing plays occur.
When such plays alter the outcome of a game, the team negatively impacted by such events must throw the ball almost exclusively while the benefitting team can then run the ball without needing to incur the risks associated with passing e.g. interceptions, incompletions, clock stoppage, sacks, etc.
Shouright - can we further this analysis by calculating the correlation between a) time of possession and rushing attempt differential and b) time of possession and win percentage?
I'm glad this point was raised. My gut reaction is that rushing attempts differential is correlated with time of possession; however, my gut reaction is that time of possession is uncorrelated with win percentage. Why? There are too many instances in which games are decided by huge plays. Not just long passing plays, but especially special teams return touchdowns and defensive touchdowns. These occurrences will reduce dramatically such a team's time of possession advantage while dramatically increasing a team's rushing attempt differential and win percentage.
This is another reason why I doubt the causal relationship between rushing attempt differential and win percentage. To do so, you would have to eliminate games in which these unanticipated game-changing plays occur.
When such plays alter the outcome of a game, the team negatively impacted by such events must throw the ball almost exclusively while the benefitting team can then run the ball without needing to incur the risks associated with passing e.g. interceptions, incompletions, clock stoppage, sacks, etc.
Shouright - can we further this analysis by calculating the correlation between a) time of possession and rushing attempt differential and b) time of possession and win percentage?
Last edited by a moderator:
2413fanphins
Active Roster
so am I to believe that pittsburgh lost to baltimore because baltimore had 5 more rush attempts than pitt?
looking at the box score pittsburgh basically won every single stat category they keep track of.
first downs, total yards, time of possession, 3rd down efficiency, 4th down efficiency, average gain per play, net yards rushing, yards per rush, net yards passing, sacks taken, punt average, penalties, penalty yards, and fumbles... they did tie for total plays...
looking at the box score pittsburgh basically won every single stat category they keep track of.
first downs, total yards, time of possession, 3rd down efficiency, 4th down efficiency, average gain per play, net yards rushing, yards per rush, net yards passing, sacks taken, punt average, penalties, penalty yards, and fumbles... they did tie for total plays...
OK, the week prior, every team who ran the ball more than its opponent won - with three exceptions: MIN/GB was a tie, and both Manning brothers lost. That's a record of 10-2-1 (not counting the TB/DET game where both teams rushed the ball 24 times).
If you're interested in the breakdown from week 12, here it is:
NO/ATL
25-22
TB/DET
24-24
MIN/GB
43-34
JAX/HOU
28-21
SD/KC
27-18
CAR/MIA
28-17
PIT/CLE
34-16
STL/CHI
29-26
BAL/NYJ
31-28
TEN/OAK
29-23
ARI/IND
30-15
DAL/NYG
20-30
NE/DEN
31-48
SF/WASH
33-27
nyashfan
For Earth Below
Not to get too technical, but if this calculation was done as a regression analysis I don't think that is optimal. What we need to do is a one-way ANOVA, splitting the data by discrete levels of turnover differential e.g. minus 2 or worse, minus 1, even, plus 1 and plus 2 or more. Then we can calculate 5 separate correlation values between rushing attempt differential and win percentage for each group.
nyashfan
For Earth Below
Shouright,
See my earlier post regarding my hypothesis that time of possession is uncorrelated to winning due to the presence of unanticipated game-changing plays e.g. defensive and special teams touchdowns.
Here, Green Bay jumped to a 14-0 lead due to the benefit of a pick-six TD. I believe these rare events alter the relevance of applying statistics to predict the outcome of such games.
It's like those academics who model the stock market as following a lognormal distribution and their model falls apart due to the rare occurrence of unanticipated crash events, which are impossible to predict.
See my earlier post regarding my hypothesis that time of possession is uncorrelated to winning due to the presence of unanticipated game-changing plays e.g. defensive and special teams touchdowns.
Here, Green Bay jumped to a 14-0 lead due to the benefit of a pick-six TD. I believe these rare events alter the relevance of applying statistics to predict the outcome of such games.
It's like those academics who model the stock market as following a lognormal distribution and their model falls apart due to the rare occurrence of unanticipated crash events, which are impossible to predict.
Last edited by a moderator:
HoneyB
Active Roster
It looks like that list is filtered by rushing yards, not rushing attempts.
From that list, you've got 8 teams under 1,000 yards:
25 Arizona 976
26 Baltimore 973
27 NY Giants 972
28 Miami 939
29 Pittsburgh 922
30 Cleveland 891
31 Atlanta 822
32 Jacksonville 735
Most of these teams have losing records. Also, 7 out of the 8, are in the bottom half of their division.
Let's take defense out of the equation. Who scores the least points per game? Let's look at the bottom half of all 32 teams, and see if we can find those 8 teams in there. If they're NOT, and they're in the top half, then there's no correlation. If they are, then that makes 2 correlations I can observe. I'm looking now...
Ok....yes, they are ALL in the bottom half:
18 Arizona
21 Pittsburgh
23 Miami
24 Baltimore
25 Atlanta
27 Giants
29 Cleveland
32 Jacksonville
What would be the rationale for that? Are you talking about on a per game basis, or overall on the season? What I did earlier in the thread was a partial correlation between winning percentage and rushing attempts differential, controlling for turnover margin on the season. I assume you're talking about a per game analysis here?
http://www.nfl.com/news/story/09000...tial-an-essential-indicator-of-success-in-nfl
Also, however, I think you have to plan your strategy on the basis of the absence of such rare events, and then adjust accordingly if they do happen.
Notice also that despite the fact that Green Bay went up 14-0 on the basis of a rare play, Pittsburgh kept running the ball.
Hadn't seen this. Interesting article. I'd be interested to see how rushing attempts correlate with that toxic differential. I'd imagine the correlation is strong - given that running the ball helps you control the game (giving the other team fewer opportunities), opens up big plays in the passing game, and limits turnovers.
I think we're getting a little lost in the weeds. We still have a team that's near-last in the league in its percentage of offensive plays that are runs, which is also using a developmental QB who isn't by any means lighting the world on fire, and has a zero turnover differential on the season. I think we're going to struggle mightily to find anything that adequately and justifiably explains that degree of abandonment of the running game. Certainly the performance of the running game alone doesn't do it, and I'm not sure we should be going beyond that much if at all, in terms of examining other variables.
I mean we can certainly put me on a statistical wild goose chase, and it would be fun to collect the data and click the buttons and see what the results are, but I'm not sure there's any sensible theoretical rationale for that. :)
nyashfan
For Earth Below
I'm looking to see if there is a monotonic relationship between turnover differential and the correlation between win percentage and rushing attempt differential. If the relationship is monotonic, it suggests that turnover differential is the driving force. I can't tell that from one single number.
2413fanphins
Active Roster
I was going by rushing attempts, not the order the poster had the list done in, basically anything over 300 attempts based off his list, is what my post referred to.
nyashfan
For Earth Below
To see 5 correlation numbers, one for each turnover differential bucket, that would just be pooling all games.
To do a true ANOVA would require per game data. In this case we wouldn't examine correlations to winning pct, but simply the rushing attempt differential for each game. The ANOVA F-Test would quantify the statistical significance of the variation among the mean values of rushing attempt differentials across the 5 turnover differential groupings. If the mean values were monotonic with respect to turnover differential, then it would show how much rushing attempt differential is a proxy for turnover differential.
To do a true ANOVA would require per game data. In this case we wouldn't examine correlations to winning pct, but simply the rushing attempt differential for each game. The ANOVA F-Test would quantify the statistical significance of the variation among the mean values of rushing attempt differentials across the 5 turnover differential groupings. If the mean values were monotonic with respect to turnover differential, then it would show how much rushing attempt differential is a proxy for turnover differential.
Last edited by a moderator: