2014 offence and defence by the numbers

[MexicoLionFan]

Great post Joe...the Vikes had all the talent in the world to win championships but they were PREDICTABLE...Bud Grant was a lot like Wally (or vice versa) in that he was NOT an X's and O's guy...he put players in their best spots and let them have at it...he kept on them for consistency of effort, which amounted to a lot of wins but ZERO NFL championships because he could be out coached, adjusted in the big game...Grant was famous for NEVER adjusting...he just got after his players to "get back to playing their game"

[South Pender]

If every TD drive this year had instead stalled at the defence's 1yd line, it would create a minimum % decrease in the total offensive yardage over what actually occurred. But scoring would drop more than 50% because all of those sevens — most of them would be sevens — would become threes at best. IOW, your question is based on a flawed assumption. You could even see scoring INCREASE against decreased offensive yardage if special teams were returning the hell out of punts and kickoffs across the board, with more drives starting already in FG range. The offence can only drive as far as the defence's goal line, however near or far away it happens to be. There's no reason why there should be any strict relationship between the two to the extent that you're implying when there are other factors that affect scoring.

Well, although the scenarios you suggest are possible in theory (e.g., drives stalling at the 1 yd. line, yielding minimal changes to total offensive yardage, but large decreases in points scored), they don't normally occur in practice. In fact we should expect a strong relationship between total offensive yardage and points scored. From NFL data (I couldn't find corresponding data from the CFL), if we use the year as the unit of analysis, there is a correlation of about .90 (which is extremely high) between (a) average yards gained per team per game and (b) average points scored per team per game, using the last 40 years of data (so that our sample size is 40) going back to the 1975 NFL season. In other words, in years in which offensive yardage was high, scoring was also high, and vice versa. No reason that I can see to expect CFL data to differ.

We're talking different things. What I'm saying is there's no formula where you can plug in an x percent change in one scoring indicator and get back y percentage change in the scoring. I've never tried to claim there wasn't high correlation (in fact I deliberately avoided use of that word), but this doesn't help when you're trying to pick some sense out of one piece of data. You don't really know if what you're seeing is an outlier from the correlation.

Re: CFL v NFL, I have no numbers to back this up, but I would expect ST to have a greater role in the Cdn game and the off yards <—> scoring correlation to be accordingly weaker.

Your statement "There's no reason why there should be any strict relationship between the two to the extent that you're implying when there are other factors that affect scoring" is synonymous with asserting the absence of a correlation between the two. In statistical data analysis, relationship between two variables is quantified as correlation. The two variables are (a) yards gained and (b) points scored. And given a non-zero correlation, we can, in fact, predict a percentage change in one of the two variables from a percentage change in the other--via a standard bivariate regression formula (the predictive use of correlation). Although this prediction won't be purrfect (this would require a correlation of +1.0 or -1.0), it will be very accurate, with little error of prediction, given the magnitude (.90) of the correlation in the present case. Re outlier detection, this is pretty easy in the bivariate case via examination of the scatterplot.

[B.C.FAN]

Significant gains in Offensive and Return Team penalties can absolutely be a factor as they negatively impact both starting field position and drive length/yardage. So let's keep that in mind as a culprit. I wonder, specifically, how much holding/procedure calls are up across the league (even despite Dean Valli getting less playing time)....

League statisticians have done some nice analysis this year since they noticed the increase in penalty frequency. Holding and illegal block penalties are up 51% in 2014 on a per-game basis. That was the biggest increase of any offensive category. The Lions took only 15 holding/illegal block penalties on offence, which was tied with Montreal for the lowest total in the league. Saskatchewan led the league with 37 holding/illegal block penalties.

Defensively, the biggest increase in penalties this year was illegal contact, which was up 75% on a per-game basis. Defensive pass interference penalties, though, were unchanged from 2013. There was nothing exceptional about the Lions' totals in these categories.

According to the league's analysis, the number of penalties has fluctuated over the past nine years from a low of 16.6 per game in 2008 to highs of 21.7 in 2006 and 21.6 in 2014. Interestingly, the two years with the most penalties were the years with the lowest scoring.

[WestCoastJoe]

Great post Joe...the Vikes had all the talent in the world to win championships but they were PREDICTABLE...Bud Grant was a lot like Wally (or vice versa) in that he was NOT an X's and O's guy...he put players in their best spots and let them have at it...he kept on them for consistency of effort, which amounted to a lot of wins but ZERO NFL championships because he could be out coached, adjusted in the big game...Grant was famous for NEVER adjusting...he just got after his players to "get back to playing their game"

Thanks, MLF.

Bud Grant was a favourite of mine. Great coach in his own right. Level, calm. He would never lose the players from over-coaching. But he did lose some big games due to vanilla preparation, it seems to some of us. On the other hand, he won a great number of championship games, just not the biggest one. The Vikings faced some monstrous teams in the Super Bowl. Chiefs. Steelers. Raiders. Dolphins. Ultra talented. Very, very well coached. Loaded with Hall of Fame players and coaches.

Bud Grant shortened Training Camp, and he did not lose sleep over any game. LOL In some ways his values were better than those coaches obsessed with preparation, to the detriment of their own health. In a big game, however, I was disappointed but not surprised, when his Vikings lost. Regular season, his Vikings usually prevailed. I thought he made the most of the talent he had. But he is a good example of a coach that held to the "execution" model, rather than the detailed Xs and Os model. Grant also left most of the practice coaching to his assistants. Some of the great ones were much more hands on.

Bud Grant's son, Mike, has an amazing record as a high school football coach. I read a very nice article a while back about the two of them. Bud frequently visits Mike at the practice facility. Be neat to hear those conversations.

Here is the article ...

http://www.startribune.com/sports/preps/220069971.html

Excerpt ... (Mike) Grant knows his obituary likely will include a comma after his name. Even now, almost every introduction begins, “Mike Grant, son of former Vikings coach Bud Grant ...”

Once, as a young teacher, he had a district administrator who was from Iowa and apparently not a football fan. The woman met Grant’s dad at an event.

“You’re Mike Grant’s dad, aren’t you?’ ” she said.

Bud didn’t flinch.

“Yep,” he answered, “that’s who I am.”

LOL Sounds like Bud Grant.

Mike Grant coaching ...

http://www.startribune.com/video/219043321.html

Mike Grant wins the Don Shula award ...

http://www.startribune.com/sports/twins ... 45901.html

[sj-roc]

Well, although the scenarios you suggest are possible in theory (e.g., drives stalling at the 1 yd. line, yielding minimal changes to total offensive yardage, but large decreases in points scored), they don't normally occur in practice. In fact we should expect a strong relationship between total offensive yardage and points scored. From NFL data (I couldn't find corresponding data from the CFL), if we use the year as the unit of analysis, there is a correlation of about .90 (which is extremely high) between (a) average yards gained per team per game and (b) average points scored per team per game, using the last 40 years of data (so that our sample size is 40) going back to the 1975 NFL season. In other words, in years in which offensive yardage was high, scoring was also high, and vice versa. No reason that I can see to expect CFL data to differ.

We're talking different things. What I'm saying is there's no formula where you can plug in an x percent change in one scoring indicator and get back y percentage change in the scoring. I've never tried to claim there wasn't high correlation (in fact I deliberately avoided use of that word), but this doesn't help when you're trying to pick some sense out of one piece of data. You don't really know if what you're seeing is an outlier from the correlation.

Re: CFL v NFL, I have no numbers to back this up, but I would expect ST to have a greater role in the Cdn game and the off yards <—> scoring correlation to be accordingly weaker.

Your statement "There's no reason why there should be any strict relationship between the two to the extent that you're implying when there are other factors that affect scoring" is synonymous with asserting the absence of a correlation between the two. In statistical data analysis, relationship between two variables is quantified as correlation. The two variables are (a) yards gained and (b) points scored. And given a non-zero correlation, we can, in fact, predict a percentage change in one of the two variables from a percentage change in the other--via a standard bivariate regression formula (the predictive use of correlation). Although this prediction won't be purrfect (this would require a correlation of +1.0 or -1.0), it will be very accurate, with little error of prediction, given the magnitude (.90) of the correlation in the present case. Re outlier detection, this is pretty easy in the bivariate case via examination of the scatterplot.

I said STRICT relationship. I'm talking in terms of y=f(x), NOTHING about correlation. When it comes to off yards and points there is NO y=f(x). OKAY?

[South Pender]

I said STRICT relationship. I'm talking in terms of y=f(x), NOTHING about correlation. When it comes to off yards and points there is NO y=f(x). OKAY?

Oh, OK. By "strict" you meant a correlation of 1.0, a condition we never encounter (except artifactually) with behavioral data. Still, in the NFL data, we come pretty close with a correlation of .90, so that we can write: Y = bX + e, with e very small.

[sj-roc]

I said STRICT relationship. I'm talking in terms of y=f(x), NOTHING about correlation. When it comes to off yards and points there is NO y=f(x). OKAY?

Oh, OK. By "strict" you meant a correlation of 1.0, a condition we never encounter (except artifactually) with behavioral data. Still, in the NFL data, we come pretty close with a correlation of .90, so that we can write: Y = bX + e, with e very small.

Yup. Don't get me wrong, I agree with you that correlation has a lot going for it. The danger comes when we start to put too much confidence (in the non-statistical sense, I suppose) in a good correlation and don't take that "e" residual into full account.

If I might relate a personal experience: for my PhD, I pursued within the same field three independent, unrelated projects. For each of these, the initial data we collected — which ultimately only formed a small percent of the total we gathered in each case — turned out by sheer fluke, to be the MOST pathological for that project in its own way (the three pathologies had essentially nothing to do with one another). These were data that with conventional means, could be easily and reasonably interpreted. But what we didn't know was that each of these initial data sets had unique confounding issues and weren't as easily understood as we first thought. So each of the projects got off to a rocky start with incorrect first analyses and this hampered our initial attempts at understanding subsequent data. For each project, it wasn't until later on when we finally had enough contradictory data, all conforming with everything except the initial run, that we convinced ourselves that our initial conclusions were wrong and we got back on the right track. In the end, those pathological initial runs served up some interesting insights in their own right, once the rest of the data gave us a better understanding from which to identify their confounding issues. But in hindsight it was clear that the work would have proceeded FAR more efficiently had we collected data in some other regime, almost *any* other regime, FIRST, and *then* gotten the other pathological data sometime later. In all three cases.

All of this is to say, the experience made me greatly skeptical of limited data sets and the conclusions we can draw from them. Overly skeptical? Maybe, but being burned three times like that was somewhat... disillusioning for a while. It was like taking three decks of cards and drawing the joker out of each one.

BTW, I had a look at scoring vs net off yards in the CFL since 1975 (I found the relevant numbers in the latest annual FFR almanac). This year is a good starting point and not only because it's the year you gave for the NFL data you cite. It happens to be the year the CFL introduced the two-point convert, but more significantly IIRC this was the year that blocking by the return team was first allowed on punt returns. You can see in the raw data that net off yards were in the same ballpark throughout the entire 70s decade but scoring underwent an abrupt, permanent increase in 1975. So best to leave pre-1975 out of it. In the NFL's case, this was around the time the goal posts were moved from the goal line to the dead line, so I'm guessing this was a factor for taking 1975 as the cutoff point in that case.

As I guessed the correlation is a bit less than in the NFL. I commented above that ST were a bigger factor in our game and speculated this would de-correlate things. I wonder also if smaller data sets are a factor? We've only twice ever had over 100 games in a season (in 1994 and 1995 during the US foray) and it's otherwise been 81 or less. The NFL season has been well over 100 games since even before 1975, and even above 200 for most (all?) of the time since. With longer schedules, perhaps things converge more strongly to the overall trend, while one or two outliers in a season would more greatly impact on where the point lands for that season when there are fewer total games to be averaged (they sure did in MY own experience!).


The yellow line joins all the points chronologically so I left it in (I was hoping excel would allow this line to progressively change shade with time to make it easier to follow but couldn't figure this out).

There's a cluster of points near the centre of the plot (50-55PPG, 660-740YPG) where scoring seems to be less sensitive to yardage than for the entire data set (i.e., the trendline would be flatter for this subset of data). I'm not sure if there's anything to this, though. It's not like these numbers are all from one sub-era (the yellow line wanders in and out of this area several times).

Looking at this plot leaves little doubt as to when Doug Flutie began making his impact. I suppose the 1991 data is also enhanced by the fact he played in six OT games that year which with the two 5min halves format at the time was like a whole extra game to rack up stats.

You can also see a significant change in 1999->2000 when we went to the shootout format. The change is enhanced by the fact that there were more OT games that season than in the three years just before it, combined. The OT was shortened from four to two "innings" maximum after one season and the curve reverted nearer to its 1999 position. FWIW, the length and slope of that 99-00 segment is comparable to what happened with 2013->2014, only in the opposite direction.

2007 and 2009 were more similar to one another than 2008 which is quite different from either of them, but I'm not sure off-hand of any good reason for this. 2009 was the year we had mandatory post-FG kickoffs, FWIW.

One final thought: is 40 years perhaps TOO long a time to study to draw out trends? I mentioned above the goal post placement and blocking issues that would hamper efforts to examine pre-1975 data in a present-day context. But are there other changes in the interim that might obsfuscate the overall trend, or perhaps even worse, create the illusion of a trend whereas things might have evolved differently under a steady, uniform set of rules/etc throughout time? The first five years of data (75-79) are all quite similar to one another, but unlike anything that has come after it. I don't think the 1981 fully interlocking schedule had everything (maybe SOMEthing) to with it as the change was already underway in 1980. Not sure what happened there.

[Rammer]

Good stats guys, nice breakdown.

Here is mine, the Lions O could not sustain drives, bottom third of the pack I suspect. The Lions took more yardage off of ST play due to penalties than most teams, once again bottom third of the league. Combine that with the poor play calling and Lions were lucky to get into the playoffs.

= boring football for a paying public.

[Ravi]

Do we have any data handy on what happened 2005->2006 as the Renegades folded? To the extent that adding a new team has been a drag on offensive productivity, there should have been a similar effect in reverse upon subtracting one.

Interesting hypothesis. This took some original number crunching, but it turns out in fact that CFL scoring declined from 48.0 points per game in 2005 to 46.6 points per game in 2006 after Ottawa folded and the league shrank to eight teams. The 2006 scoring was the lowest of the past decade until this year.

CFL SCORING (points per game)
2005: 48.0
2006: 46.6
2007: 49.1
2008: 51.8
2009: 51.3
2010: 52.9
2011: 50.3
2012: 51.8
2013: 52.4
2014: 45.5

The numbers show 2013 was the second highest-scoring year in the past decade so some drop-off would be expected, but to go from the second highest total in a decade to the lowest total is exceptional.

BTW, scoring per game in the CFL actually went up from '01 to '02 when the Renegades entered the league. In 2001, scoring was at 48.29 points per game (over 9 points less than the year before). In 2002, scoring actually went back up to 51.22 points per game so I am not sure that expansion is a valid reason for the drop in scoring this year. Here is a compilation of points per game for the last forty years that I did on argofans.com at the end of last season:

1974 36.93
1975 43.05
1976 41.94
1977 40.93
1978 42.75
1979 40.22
1980 46.47
1981 48.55
1982 51.88 (the elimination of the tight end is complete with everyone playing two slotbacks instead and the run-and-shoot emerges in Toronto)
1983 51.61
1984 50.67
1985 45.36 (a drop in the level of quarterbacking, especially in the east where Ken Hobart was MOP and Condredge Holloway was injured)
1986 47.90
1987 53.07 (the rise of some good, young QBs such as Renfroe, Austin, Burgess, Dunigan, and Allen. The # of teams in the league reduced from 9 to 8)
1988 50.28
1989 54.57
1990 61.86 (the peak years of offence are '91 and '92 with outstanding quarterbacking being the key. Doug Flutie is "lights out" during these two seasons)
1991 64.15
1992 57.39
1993 55.25 (the beginning of U.S. expansion)
1994 57.19
1995 53.61 (the final season with American-based teams)
1996 50.96
1997 53.01 (the final year for Doug Flutie and the number of teams is once again reduced from 9 to 8)
1998 49.76 (concern over the seeming dearth of promising, young QBs arises)
1999 50.10
2000 57.68 (Anthony Calvillo emerges as a bona fide star QB in Montreal and Henry Burris lights up the scoreboard in Regina)
2001 48.29 (Calvillo misses much of the second half of this season with an injury. Dave Dickenson and Burris have both left for the NFL)
2002 51.22 (rookie Ricky Ray emerges in Edmonton and the Ottawa Renegades make their debut)
2003 52.68
2004 53.06
2005 53.90 (the final year for the Ottawa Renegades)
2006 46.67 (blocking rules are tightened on kick returns for this season resulting in fewer good returns; concern over quality of QBs returns)
2007 49.11 (Kerry Joseph is MOP and only QB to start every regular season and post-season game for his team)
2008 56.24 (Calvillo and Burris lead the return of the offences. Stubler's defence peters out in Toronto finally)
2009 51.26
2010 52.92
2011 50.28
2012 51.81
2013 52.40 (despite so many injuries to quarterbacks, offensive coordinators do an excellent job preparing backups and general managers and scouts do an excellent job of finding quality quarterbacks)

[sj-roc]

Ravi, I think part of the reason for the spike in 2000 was the adoption of the OT shootout format. Unfortunately I don't have numbers handy for the amount of OT points scored in the OT games that year to support this view but I recall one OT game that went the distance and ended something like 52-52. There were more OT games in 2000 than in 1999, 1998 and 1997 combined which only served to more strongly expose the folly of the 4-"inning" shootout and they dialed it back to 2 the very next year.

Also perhaps worth noting in the spirit of your comments: in 2007, AC missed most of the 2nd half to tend to his ailing wife.