clock menu more-arrow no yes mobile

Filed under:

Stat Relevance Watch: Wrap-up and Analysis

New, 5 comments

There are, as they say, lies, damn lies, and statistics. The numbers mean something, yet often we're not sure what. Here SMQ will look at the final regular season statistics in more than a dozen major categories to suss out who succeeded in what and how that statistical success correlated to overall success in terms of final record. SMQ does not have the luxury of a high-powered supercomputer or degree-type qualification in mathematics or statistics, but his analysis will be driven as deep as his egghead, tinfoil cap curiosity and cell phone calculator will take it. That is to say, quasi-scientific at best.

Part One: Which stats most closely correlate with success?
Part Two: What do the best teams do best?
Part Three: ACC Game-by-Game Results
Part Four: Big East Game-by-Game Results
Part Five: Big Ten Game-by-Game Results
Part Six: Big XII Game-by-Game Results
Part Seven: Pac-Ten Game-by-Game Results
Part Eight: SEC Game-by-Game Results
Part Nine: Non-Conference Game-by-Game Results
Part Ten: Bowl Game-by-Game Results


Part eleven: Final Wrap-up and Analysis

The Method: SMQ used ESPN box scores to pull out specific numbers from every game in each of the above editions, and developed a winning percentage for each of eleven major statistical categories. That is, if the winning team outgained its opponent running the ball, that game was marked as a "victory" for the rush offense category; if the loser had a higher conversion rate on third down, the game was marked as a "defeat" for the third down efficiency category (the stats below are listed in offensive form, but the records are identical as from a defense-centric point of view, i.e. the team that gained more yards also allowed fewer). And so on for each of the categories in each game until the sample was dry. After which each category's "record" was added up to determine its correlation to victory among the group as a whole.

First, the range of each category's relevance by conference. The number in bold is the "rank" of the category within the conference, followed by the actual winning percentage of teams that outperformed opponents in said category:

Total Offense Passing Offense Yards Per Pass Rush Offense Yards Per Carry Penalty Yards Time of Possession First to Score Teams at Home Turnover Margin
ACC 7 (.546) 10 (.424) 3 (.656) 2 (.706) 3 (.656) 9 (.485) 8 (.519) 5 (.588) - 1 (.748)
Big East 2 (.826) 9 (.522) 5 (.696) 1 (.913) 2 (.826) 11 (.364) 10 (.471) 5 (.696) 7 (.652) 4 (.706)
Big Ten 2 (.860) 9 (.605) 1 (.907) 5 (.721) 6 (.682) 11 (.390) 3 (.763) 7 (.636) 10 (.477) 8 (.630)
Big XII 3 (.708) 10 (.458) 2 (.714) 6 (667) 8 (.625) 11 (.435) 7 (.646) 5 (.674) 9 (.575) 4 (.676)
PAC Ten 3 (.756) 8 (.667) 2 (.810) 1 (.822) 6 (.698) 11 (.436) 9 (.618) 5 (.733) 10 (.558) 7 (.686)
SEC 5 (.617) 9 (.532) 1 (.766) 3 (.674) 7 (.609) 11 (.311) 5 (.617) 2 (.681) 10 (.489) 8 (.605)
Non-Conference 1 (.584) 4 (.762) 2 (.850) 4 (.762) 6 (.718) 11 (.439) 9 (.675) 7 (.691) 10 (.643) 8 (.686)
Bowls 7 (.613) 7 (.613) 2 (.839) 6 (.625) 4 (.645) 10 (.385) 4 (.645) 3 (.688) - 1 (.870)
Average 'Rank' 3.75 8.25 2.25 3.5 5.25 10.63 6.88 4.88 7.0 5.13

Granted an outlier or two, most categories fall into fairly predictable patterns, the most notable being that every one of them is, ultimately, the mark of a winner: aside from `Fewest Penalty Yards,' always a telltale sign of defeat on the whole, the only occasions success in any category had a slightly higher correlation to defeat were in `Pass Offense' in the ACC and Big XII, in `Time of Possession' in the Big East and in the rough "control" element of `Home Team' in the Big XII and SEC (that category was not part of the first game-by-game study, of the ACC).

Otherwise, there is more some consistency in terms of relevance (as measured by the 'rank') than in terms of actual winning percentage, which tended to fluctuate within windows as large as 25 or 30 percent. Simply outgaining your opponent, for example, didn't mean a whole lot in the low-octane ACC (.546), but it did in the Big Ten (.860). Even the success of the home team fluctuated as much as 17 percent. The only exceptions were 'First to Score,' which usually resulted in a correlation somewhere in the high sixties, and 'Fewest Penalty Yards,' which never showed the slightest indication of being important to victory. The complete irrelevance of penalty yardage again comes with the caveat that penalties are situational killers (see Louisville jumping offsides on Rutgers' first ill-fated field goal attempt to win in November or any number of questionable flags - like LSU "holding" to negate a critical early touchdown at Florida, perhaps, or the same call against Missouri late in the Tigers' loss to Iowa State - for a similar example of defeat somehow via flag), but cumulatively, there's no reason whatsoever to sweat a lot of penalties. It's when flags are thrown, not how many. The debate over the explanation has already begun elsewhere, centering on the idea that penalties are most closely linked to time of possession (offenses draw more flags than defenses), which is correlated to victory. The numbers are iffy on that, though, and SMQ is skeptical because he considers time of possession itself largely an ancillary indicator.

It is interesting to note that higher totals in `Yards Per Pass' correlated more strongly to victory in every case than cumlative 'Passing Yards,' usually dramatically so, by these relatively small margins (success in the two was very often coupled, of course), but the exact opposite is the case with running the ball: amassing more total rushing yardage has a stronger correlation to winning across the board than does yards per carry, though it's a closer matter there than with the passing numbers. Both of these categories finished with slightly higher correlations to winning than success in `Total Offense,' though the combination of these three elements at the top of the following table represent a nice symmetry: "throw to score, run to win," as they say, or, for you Matisse fans, "an art of balance":

Rank Category Win %
1. Yards Per Pass .785 (241-66)
2. Total Offense .740 (231-81)
3. 3rd Down Efficiency .709 (212-87)
4. Turnover Margin .681 (156-73)
5. Rush Offense .678 (213-101)
6. First to Score .674 (213-103)
7. Yards Per Carry .673 (206-100)
8. Time of Possession .624 (179-108)
9. Pass Offense .556 (173-138)
10. Home Team .539 (126-108)
11. Fewest Penalty Yards .405 (119-175)

Ah, the straight winning percentages are not exactly the same as the averages. The OCD version:

Stat Category Win % Stat Category Win %
Total Offense .740 (231-81) Yards Per Carry .673 (206-100)
> 500 .796 (35-9) > 6.0 .708 (34-14)
450-499 .737 (28-10) 5.5 - 5.9 .526 (20-18)
400-449 .870 (80-12) 5.0 - 5.4 .650 (39-21)
350-399 .548 (74-61) 4.5 - 4.9 .676 (48-23)
300-349 .513 (60-57) 4.0 - 4.4 .569 (41-31)
250-299 .343 (34-65) 3.5 - 3.9 .522 (35-32)
< 250 .132 (14-92) 3.0 - 3.4 .485 (33-35)
Pass Offense .556 (173-138) < 3.0 .259 (50-143)
> 400 .583 (7-5) Yards Per Pass .785 (241-66)
350-399 .722 (13-5) > 12.0 .875 (14-2)
300-349 .566 (30-23) 10.0 - 11.9 .793 (46-12)
250-299 .548 (68-56) 8.0 - 9.9 .739 (99-35)
200-249 .529 (73-65) 6.0 - 7.9 .479 (104-113)
150-199 .486 (70-74) 4.0 - 5.9 .285 (47-118)
< 150 .392 (58-90) < 4.0 .167 (10-50)
Rush Offense .678 (213-101) Third Down Efficiency .709 (212-87)
> 300 .786 (11-3) > 70% .875 (7-1)
250-299 .708 (17-7) 60 - 69% .875 (28-4)
200-249 .723 (60-23) 50 - 59% .763 (74-23)
150-199 .624 (63-38) 40 - 49% .560 (70-55)
100-149 .544 (86-72) 30 - 39% .464 (77-89)
50-99 .341 (47-91) < 30% .306 (60-136)
< 50 .222 (20-70) Turnover Margin .681 (156-73)
> +3 .867 (52-8)
Fewest Penalty Yards .405 (119-175) + 2 .672 (43-21)
+ 1 .625 (70-42)
Time of Possession .624 (179-108) 0 .500 (75-75)
-1 .375 (42-70)
First to Score .674 (213-103) -2 .328 (21-43)
< -3 .133 (8-52)
Home Team .539 (126-108)

Discrepancies due mainly to ties and virtual ties in some categories, neutral site games and a weird absence of time of possession numbers for many ACC and Big East games.

"Big players make big plays in big games," right? So it should be no surprise that producing big plays in the passing game - or at least bigger plays, on average (the category also includes preventing big passes, remember) - was the single greatest indicator of success over the course of the entire season. Looking at the more specific subcategories, there is a clear stratification overall that was hinted at but not demonstrating as outright in the conference-by-conference looks: the better the per pass average, the better the winning percentage. Teams that averaged below eight yards per pass were losers; eight yards or more, big winners; starting at zero, the line is on a steady, uninterrupted climb.

Three of the top four categories, `Yards Per Pass,' `Third Down Efficiency' and of course `Turnover Margin' - along with `Rushing Offense,' too, really - all show this trend of winning percentages within the category steadily increasing with better numbers. `Total Offense,' `Pass Offense' and `Yards Per Carry' do not show this; in each, there is a point of diminishing return where, although still correlating to victory, the percentages are slightly lower above a certain threshold. Throwing for a lot of yards was mostly a crapshoot, unless you were one of the few teams that managed to top 350 yards, the only situation in which the winning percentage in `Pass Offense' topped .700 (or .600, for that matter). At 400 yards, though, the percentage dips again. In some cases, this was because the opposing team also threw for a chunk of yardage, but at any rate, passing for 400 virtually always means a tight game (teams wouldn't keep throwing otherwise, unless they are Texas Tech or Hawaii) that is really decided elsewhere. The same is true of `Total Offense,' but note that the winning trend begins low, at 300 yards, as it does with `Rush Offense,' at 100 yards. This could be another case for balance: teams that could pass for 200 and run for 100, if they took care of the ball, generally won. That's not a lot of offense (only 22 teams averaged shy of 300 total yards per game, only two of them - Middle Tennessee State and Ohio U. of Ohio - winners). If that production didn't result in a win,  it was a good bet barring turnovers the defense allowed significantly more.

From a defensive angle, notice this trend as well, and it may be the most important: the failure in the bottom subcategory of each larger measure is nearly always more pronounced than the success at the top, which would make the case for defense where the presentation has been from an offensive ppoint of view. Flipped to a defensive perspective, no teams were better than those that allowed fewer than 250 total yards (.868 over 90 more games than the 12-plus-per-pass teams that won .875), and their rate of success was even greater than the success of the teams at the top of any subcategory. On average, a team that was held under four yards per pass was twice as likely to lose than a team that finished below 150 yards total passing.

In general, SMQ thinks the numbers are an ode to balance and consistency: yards per pass is a sign most of all of a high completion percentage, regardless of the number of big plays, and is probably inseparable in a lot of instances from third down efficiency just below it. There may be a sort of addendum in there related to the effects of balance, which SMQ had considered before but scrapped because it was deemed to time-consuming. But now he's interested in what the margins between categories may be able to tell us. Any further thoughts? Remember, before establishing an ideology of some sort, every category save penalty yards correlated to vcitory, and the differences between them aren't chasms.