Stats Relevance Watch, Part Three: Into the Trees In the ACC

Wherein SMQ examines 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. I do not have the luxury of a high-powered supercomputer or degree-type qualification in mathematics or statistics, but analysis here will be driven as deep as my egghead, tinfoil cap curiosity and cell phone calculator will take it. That is to say, quasi-scientific at best. If you've ever said "the only number that matters is the one on the scoreboard" or anything to such effect, click here and don't be such a philistine.
- - -

Part One: Which stats correlate most closely with winning?
Part Two: What do the best teams do the best?

Part Three: ACC Game-by-Game Results

Cumulative statistics and statistical rankings, as assessed in parts One and Two, are nice, neat and telling in their own right, but the overall averages also have the power to mislead: what's true in the macro is not necessarily true in the micro. It's fine to establish, for example, that the top 20 rushing teams in the country win, say, 59 percent of the time over the course of the season, as I reported in Part One. But that big picture doesn't tell us anything about how rushing stats correlate to victory on a game-by-game basis - because there's so much variability within the season-long average, it's possible the team with better rushing numbers in any given game won 80 or 90 percent of the time. Think of it this way: if a team finishes 20 percent better than its average in a given category every time it wins and 20 percent worse than its average every time it loses, that category has a perfect correlation with winning and losing. But that may not necessarily show up in the final average.

So in addition to establishing the impact of numbers from the "forest" perspective, it's probably more useful to get down into the trees.

To do that, I'm looking hard at game-by-game data (using very useful box scores from ESPN) in each BCS conference to put into a catch-all chart covering hundreds of games. Here are the results from the 2006 season, with each conference's results linked therein; they're also linked on the left sidebar.

For each of the 49 games played among ACC teams last year, I developed a winning percentage for each of eleven major statistical categories (the stats below are listed in offensive form, but merely flip the records for a defense-centric point of view) as well as a pair of "control" categories, "home team" and "first to score." 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. And so on for each of the categories in each game until the supply of examples was dry. At that point, each category's "record" was added up to determine its correlation to victory among the group as a whole.

The ACC results may be the least representative of the national results because the conference is easily the most defensive-oriented and conservative in the country. If not for Matt Ryan and a solidly balanced attack at Clemson, the entire league would have almost no major offensive weapons to speak of. Keep that in mind when scanning the quick and dirty, portable results:

Category	Win %	Record
Rush Yards	.833	40-8
Yards/Rush	.830	39-8
Yards/Pass	.771	37-11
Yards/Play	.702	33-14
First Downs	.698	30-13
Total Yards	.688	33-15
First to Score	.688	33-16
Turnover Margin	.676	25-12
Home Team	.583	28-20
3rd Down %	.542	26-22
Time of Possession	.531	26-23
Passing Yards	.521	25-23
Fewest Penalty Yards	.323	10-21

If it's less relevant then "Home Team," it's almost random. So where the ACC is concerned, at least, I would regard everything below "Turnover Margin," which correlated to winning a little more than two-thirds of the time, with extreme suspicion. Every yardage-based category except passing offense had a strong correlation to victory, predictably, and even more predictably, teams with better total rushing numbers (remember: a result of the defense stopping the run as much as the offense running well) were dominant, winning at better than 80 percent clip. Given the quarterback situation here past Ryan, it's no surprise that efficiency rather than sheer yardage apparently played a much greater role in winning and losing.

The OCD version follows the jump...

This year, I categorized the specific numbers in each section according to margins rather than raw production. Where you see + 200-249 under rushing yards, for example, the number is the record of teams that outrushed opponents by somewhere between 200 and 249 yards.

Total Yards	.688 (33-15)	Yards/Play	.702 (33-14)
> + 300	3-0	> + 3.0	4-0
+ 250-299	2-0	+ 2.5-2.9	2-0
+ 200-249	4-0	+ 2.0-2.4	4-1
+ 150-199	4-1	+1.5-1.9	4-0
+ 100-149	7-2	+ 1.0-1.4	8-1
+ 50-99	8-4	+ .5-.9	6-2
< + 50	5-8	< + 0.5	5-11
Pass Yards	.521 (25-23)	Yards/Pass	.771 (37-11)
> + 250	0-1	> + 3.0	12-3
+ 200-249	2-1	+ 2.5-2.9	2-0
+ 150-199	3-1	+ 2.0-2.4	3-0
+ 100-149	7-5	+ 1.5-1.9	6-1
+ 50-99	8-8	+ 1.0-1.4	2-2
< + 50	5-7	> + 1.0	12-5
Rush Yards	.833 (40-8)	Yards/Rush	.830 (39-8)
> + 250	1-0	> + 3.0	11-0
+ 200-249	2-1	+ 2.5-2.9	-
+ 150-199	5-0	+ 2.0-2.4	1-1
+ 100-149	8-0	+ 1.5-1.9	9-1
+ 50-99	11-3	+ 1.0-1.4	4-1
< + 50	13-4	< + 1.0	14-5
3rd Down %	.542 (26-22)	Turnover Margin	.676 (25-12)
> + 25 %	7-0	> +3	12-1
+ 20-24 %	3-1	+ 2	4-1
+ 15-19 %	5-3	+ 1	9-10
+ 10-14 %	2-4	0	12-12
+ 5-9 %	5-5	-1	10-9
> + 5 %	3-9	-2	1-4
		< -3	1-12
Possession	.531 (26-23)	Home Team	.583 (28-20)
Penalty Yards	.323 (10-21)	First to Score	.688 (33-16)
First Downs	.698 (30-13)

In some sections, we see that a basic margin doesn’t matter at all. In fact, being slightly better in total yards, pass yards, yards per play, third down percentage and even turnover margin had a significantly stronger correlation to defeat than victory. In every single category, though, we see the trend reverse as the margins get wider: each one of them (again, with the exception of total passing yards) is a tremendous indicator of wins and losses as the gap widens. If you graphed these results, the line in every category would be on an uninterrupted incline.

Take-away. Nobody in the ACC except Boston College could pass, so running determined everything. Teams that successfully established the run, even just a little, and shut down opposing backs were essentially impossible to beat; even slight margins on the ground, differences that didn’t show any relevance in other categories, correlated overwhelmingly to victory in the cases of rushing yards and yards per carry. Outrushing an opponent by just 50 yards (27-4) and/or a yard per carry (25-3) was an almost foolproof route to a win.

Next: The Big East.