Non-Controversial Fun with Non-Controversial Stats

With Georgia appearing to do the impossible this past weekend by managing to lose a game while also making most fans feel glimmers of hope (this is not a moral victory this is not a moral victory this is not a moral victory), I started thinking about games in which teams seem to play well but still lose. In that spirit, I got to thinking about general trends that show up in all games, rather than Georgia specifically. This week, I am determined not to touch Georgia specifics with a ten-foot pole.

(For the details of our "we-own-this-field-except-for-the-endzone" game against the Cocks Saturday night, I expect we will all be gnashing our teeth....erm, I mean, we will all be whistling optimistically at what might have been, when we read Bill C's in-depth analysis of this game later this week. I'm just guessing here, but I imagine that UGA is going to look dominant in terms of EqPts, F&P+, and Field Position. I wouldn't be surprised if the EqPts is a 20-point spread or more, in fact. Our downfall, as we know, was turnovers, and it really is that simple. A few redzone stutters in the first half didn't help matters either, but they wouldn't have mattered if a single turnover had not taken place.) 

In any case, as I thought about this for myself, I decided I'd throw together something a little more simplistic that uses the kinds of statistics we're all familiar with:  total yards and points. How often do teams do well in either category, and still lose? And, in the spirit of Erk Russell, is it better to be dominant on one side of the ball or the other? To the tale of the tape, after the jump.

This google doc summarizes every single game played between FBS teams from 2000-2010. For each year, it shows the average yards and points per game for that season, as well as the standard deviation for that season. 

It then goes on to display the winning percentage of teams that turn in a dominant offensive or defensive performance, respectively, both in terms of yards gained/given up and in terms of points scored/yielded.

To understand the chart, I defined a "dominant" peformance on offense as any game in which a team gained at least one standard deviation more than that season's average yards per game, or points per game (as the case may be). Inversely, a dominant defensive performance is any game in which a team yielded at least a standard deviation fewer than the season's average yards per game, or points per game (as the case may be). 

In other words, looking at 2005, the average yards per game was 377.2, and the standard deviation was 115.4. Thus, a dominant offensive performance in 2005 was any team that gained at least 493 yards of offense (377.2+115.4=492.6). Similarly, a dominant defense in 2005 was any team that allowed 261 yards or fewer in a game. And we use the chart to determine a dominant performance in terms of points scored or points given up as well, and for any season from 2000-2010.

The reason this is a nice way to organize the data is that it is similar to standardizing test scores to produce a percentile ranking. For any nice large sample of data, such as all the games played in an entire college football season, you know that around 1/6 of all performances are going to be dominant defensive performances, and 1/6 are going to be dominant offensive performances. 

Also, with close to 2500 performances over the past 11 seasons that fit in these dominant categories, our sample size is plenty large enough to have confidence in the results. In other words, it is highly likely that the "real" winning percentage of dominant offenses really is very close to the .803 in the chart.

So, summing up the data, we see that teams that "dominate" on offense in terms of yards gained can expect to win 80.3 percent of the time, and teams that dominate in terms of points scored can expect to win 93.2% of the time. (Such a high success rate for points is to be expected, of course, because scoring more points than the other guy is the very definition of winning, and obviously any team that is a standard deviation above average in points for a given game is very likely scoring more points than their opponent in the same game. However, there is some "wiggle" room here, as we see when we compare offensive dominance to defensive dominance. But pure blind luck would expect to see a winning percentage of about 92-93% for all dominant teams on either side of the ball.)

When we look at the defensive side of the chart, we see a favorite cliche confirmed, though perhaps not by as large of a margin as we might have expected. When a team dominates on defense in terms of yards, it can expect to win (all else being equal) 84.1% of the time, and when it dominates in terms of points, it wins 96.8% of the time.

So, if you have your choice as a fan or a coach of putting a team on the field that dominates on offense or that dominates on defense, your chances of winning will be great in either case. But your odds are even best of all if you dominate on defense rather than on offense. Dominant defenses in terms of total yards win 3.8% more of the time than dominant offenses. Similarly, dominant defenses in terms of scoring win 3.6% more of the time than do dominant offenses.

Another way to think of the data is to look at the very last row, which shows the number of times per season we can expect to see someone lose after turning in each of the four types of dominant performances. When it comes to points scored/allowed, we can expect to see about 16 teams dominate on offense and still lose. (From week 1 this season, hello TCU! From week 2, hello, Georgia!) But we only expect to see this happen about 7 times for a team that dominates on defense. (It hasn't happened yet this season).

What is most frustrating/numbing about Georgia's performance on Saturday is that, unlike TCU from week one in their 50-48 loss to Baylor, the Dawg's defense did not get shellacked. If we include the 68 yards on the fake punt as total offense, which is how the statistics are officially kept, then the Dawgs did an average job on defense. Considering that SC is expected to be a better-than-average offense, however, this is still a good performance. If we just give SC credit for a first down on the fake punt, but assume that it was something of a fluke for it to go 68 yards like that, then Georgia turned in a very good defensive performance (in terms of yards allowed), though not quite a "dominant" one.

So, an above-average-to-good defensive performance, in terms of yards allowed, coupled with a dominant offensive performance in terms of points scored (and we earned our points the "hard" way, and left some on the field as well), added up to a 3 point loss. That is, indeed, frustrating.

The score understandably reminds of the 2008 Georgia Tech debacle, as does the fact that the Dawgs played very well on offense but couldn't quite finish in the end. However, the defensive story really is different, in that we didn't arm-tackle and shoulder bump our way to a fundamentals clusterfark, we didn't get beaten over and over again by the same basic group of plays, and we didn't look out of position for half of the afternoon.

There is legitimate disagreement about whether a team can lose but still be the "better" team. I know many around here are fond of saying that if you lose than you weren't better "on that day." I understand that sentiment, but I also see the argument that sometimes one team is objectively better but just doesn't happen to get the win. There is no doubt that we didn't lose like a bad team Saturday night. We lost like a good team (potentially very good) that made a few too many heartbreaking mistakes.

I make no comments whatsoever about how we are supposed to feel about that.

It's great to be a Georgia Bulldog, and go Dawgs!