The SEC in 2006 – Statistical Comparison

Behind the win-loss records for the SEC conference games last year is a wealth of statistical information on how those records came to be.

And, if you look closely enough, what we might expect this year.

Some time ago I had taken the “Points For” and “Points Against” records of SEC teams for SEC play only. Not surprisingly, the statistics were much stingier for SEC only play verses non-conference play, particularly points scored on offense (reflecting tough SEC defenses). I believe SEC only statistics are much more accurate in analyzing predicting league play, for obvious reasons.

For each team, I calculated average points for and points against. In the case of Florida we found the following –

Average Points For - 24
Average Points Against – 17.1

Then, I calculated the standard deviation for both above statistics to get a “normal distribution” for each. By doing so, we can determine what the most likely set of outcomes, within one standard deviation, for a given team in the past season. For Florida, we found the following ranges within one standard deviation –

Average Points For – 17.5 to 30.5
Average Points Against – 9.96 to 24.3

What this tells us is that in the 2006 season, Florida was 68.2% likely (1 standard deviation either way) to score between 18 and 31 points in SEC competition, and to give up between 10 and 24 points.

Rather than just giving this information in numbers, I find it useful to provide a chart that shows how the SEC teams stacked up in 2006 in these terms. First the offense (click on charts for bigger version)–



A few notes on what you see here, using Alabama as an example. By our statistical analysis, Alabama scored an average of 17 points per game last season (the “A”). One standard deviation higher ("H") shows ‘Bama scoring as much as 22, and one standard deviation lower ("L") shows ‘Bama scoring as low as 12. 68.2% of the time we can expect the Tide to score between 12 and 22 based on 2006 stats. Point ranges are the numbers across the top.

In the simplest form, the chart shows how likely a team was to beat another team based on last years stats. Take Alabama compared to Arkansas for instance. Alabama’s high mark of 22 points is 1 point below Arkansas’ low end of 23. Thus, within the 1 standard deviation range, it is highly unlikely Alabama could have outscored Arkansas.

Also a note on tight and wide dispersions. While we are dealing with only a small sample size (8 games for everyone but Arkansas and Florida who had 9), a tighter range is good if it is higher up the point scale, and bad if lower. While tight is consistent, where it is located shows whether it is good consistent or bad consistent. A wider range merely shows an inconsistency of results. LSU’s wide dispersion reflects the scores of both the Auburn game (LSU lost 7-3) and the Kentucky game (LSU won 49-0).

Lastly, looking at the chart from left to right, we can see the SEC teams that were hopeless scorers (Alabama, Ole Miss and Vandy) and those that could put up huge numbers (LSU and Tennessee).

In the case of Florida, the Gators never scored a great deal of points on any SEC team. However, they rarely scored only a few either.

The following is the defensive chart using the same analysis –

Obviously here, an average and range to the left is better. Here also you can see Florida’s success, with a top end of the range lower than any other SEC team (though LSU is close). And while an Auburn could potentially hold opponents to only 5 points, their inconsistency allowed them – within the same probability – to give up as many as 28. A tighter range, and to the left, is better here.

This also shows where Kentucky got into trouble. While their offense is among the league best (top chart), their range in defense is 2nd worst to only Mississippi State.

Lastly, this chart shows the comparison of both offense and defense for the same teams, and across the league –

League records are also included.

Using Alabama as an example again, if the best you are likely to score is equal to the average you are likely to give up, you are going to end up about 2-6 in league play.

Florida’s 2 bars show good consistency by their narrow range. Florida didn’t score a lot, but they didn’t fail to score either, and they didn’t give up a lot of points. This range seems to reflect the ball control/field position philosophy of Coach Meyer.

These charts also beg another question – how did Florida beat LSU?

Well, in large part Florida got lucky. LSU was held to the low end of their normalized one standard deviation range (10 points), while giving up nearly the top of their range on defense (23 points). Accounting for whatever point advantage home field gives, lets just say Florida didn’t want to play LSU again last year.

Finally, how does this help with the most important question – the upcoming season. Well, by looking at the above, and making a judgment call on whether a team has improved or not, we can see what to expect this year.

Using Florida as an example, we know the Gators return 6 starters on offense, and only 2 on defense. I’m going to say our offense, with our experience line and very talented receivers, will be a little better. I am also going to assume or defense will be worse. I also assume the defense will be less consistent. Based on those assumptions, we might have something like this –

Looking at this, 7-1 for the Gators could be a challenge this year.

Also what does this show for Alabama? Hopes are high in Tuscaloosa, but marked improvement on defense - and especially on offense - will be necessary for the Tide to be remotely competitive.

On a final note, how scary might LSU just be this year? The Tigers return 6 starters on offense, and 8 on defense including all-everything Glenn Dorsey. If you imagine the offense staying the same as the charts, and the defense improving (and it should), LSU is going to be very tough this year indeed.