Very good read you wrote here LT

bigboydan · 11-13-2007, 06:27 PM

I see you noticed the low scoring non-division games numbers as well.

Quote:

Oakland Raiders, AFC West flying Under the radar

By: LT Profits

NFL division games are generally lower scoring then non-divisional games, so the Under is usually a good play in these contests. We have broken out the divisional total records by conference and division over the last three years plus the first half of this one, and have found some lucrative situations.

Now it has been a long-held belief that NFL divisional games are lower scoring than other contests, thus making the Under a good play in these battles.
Oakland Raiders, AFC West flying Under the radar

Two main reasons for this are that the offenses generally play closer to the vest, since division games are the most important games on the schedule, and that the teams face each other so often that the defenses usually know what to expect. Looking at the numbers the past three seasons plus the first half of 2007, this theory has generally held up, as the Under is 171-157, 52.1 percent (omitting pushes) in division wars since the start of the 2004 season. While this figure is not enough by itself to turn a profit, it can be turned profitable with some filters, and it is still better than the overall Under record of 458-440, 51.0 percent during this time.

When we break out the Under results between the American Football Conference and the National Football Conference, it is clear that the NFC is the more Under-friendly conference, as it was a slightly profitable 88-77, 53.3 percent for the low side. However, the single division with the highest Under percentage comes out of the AFC, and it is surprisingly the AFC West at 63.4 percent! Because of this phenomenal percentage, the AFC Under comes in at 83-80, 50.9 percent overall, despite the other three divisions in the AFC all favoring the Over.

Here is the breakout by division, with the last number on the right being the average combined total score in game played inside that division:

Code:

Under Records, Division Games Only – 2004 to 2007 (through games of 10/29/07)
AFC East 20-22, 47.6%, 39.4
AFC North 17-22, 43.6%, 43.2
AFC South 20-21, 48.8%, 43.6
AFC West 26-15, 63.4%, 41.8
ALL AFC 83-80, 50.9%, 42.0

NFC East 22-17, 56.4%, 41.0
NFC North 21-21, 50.0%, 40.2
NFC South 25-16, 61.0%, 39.8
NFC West 20-23, 46.5%, 44.4
ALL NFC 88-77, 53.3%, 41.4

ALL Division Games 171-157, 52.1%, 41.6

In actuality, while division games have been slightly lower scoring than non-division games since 2004, the difference is marginal, as the 41.6-point average for division games is just 0.4 points less than the overall league average of 42.0 points over this time span.

It is interesting that the AFC West Under has been so strong despite the fact that these division games are averaging a combined 41.8 points, suggesting that the success of the Under has more to do with inflated posted totals than with conservative play. That is fine with us, as long as the linemakers keep putting up totals of 42 points or more. The Oakland Raiders are the main reason for the West’s Under success, as they have gone below the total in 15 out of 21 division games in this span, with an average combined score in their division contests of 39.2 points.

Our advice would be to concentrate on division games involving the AFC West, NFC East and NFC South, and pay special attention to games involving these divisions that have a posted total of at least one point higher than the average total scores above, as these games would be prime Under candidates.

Ganchrow · 11-13-2007, 08:14 PM

Certainly another interesting article by LT.

I will point out, however, that the 52.1% win rate of the 171-157 record has a standard error of 2.77%, meaning that it's only about 0.77 standard deviations from a coinflip.

If the true divisional over/under probability were actually 50%, then over the course of 328 games we'd expect to see a win rate of 52.13% (on either of the over or under) or higher about 47.29% of the time (44.05% if approximating using the normal distribution).

In other words results of this magnitude would happen by chance alone almost half the time.