Predictability of 2nd Half Totals From 1st Half Results

Being the geek that I am I took a look at the first and second half totals of each NFL game to see if there was some predictability for second half totals based on the first. Without sausage grinding p-values and sds of such a small sample size, I find there are some strong trends, and not exactly the trends I would have expected.

I broke each half into three groups, the under group with half totals of 20 or less, middle group of game totals 21-24, and an over group with half totals over 24. This is to reflect most game totals falling in the 42 - 48 range.

Not surprisingly, the least predictive group was the middle group. It also had the smallest sample by far. There were a total of 49 first halves in the NFL with a first half total of 21-24. Of those 49 games, 5 also had a second half total of 21-24. The game total shaded to the under in this group with 25 second half totals falling under 21. As anyone with a calculator can now determine, 61% of second half totals in this group fell under 25 and 49% more than 20 points.

In the first half over group there were 80 game scores over 24. In those 80 games 54 had second half totals under 25 with a whopping 40 of them under 21. I plan to look at these more closely but my eyeball indicated that heavy unders were most likely in first half blowouts.

The under group had the largest sample with 121 NFL games under 21 total in the first half. This group also had the strongest predictive strength for second half totals. Of those 121 games 63 had second half totals hit the u21 (52%) and another 24 were under 25 (20%). This is a group that I will invest some time dissecting as well, and will probably break it down completely in a later entry.

Looking at the big picture there were a total of a little over 11,000 points scored in the NFL in 2008, an average of 43.5 per game. I probably should re-sort the totals over a slightly lower range, since my current spreadsheet uses a mean game total of 45. A range of 20-23 would reduce the skew to the under and certainly alter the results, as 20 is possibly the most common halftime total. Even so, with a mean expectation of ~ 21.5 points per half there would still be a strong trend toward first half unders.

Up next, I add in a couple more seasons and break down totals at both the margins and the extremes.