As you may have read, I have recently used regression for estimating totals in the NFL and realized that betting the under is consistently the better choice in certain games.

I have used regression from 2000-2011 to create a regression formula and used it to estimate years 2012-13 weeks 7 and on. This has produced roughly a 62% win rate over 100 games in that period.

Looking further into these two systems I found the following:
System 1: is 23W-8L in weeks 15,16,17
System 2: is 30W-8L in weeks 15,16,17

Since I am only playing unders, I figured maybe the league was just hitting unders in these weeks, but I've found that in these three weeks in 2012 and 2013, the games have gone under 48 times, and over 44 times.

Definitely encouraging.

Anyone else had any similar revelations with later weeks in the season producing better results?