Cool little tool you posted. It helped answer some of my theories around if early uncompetitive games (blowout games) are actually leading to a slow down in points in the second half or 4th quarter. Here are the Dec and Jan queries I ran to prove that out.
Jan -
www.killersports.com/.../query
Dec -
www.killersports.com/.../query
Also, I am not surprised to see there being no preponderance of Overs over a 3 year period when just looking at the dimension specifically, I would think similar results would come out on Favorites as well. Given that Vegas understands the casual bettor likes high scoring games and favorites.
My models are much more complex and just don't look at 1 dimension at a time.
For example purposes to try and explain my approach ... this is a simple fictional model ... Looking at 3 years of data, when a game consists of non-divisional opponents where both the home and away team are averaging >= 102 points per game in their last 10 games, each are shooting >= 47% from the field, home team is the favorite, home team just lost on the road in their last game and are coming off 2 days rest ... when all of those things are true, the game has gone Over 61% of time over the past 3 years. That scenario may of happened 120 times in the last 3 years, and 73 of those 120 games the game has gone Over.
The reason I believe in this approach is because my theory is Vegas also uses computer/algorithms to produce majority of their lines now a days. As a result, a computer is more likely to produce a pattern/trend. Humans setting every line every day would be more erratic and unpredictable. To shine more thoughts on my theory, I think 2014 is when Vegas introduced computers, reasoning is because there were major swings in all sports data starting around that time. When comparing data patterns between 2007 - 2013 to 2014 - Present, it looks like some approach to setting lines had changed.
All this being said, I could just be crazy! haha I haven't ruled that out yet either. BOL everyone