I made a post
in the January chatter thread about how I was dipping into NCAAB seriously for the first time. Basically, I've used widely available stat projections to try and hammer lines off by 4 or more points.
I've done this from 1/18 through today every day.
Here's the results so far
192 picks ATS: 100-86-6 for 53.76% winners
71 line moves in favor of the picks, 77 against
247 picks O/U: 126-119-2 for 51.4% winners
80 moves in favor, 52 against
439 picks overall: 226-205-8 for 52.4% winners
151 line moves in favor, 129 against
so....pretty meh alltogether. However, I have been clearing some decent bonus, so it has been profitable.
I am just blankly reporting the system. A reasonable bettor could do better. Here are a few factors:
1) the Duquesne effect. This team has really changed throughout the season. The stat projectors just don't account for their tempo/scoring surge in the 2nd half of the year. The projections have been solidly under time and time again during the last 6 or 7 games. the totals have moved 6.5, 11.5, 8, 3.5, 4.5 points higher for duquesne over the last 5 openers IIRC!!! IN that time, the system has been 1-4. A rational better would just avoid Duquesne under recommendations like the plague right now.
2) Home unranked fav against ranked road opponent.
The system has been monstrous of late, while the stat projections keep going with the ranked road doggie. I believe this actually highlights an interesting subset to avoid. By and large, the stat crunchers use a set homecourt adjustment or multiple. It's my intuition that this is incorrect. Home court is not a linear effect on the game. It's my theory that home court plays a larger role in projected close games with big crowds. During the offseason, I'll dive in and see what adjustments I can make.
In essence, though, just avoiding the Duquesne under stuff and the home unranked fave verse road ranked doggie would push the system to near 54%....which is nice for the amount of work it takes.
Of course, with it being this close, the sample size is not enough to say anything definitive from a stat basis.