Pirateboy's NCAAF ML Dogs Experiment - W9

[knicknut]

I don't know if you've addressed this, but you're wrongfully assuming that average point spread and winning probability are directly related. While they're certainly highly correlated, they're not perfectly linked. You don't think Pinnacle is messing up every time it offers two teams at the same spread but with different moneylines (or vice versa), do you?

That's because the distribution of scores is not uniform in each game. If we assume every game will have an approximately normal distribution centered on the point spread, the standard deviation of this distribution will have a large effect on the percentage of the time each team wins.

For example, if you have an underdog that is very erratic (high SD), they will be more likely to win than a team that has the same point spread but is much less erratic (low SD). As a result, the more erratic team will have lower + odds ML than the consistent team (in an efficient market). And in you\r system, you're betting the low SD games, thinking all games are equal.

You may show short term success, but not all X point underdog games are created equal. I can't see this being a profitable strategy longterm.