Let me repost the financial rankings side by side with the statistical rankings to better clarify how to use and read them both. I've highlighted the NO v. HOU game, a game in which the Box chooses NO and suggests a double bet.
First thing to consider is HOW the Box chooses NO. Highlighted in red are the best ranking in each category between NO and HOU. NO, as you can see, has a statistical advantage, even though it's slight. If the spread were in NO favor by anything but the slimmest margin, there is an excellent chance the system would pick HOU to cover because there is such a small statistical advantage. Right now, because the spread is in HOU favor, the Box will pick NO all day long on a statistical level. The initial picks are based entirely on statistics - ability to score (alpha and Sharpe) and volatility (beta) have little or nothing to do with a team's advantage over another team when going against the spread.
So that's how we picked NO in the first place. Sharpe, alpha, and beta are non factors in the initial pick. I've run some models that add each financial metric in various weightings, only to find that historically it performed worse in the last two seasons that my original statistics only approach. So, for the time being, NO is the pick, and the Box loves the underdog in general.
The "Double the Bet" (I'll call it DB for simplicity) suggestion that follows doesn't employ either Sharpe or alpha, though perhaps in the future it will. For now, it uses only beta (volatility). More than that, it uses the GAME beta, averaging the beta of the two teams. I do this because beta of a single entity in a matchup wouldn't make sense. If HOU played CHI, the 1st place beta versus the 32nd place beta, even though HOU is 1st and will likely play with the same consistency it's played all season, CHI could score 40 points or 5 points and are such a wildcard, the game could go either way. When I figure the beta of the game, it makes it seem like much less of a sure bet.
The DB takes a given parameter (number of bets I want to place, in this case), and tells me which, if any, of the top picks, are low enough in game beta to have outcomes that are more predictable. There are (and will be) weeks that DB suggests doubling none of my bets, since they are all too volatile. However, you can't argue with the results - since 2006, the DB record in suggested doubles is 37 - 18, a staggering 67% win rate. An argument could be made to ONLY bet the DB games, but what's the fun in that?
Alpha and Sharpe have been more or less discarded from spread bets, as I have yet to find any historical win rate that even matches up to the purely statistical system I have in place. What they are useful in determining, however, are moneyline winners. I've been working on the perfect combination of the two, but right now, I have complete a DB model for moneyline bets that would have won 5 times as much money since Week 3 on my moneyline choices. Outside of moneyline, alpha and Sharpe act as useful guides in determining the value of a team. I'm hoping to show that alpha bets are better for parlays and over/unders, while Sharpe are useful in moneyline parlays. Making picks has been the easy part for the system, the hard part has been prioritizing the picks such that I have the most winners at the top - this is where the financials can be the most utilized.
So when you're considering your moneyline parlays, look at the financials as a guide to help weed out some of the less favorable options.