Thursday, November 15, 2007

Week 11 - Saintly Saints, Flaccid Texans

That's right, I said flaccid.

So, you ask a good question, but I of course have an amazing answer.

The way the double down is figured is Beta only - so I throw Sharpe and Alpha out the window, since they are useful primarily in the moneyline bets. The system picks NO based on the statistics - but the double down is chosen by picking the best possible GAME beta (HOU beta + NO beta). So thanks to HOU being a solidly consistent team, it brings down the overall beta rank. If the game beta ranks in the top 50% of games, PLUS the pick is one of the top 5, the system suggest a bet doubling.

My fear in this game is the return of Schaub and Johnson at the same time - it inevitably will alter the beta if they perform really well, since there will be some volatility. This makes this double down risky, but a necessary evil - hard to argue with results (17x more money on double down bets in the last two years). Fortunately, I have doubles on two other top 5 games, so hopefully something hits.


The Mad Capper said...

Something still seems amiss...

The sharpe and alpha ratios are supposed to take into account both scoring and volatility, and the results of these ratios say pick the texans. meanwhile...

The box takes into account scoring, and you cross check it with the beta for volatility and pick a double bet on NO.

Same two criteria - two different results?

Matty and the Box said...

The box actually doesn't take scoring into account at all - it's based purely on statistical averages over the season. Basically, the box says statistically, when correcting for league average, NO is a superior team (to the tune of almost 1.5 pts, says the box). the double down is separate entirely, and is beta for the game. So the box says: pick NO, they are statistically better. And beta says, NO vs. HOU as a game has lower volatility than half the games this week, so double the bet.

The criteria are actually different for measuring each. The alpha scores are pulled in HOU favor ONLY because of their beta (beta is a component in figuring alpha - greeks stick together i guess), not their statistical advantage or scoring advantage. If you look at just the sharpe, which doesn't use beta but uses a different measure of volatility, NO actually has the edge.

Make sense?

The Mad Capper said...

did you ever try my idea and blend the box with the ratios? Meaning, define "performance" in the sharpe/alpha ratios as the box's statistical analysis?

Matty and the Box said...

I tried redefining the components of them with statistics, and had some decent results. right now, alpha and beta are figured using a ratio of points score to statistical advantage as it's "performance". sharpe is purely points scored, and actually has benefited from it for moneylines.

i ran into some problems (ie, NE was in the bottom third of alpha) when i did pure statistics and i haven't figured out how to resolve it. the problem may exist because of the variables used in determining alpha and beta - you need "market" data and "risk free" returns, both of which don't make sense on a statistical level. there's no "returns" when comparing the teams unless i can use the score.

it's a little complex, but i like the compromise i've struck so far. when i see you in december, i can show you the stuff i can't explain.