Wednesday, November 14, 2007

Week 11 - SPAM Rankings - Slight Change to Beta Algorithm

So here in Boxland, there are slight tweaks that happen on a constant basis. As of today, there is an ever so slight tweak to the Beta algorithm that had a minor impact on double down bets (went from winning 15X more in the last 2 year span to 17X more on the games on which bets are doubled), and had virtually no other effect. I will repost the game rankings with updated Vegas spreads later today, but for now we'll take a look at the Financial Rankings.

Some notes on how these are figured and what they mean, in case you're new to this blog or these metrics:

1.) Sharpe Ratio - a measure of risk adjusted performance in finance, I define performance here as the moving average of points scored per game. This is the closest to "pundit" rankings out of the financials, as it is a purely point driven derivative.

2.) Beta - a measure of volatility in finance. I am using a bastardized version of DVOA, the genius work of the team at Football Outsiders (a MUST read) in figuring this, and it has, with some success, measured how consistent teams have performed. It does NOT measure how good the team is - a team can be consistently BAD and therefore have "low volatility".

3.) Alpha - another measure of risk adjusted performance, but uses Beta in its algorithm, and therefore uses DVOA rather than a pure point performance. In a way, this measures "bang for your buck" in terms of a team's ability to "outperform" it's statistics or expected performance. Usually has the most interesting results.

So with that said, here are the rankings (organized by team, not by rank):

Sharpe Alpha Beta
Team Rank Rank Rank
ARI 18 14 3
ATL 30 32 30
BAL 26 19 10
BUF 27 30 6
CAR 24 11 15
CHI 23 27 32
CIN 9 5 13
CLE 4 7 28
DAL 2 1 31
DEN 29 25 2
DET 7 6 23
GB 8 12 12
HOU 11 10 1
IND 3 3 25
JAC 19 28 22
KC 28 31 29
MIA 15 15 24
MIN 14 16 11
NE 1 2 5
NO 12 24 21
NYG 6 4 26
NYJ 25 13 16
OAK 20 9 4
PHI 17 22 20
PIT 5 8 18
SD 10 18 27
SEA 13 17 8
SF 31 23 14
STL 32 29 7
TB 22 26 19
TEN 16 20 17
WAS 21 21 9

What pops out at me is the Alpha score - according to Alpha, CIN is the number five overall team. It's hard for me to believe that CIN would provide so much bang for your buck, but if nothing else, they have consistently scored this season. And more surprising to me is that NE is actually SECOND in SOMETHING this year. I think it's actually just a function of competition - DAL has none in the NFC, and NE has... well, none, but at least more than DAL.

The beta rankings are close to expected, though I'm a bit surprised by HOU. Despite QB issues and Andre Johnson being out, they've been very consistent this year. I expect them to drop with the return of Schaub and Johnson this week, as it may throw off the average. On the bottom half, I'm mildly shocked at the number of top tier teams - CLE, DAL, DET, IND (after the last two weeks), JAC, NO, and NYG - all playoff contenders or locks, all playing with some inconsistencies. This can be attributed in part to injuries (IND, JAC), and in part to inability to put together back to back weeks (NO, NYG, DET). DAL and CLE are mysteries to me, since DAL seems to be pretty steady and CLE has been playing at such a high level. My best guess is the dropoff in points on DAL part, which has been steady over the season, and a terrible Week 1 performance from CLE keeping them down. I may have to add a parameter that weights current weeks greater than early weeks, but the sample size in a season is already so small, I'm hesitant to do so.

A note if you're betting this week, the Alpha and Sharpe are nice tools for moneyline bets. The Beta is a nice overall tool if you're worried about your team's consistency (or as Mad Capper puts it, "multiple personality syndrome"). As noted, some top tier teams all rank in the bottom half of Beta, indicating a bet on any of them could lead to trouble depending on which version of the team shows up. Look out tomorrow for the updated game rankings and best bets.

No comments: