I call him GAMBLOR! (Part 3)

Written by Football, Sports

On the cusp of Week 10 in the 2009 NFL season, Zack Dennis pauses to reflect on the performance of his game-calling supercomputer, Gamblor.

I promised a review of the year so far, and I’m happy to see that this last week’s performance has restored Gamblor to profitability for the season. And even though the model is slightly underperforming its target so far this year (winning at a rate of 53.6% compared to a goal of 55.0%), it has been behaving exactly the way it was designed to.

I summarized the two major models at the beginning of the season (in parts one and two), but to refresh your memory on how Gamblor works, when it feels the spread has been set accurately, it defaults to the underdog – these picks are referred to as “hedge” bets. Last week’s surge by the underdogs took them over the top for the season (to 64-63 for a percentage of 50.4%) but they’re still just a shade below the 51.1% that can be expected over the long haul. And so far this season the hedge bets have been a weak spot for Gamblor, as they have delivered a winning percentage of just 44.4%. Fortunately, this has been more than compensated for by the “flip” bets, where Gamblor thinks the spread is significantly skewed from where the statistics say it ought to be and thus presumes that Vegas is setting a trap – leading to a bet in the opposite direction. The flip bets haven’t been very active this season, but when they have, they’ve been delivering at an unbelievable (and certainly unsustainable) rate. Flip bets where the confidence level is above 85% have been correct 78.6% of the time. And where the confidence level is above 90%, the flip bets have performed at been a truly ridiculous rate of 95.8%.

It’s important to remember that confidence level does not equate with expected win percentage – the 85% confidence level corresponds to an expected win percentage of 54.9% (based on historical data) and the 90% confidence level corresponds to an expected win percentage of 58.0%. It’s hard to say whether its overachievement this year this is just a statistical fluctuation, or whether the sportsbooks have become a bit more conservative in attempting to set traps – waiting until they feel like they’ve got a sure thing before pulling the trigger. There hasn’t been a single model that’s delivered a confidence level of greater than 95% this season (which anticipates a win percentage of 70.1%), and it seems like the 95% bets have become more and more rare over the last several years (only ten games in the last three and a half years compared to 55 in the eight seasons prior). It could be that 90% is the new 95%, or it could have no significance whatsoever. It’s too soon to tell.

The following graph shows Gamblor’s performance over the entire time it’s been active.  I didn’t finalize the model and start betting until about halfway through the season in 2008, which is actually quite fortunate because it would have suffered a huge setback in Week 3 if I had.  There was a cripping loss on a 95% flip bet that would have seriously shaken my faith in the program – but fortunately I picked a very good time to get started.   In the graph below, the blue line represents the hedge bets, and the red line represents the flip bets, each based on a starting bankroll of $500.  The columns represent the overall performance once you combine the two.  It doesn’t seem like all that much has happened so far in 2009 – the overall growth of my bankroll has been just 10% since the start of this season – but that all has the potential to change this week…

flashfiller

Week

Zack

Gamblor

Weighted Wins

Weighted Picks

Weighted Win %

Profit

Win

Lose

Push

Win

Lose

Push

1

7

9

9

7

1339

1727

77.5%

$155.71

2

9

7

11

5

970

1292

75.1%

$98.97

3

9

7

6

10

607

1552

39.1%

-$61.75

4

8

6

5

9

288

963

29.9%

-$70.33

5

6

8

6

8

655

1217

53.8%

$15.46

6

7

7

8

6

625

1231

50.8%

-$10.53

7

8

4

1

3

9

1

246

1161

21.2%

-$127.43

8

3

10

9

7

709

1548

45.8%

-$34.43

9

7

6

10

3

1152

1609

71.6%

$119.67

Total

64

64

1

64

64

1

6591

12300

53.6%

$85.34

For legality purposes, this website does not promote or advocate gambling. This is solely for entertainment purposes only.