Good news everyone, I spent all day Sunday writing the code to get the custom stats I need. It took many hours, but now I have the ability to run the algorithms using any range of stats, beginning on any day. I just started backtesting the algorithms using a 30-day moving range of stats, starting January 2, when things fell off the cliff. I chose 30 days because over the approximately 60 days from November to December, when accuracy was at 67%, the average range was 30 days. If there are any stats guys that have a better suggestion, speak up.
| NO -1.5 at ATL |
|
| Winner: | New Orleans |
| Underdog: | N/A |
|
|
| CHI at MIA -5 |
|
| Winner: | Miami |
| Underdog: | N/A |
|
|
| ORL -2 at DET |
|
| Winner: | Orlando |
| Spread: | ORL -2 ***** |
| Underdog: | N/A |
|
|
| WAS at MIN -4 |
|
| Winner: | Minnesota |
| Underdog: | N/A |
|
|
| HOU at DEN -4.5 |
|
| Winner: | Denver |
| Underdog: | N/A |
|
|
| LAL -3 at POR |
|
| Winner: | LA Lakers |
| Underdog: | N/A |
| Correct | Games | Win% |
| STRAIGHT-UP | 1 | 6 | 16.7% |
| SU UNDERDOG WIN | 0 | 0 | ---
|
| ***** | 0 | 1 | 0.0% |
FYI, using previous 30 days, the resuts were:
| Correct | Games | Win% |
| STRAIGHT-UP | 1 | 4 | 25.0% |
| SU UNDERDOG WIN | 2 | 3 | 66.7% |
| ***** | 1 | 1 | 100.0% |
3 comments:
Thanks for doing the work. So will the picks from now on be based on stats from last 30 days?
new Orleans is a play tonight.
They'll be based on full season stats until I finish backtesting. I still don't know if the shortened range will offer any improvement.
I also automated data acquisition, so things should go quite speedily now.
Post a Comment