I've developed a bayesian probabilistic model that has done really well in back testing, paper trading and now live testing. So far under a really small live testing sample size it has a 60% win record. It has also done pretty well with picking home spread covers. With that said I wouldn't blindly trust it just yet but instead use this as some useful information to help your existing handicapping.
Here are some stats for this season regarding totals and spreads compared to pinnicale's lines:
Pinnicale spread r2 score 0.2268
Model spread r2 score 0.23
Model home points r2 score 0.222
Model away points r2 score 0.180
Model total points r2 score: 0.1872
Pinnicale total points r2 score: 0.1618
R2 score is a regression metric used to determine how much of the variation in the target variable the model explains. For example, so far this season Pinnicale's spreads explain about 22.68% of the variation in the actual point differential at the end of the game whereas my model is just slightly higher at 23%. Pinnicale's totals explain about 16% of the variation in the actual totals and my model is 18.72%.
When posting picks I will be using implied probabilities instead of american odds/decimal odds as I believe they are easier to understand potential edges. I will always use Pinnicale's lines/totals when posting predictions but will include the predicted home and away totals for completeness sake. I will also only post totals picks where the model says we have a 6% or greater edge over the book's implied probability. With spreads I will post when the model says we have any edge over the book's implied probability.
Additionally I have one spread prediction for today:
Toronto Raptors @ Oklahoma City Thunder, home cover -1.5, implied prob: 50.25%, predicted prob: 54%
Here are some stats for this season regarding totals and spreads compared to pinnicale's lines:
Pinnicale spread r2 score 0.2268
Model spread r2 score 0.23
Model home points r2 score 0.222
Model away points r2 score 0.180
Model total points r2 score: 0.1872
Pinnicale total points r2 score: 0.1618
R2 score is a regression metric used to determine how much of the variation in the target variable the model explains. For example, so far this season Pinnicale's spreads explain about 22.68% of the variation in the actual point differential at the end of the game whereas my model is just slightly higher at 23%. Pinnicale's totals explain about 16% of the variation in the actual totals and my model is 18.72%.
When posting picks I will be using implied probabilities instead of american odds/decimal odds as I believe they are easier to understand potential edges. I will always use Pinnicale's lines/totals when posting predictions but will include the predicted home and away totals for completeness sake. I will also only post totals picks where the model says we have a 6% or greater edge over the book's implied probability. With spreads I will post when the model says we have any edge over the book's implied probability.
| Home Team | Home points | Away Team | Away points | Total | OU | Implied Prob | Predicted Prob |
| Philadelphia 76ers | 112 | Brooklyn Nets | 106 | 218 | Under 223 | 50% | 58.29% |
| Miami Heat | 115 | San Antonio Spurs | 111 | 226 | Over 220.5 | 51% | 58.67% |
| Minnesota Timberwolves | 111 | Indiana Pacers | 113 | 224 | Over 218.5 | 51% | 58.98% |
| Houston Rockets | 120 | Portland Trail Blazers | 111 | 231 | Under 236 | 51% | 58.29% |
| Sacramento Kings | 105 | Dallas Mavericks | 112 | 217 | Under 228 | 51% | 69.3% |
| L.A. Lakers | 106 | Orlando Magic | 102 | 207 | Under 212.5 | 51% | 58.47% |
Additionally I have one spread prediction for today:
Toronto Raptors @ Oklahoma City Thunder, home cover -1.5, implied prob: 50.25%, predicted prob: 54%