Week 7, Wagers Accepted
The models were built and trained on data going back to 2012 and the most recent 8 games for each prediction. Since the most recent 8 games could go back to last season, an interesting trend came out:
During testing on the 2018 season, the model did better as the year went on.
With this in mind, and before putting actual money on an unproven model, I held off making actual wagers on these games. Even with last week being the first losing week of the model, I am now confident enough to start putting money down, starting with $20 per bet and increasing later in the year if $20 become lower than 2% of bankroll (currently using a more aggressive 3%).
Last week was our first losing week at 3-4-1, bringing down our 70% to 68% on filtered games (ones I deem as betable).
1. Mia +17 (Mia @ Buf -17) - Top ATS Pick, Filtered
2. Ari +3 (Ari @ NYG -3) - Filtered
3. LAR -3 (LAR @ Atl +3) - Filtered
4. Det +1 (Min @ Det +1) - Filtered
5. NYJ +9.5 (NE @ NYJ +9.5)
If you think some of these look painful because they are terrible teams (Mia, Ari, and NYJ...) then you are right.
It is very painful to start putting money down and see the model pick gong-show type teams, however the whole pain of using a data-driven model is let it help us avoid our own personal bias. It also helps that Fitzpatrick is starting for Miami, but not much...
1. LAR @ Atl - Under 55 - Top Totals Pick
2. Jac @ Cin - Under 44
3. Ari @ NYG - Under 50
4. LAC @ Ten - Over 39.5
5. KC @ Den - Over 48 - Thursday night
6. Mia @ Buf - Over 40.5
7. Oak @ GB - Over 47