Redskins or Deadskins?
Five weeks into the season and every week has been a winner. Last week was the lowest though, going 5-4 on filtered picks with Totals carrying the load (Week 5: ATS 0-3, Totals 5-1).
Total for the year, using filtered picks (the ones worthy of a bet based on model accuracy) we are 32-14-1 (69.6%) and up 15.1 units. Meaning if we put $20 on each bet we'd be +$302.
Time to Ignore
For the third straight week, the same game tops both the ATS and Total models. LAC and Cle, both with Unders, were for week 4 and 5, respectively. This week is a hard one to swallow, as the game I want no part of sits atop the list: Was @ Mia +3.5, 41.5.
Looking at how the model works gives us insight into when to trust the model and when to avoid the pick. We cannot blindly trust the model knowing if the underlying input data is not indicative of what will go into the next game.
The model uses data going to 2012, most of this is used for training the model. For each game, the most recent 8 games are used as input for the next game. This means that the most recent 8 games are used for predictions.
BUT WAIT WAIT WAIT, we are only at week 6?
Correct, meaning some games from last year are still being used for predictions. Additionally, if there is a change in coach or players, the model will miss it because it assumes the "team" is operating at the given team stat level. Early in the season is the biggest risk, where coaches and players have a greater chance of changing, however a major change midseason can also have an impact.
For example, a starting QB gets injured or a coach is fired.
Now back to the Redskins... who just fired their coach and have had QB injuries. After poor outings from Dwayne Haskins and Colt McCoy (albeit he got his start in the NE game), Case Keenum is coming back after missing nearly 3 games and coach Jay Gruden is out, replaced by OL coach Bill Callahan.
On the other side of the ball, Miami is off a bye and apparently is in full tank mode (trying to lose games). Will their players and coaches show up and fight for one of the only wins that may get this year OR will they continue to lose by double digits...
Another game with a QB returning is Sam Darnold and the Jets. We have over 3 games of extremely poor backup play while he was out, yet he is set to return this week (Dal @ NYJ +7).
Coach and QB replacements are not what the model is built to factor in. Therefore, Was @ Mia +3.5 and Dal @ NYJ +7 are being removed from consideration.
1. Sea -1 (Sea @ Cle +1) - Top ATS Pick, Filtered
2. Atl -2.5 (Atl @ Ari +2.5) - Filtered
3. Min -3 (Phi @ Min -3) - Filtered
4. SF +3 (SF @ LAR -3)
5. GB -4 (Det @ GB -4)
1. Atl @ Ari - Under 51.5 - Top Total
2. NO @ Jac - Under 44
3. Sea @ Cle - Under 46.5
4. Ten @ Den - Over 39
5. Pit @ LAC - Over 41
(All posted totals are "filtered", i.e. worthy of a bet)
Earlier we ignored Was and NYJ games due to QB injuries, yet Pit we are accepting (and did earlier in the year after Roesthlisburger went down). We trusted Rudolph was going to give good replacement value, and Big Ben had poor games before going down, which both combined to accept. I am using the same premise here for the unknown Delvin Hodges. He appeared to be competent in replacement, telling me the drop-off is not significant enough to ignore in this case.
Surprising on the OVER list is Ten @ Den. With no major team changes, we will trust the model and the billions of math operations it does to bring us here.