ANALeyesing historic football scores against DK

Using expected value to determine if there are positive values associated with previous total scores in games and DK expected total score in upcoming games.

We are a few weeks into this, ironing out the process since NFL week 1. Based on the information here, we're up 4.5% on college and 16% on NFL, 7.3% in all.

Perhaps this continues to have positive results, but also perhaps not!

There's a few ways to select data, to determine which previous games should be included. Do you use all of last years games? Only this years games? Some sort of blend? Recently, we're looking more at how each of those data sets would perform as total correct vs total bets. The chart above indicates what my activity has been. Performance could be better once the data used to evaluate expected returns is more well vetted.

Past year - games in the past 365 days

Full Last - all of last seasons games and all of this seasons games

This year - only this years games

2wk scale - drop 2 weeks of last season for every week that can be included for this season

3wk scale - drop 3 weeks of last season for every week that can be included for this season

"Push" indicates that neither the over or under has positive expected value. My interpretation is that if you were to bet on that game 100 times, choosing either the over or under every time, you would end up losing money because of the Vig or whatever Dirt calls it.

     Week 4 5 6 7   8

Full Last 61.8% 53.7% 61.7% 58.1%   53.2%

Past Year 60.5% 53.7% 63.6% 51.2%   50.0%

2wk scale  52.4% 61.0% 55.3%   55.3%

3wk scale  40.5% 35.0%  - becomes "This year" becuase all of last year games dropped   

This Year  65.5% 37.5% 42.9%    51.1%

Florida International vs Kennesaw   State O/U-49.5, O-odds--110, U-odds--110, Past Year - Push, Full Last-U, This   Year-U, 2wk scale-O

Louisiana Tech vs Western Kentucky O/U-49.5,   O-odds--112, U-odds--108, Past Year - U, Full Last-U, This Year-U, 2wk   scale-U

Georgia State vs South Alabama   O/U-55.5, O-odds--112, U-odds--108, Past Year - O, Full Last-O, This Year-O,   2wk scale-O

Nevada vs Boise State O/U-50.5,   O-odds--115, U-odds--105, Past Year - U, Full Last-O, This Year-Push, 2wk   scale-U

Virginia Tech vs California   O/U-49.5, O-odds--115, U-odds--105, Past Year - U, Full Last-U, This Year-U,   2wk scale-U

Charlotte vs North Texas O/U-61.5,   O-odds--105, U-odds--115, Past Year - O, Full Last-O, This Year-O, 2wk   scale-O

      4 5   

Past Year 63.6% 91.7%   

Full Last 42.9% 91.7%   

This Year 50.0% 83.3%   

2wk scale 46.7% 91.7%   

3wk scale 50.0% 70.0%    

     Los Angeles Chargers   vs Minnesota Vikings O/U-44.5, O-odds--112, U-odds--108, Past Year - Push,   Full Last-U, This Year-O, 2wk scale-Push    

This one is pretty wild. This will probably never happen again, but for week 5, based on the data we had, we went 17 for 18 on the O/U evaluation. 

Total Games is how many games the QB played that are being used to evaluate their chances of throwing for more or less passing yards than the O/U.

   tbd

I am not Josh Allen. I'm just an idiot that looks like him and you absolutely should ignore anything I do and say.