I'm in an NFL Pickem league. I don't care how I finish, and in fact I kinda believe that Pickem is all luck and most people forced to pick all games, end up converging on a 50%. But I digress. I am here to talk about something different.
I didn't feel like making my own picks this year, I made it my mission to create a neat little algorithm that would make my picks for me. That way I would have to invest as little time as possible into a league that I entered so that the other guys would stop hassling me to join. So here is what it does:
Take Dodds' game predicted scores. Takes the difference between the 2 teams for each matchup. Match this difference with the spread. This determines which team to pick.
The next step is to assign confidence intervals. The way it does this is, the farther that Dodds' absolute predicted difference is from the absolute spread value, the more weight it gives the pick as a confidence interval. I made the entire model in an Excel spreadsheet. now I simply copy/paste spreads, copy/paste Dodds' predicted scores, and magically all my picks and CIs are done for me.
Seemed like an interesting idea. So anyway, the moral of the story is that this method has produced disastrous results. I started using the model in Week 3. In week 5, I was out of town, so missed using it.
Results:
wk 3: 43/105
wk 4: 41/105
wk 6: 29/91
wk 7: 26/91
wk 8: 61/105 (added since this post was originally created)
Notation: (wins*confidence intervals / all games * confidence intervals)
In a nutshell, not only has the Game Predictor not had even a +50% week, its actually been getting progressively worse, to the point that it would be mathematically difficult to sustain a losing streak this long.
I realize from reading old forum posts, that Dodds does not do the game predictor to beat the spread. And frankly I don't care if I lose my league, it cost me 10 bucks. But after having so much fun making the spreadsheet model, I have been stunned by its performance, particularly because I figured at worst it would be 50%.
I figure that the model will rebound over the course of the season and probably converge back to 50%, but I thought I'd throw it out there and see if anyone has any thoughts.
EDIT: Edited to add Week 8 results.
I didn't feel like making my own picks this year, I made it my mission to create a neat little algorithm that would make my picks for me. That way I would have to invest as little time as possible into a league that I entered so that the other guys would stop hassling me to join. So here is what it does:
Take Dodds' game predicted scores. Takes the difference between the 2 teams for each matchup. Match this difference with the spread. This determines which team to pick.
The next step is to assign confidence intervals. The way it does this is, the farther that Dodds' absolute predicted difference is from the absolute spread value, the more weight it gives the pick as a confidence interval. I made the entire model in an Excel spreadsheet. now I simply copy/paste spreads, copy/paste Dodds' predicted scores, and magically all my picks and CIs are done for me.
Seemed like an interesting idea. So anyway, the moral of the story is that this method has produced disastrous results. I started using the model in Week 3. In week 5, I was out of town, so missed using it.
Results:
wk 3: 43/105
wk 4: 41/105
wk 6: 29/91
wk 7: 26/91
wk 8: 61/105 (added since this post was originally created)
Notation: (wins*confidence intervals / all games * confidence intervals)
In a nutshell, not only has the Game Predictor not had even a +50% week, its actually been getting progressively worse, to the point that it would be mathematically difficult to sustain a losing streak this long.
I realize from reading old forum posts, that Dodds does not do the game predictor to beat the spread. And frankly I don't care if I lose my league, it cost me 10 bucks. But after having so much fun making the spreadsheet model, I have been stunned by its performance, particularly because I figured at worst it would be 50%.
I figure that the model will rebound over the course of the season and probably converge back to 50%, but I thought I'd throw it out there and see if anyone has any thoughts.
EDIT: Edited to add Week 8 results.
Last edited by a moderator:
At least they took the boasting out of the intro this year I hadnt really looked at it this season.
