RE: Week 2 NFL Daily Fantasy Football Analysis - Positions, Salaries, Projected Point Values, Rankings and More!
This is cool. I wish you joined our Fantasy League..
Anyways great to see this analysis. I'll definitely follow to see how well this algorithm works.
BTW long time ago I wanted to find a good way to rank offenses and defenses based on the strength of schedule or strength of opposing defenses and offenses respectively. It seemed to me though that as the season goes by you'll have more information about the strength of previous team matchups.. but it seems like if you updated the data based on previous matchups that would in turn depend on other matchups and you'd get some kind of infinite recursion. I was wondering how you would solve that and use the entire matchups database to rank offenses and defenses? Not sure I'm describing the question the right way, but hopefully you get the gist of what I'm asking. Thanks.
Thanks for the kind words! I saw the fantasy league starting but only declined because I am already in 10 yearly leagues this year. Needed some balance with that and everything I am doing here with DFS analysis.
I totally understand what you are asking with infinite recursion, that would be very tough to model. You wood have to be very careful with how you quantified that information so it came across in a meaningful way. I really like the FiveThirtyEight blog and how they use a chess model (ELO) to calculate strength of teams. I opted for differences in actual versus predicted. Both rely on numbers that are easier to pin down.
Have you ever seen the NFL parity wheel? Pretty cool graph that comes out every year showing exactly the same problem you described:
Wow 10 leagues... no problem. Maybe you can join us next year.
Yeah ELO is interesting for sure. I like that it's simple and just uses points & location.
I've never seen the parity wheel, but yeah it does describe the problem.
Hey I just had an idea. To keep it simple to start just use Avg Points Per Game for offense and Avg Points Per Game allowed for defense. To prevent the infinite recursion we can identify the median 8 teams (25% of the league) on both Offense & Defense from 2015 and give them an average strength rating to start in 2016. Once we fix certain team ratings as average we probably won't have a recursion problem. You can also adjust the initial ratings based on major changes (coaching changes/player acquisition) so that the 8 median teams may not be purely based on 2015. What do you think?