Quantitative Analysis: Fantasy Football

As most of you know, I’m a huge fantasy football junkie.  I spend a disproportionate amount of my time on Sunday mornings getting my lineups ready based on research I’ve done during the week thru various forms:  XM Radio Fantasy Football Channel, FFLibrarian, Rotowire (Subscription), The Huddle (Subscription), and even simply talking to players.

I stumbled across Numberfire (I hate the name) from a friend of mine and did some research and saw that they were recently written up on TechCrunch.  Numberfire applies data mining and statistical analysis to determine which fantasy football players to sit and/or start each week… they accomplish this by looking at many different data points that affect a players outcome such as the player itself, his team, his competition, and other factors.

They just released their Week 1 analysis of performance and it turns out:

The two key metrics are that I beat both ESPN (59%) and Yahoo! (77%). There’s not much analysis that needs to be derived from that statistic. All of made projections; I was right more often and in the case of Yahoo!, almost embarrassingly so.

The smaller, but equally interesting metrics are the deltas, or the difference between a projection and the actual result. My deltas were lower than both (ie: were more accurate), and the difference between the projections themselves would have caused a 10.8 point difference over Yahoo! and a staggering 16.2 point difference over ESPN.

Not bad.  I’m thinking of converting one of my teams (out of 3) to be based on the numberfire analysis system and seeing how I perform this season.

What I don’t like about numberfire (maybe because I haven’t found it) is that the algorithm they use to score the players is a black box.  I hate black boxes.  Maybe this will change over time.

Needless to say, quantitative analysis can help make decisions, especially when there are lots of data points.  We obviously used this thesis when creating Varick Media Management and we’re seeing similar awesome results.

Enhanced by Zemanta
  • http://www.dlewis.net/nik Dan Lewis

    Have you looked at http://footballoutsiders.com? They’re the big trailblazers here (disclosure: I’ve written for them and contributed to their annual).

  • http://twitter.com/nikbonaddio Nik Bonaddio

    Darren,

    Thank you for the write-up. I’m actually not wild about the name either, but the domain was available for the right price (free) and I didn’t feel like writing some squatter a $50k check for predictable.com. It is what it is though :)

    I pride a lot of things I’m doing on nF in being transparent, which is why I go through the trouble of exposing the data points of players, matchups, and scenarios. That said, I can certainly look into exposing the methodologies behind where those data points come from. For purposes of not wanting to get into data overload, I erred on the side of only showing the most pertinent information (the predictors that most strongly correlate to our performance projection) and not the relatively extraneous ones (similar players, matchups). I am more than happy, however, to be equally transparent about all of my algorithms if critical mass dictates that level of transparency.

    Any other comments or questions, please do feel free to reach out to me personally. nF is a one-man band, so I can assure you that it won’t be lost in the ether. :)

    Nik

    • http://www.darrenherman.com dherman76

      Nik, thanks for stopping bye! As you can see, I’m a huge fan. I’ll be in touch.

  • http://www.dlewis.net/nik Dan Lewis

    Have you looked at http://footballoutsiders.com? They're the big trailblazers here (disclosure: I've written for them and contributed to their annual).

  • http://twitter.com/nikbonaddio Nik Bonaddio

    Darren,

    Thank you for the write-up. I'm actually not wild about the name either, but the domain was available for the right price (free) and I didn't feel like writing some squatter a $50k check for predictable.com. It is what it is though :)

    I pride a lot of things I'm doing on nF in being transparent, which is why I go through the trouble of exposing the data points of players, matchups, and scenarios. That said, I can certainly look into exposing the methodologies behind where those data points come from. For purposes of not wanting to get into data overload, I erred on the side of only showing the most pertinent information (the predictors that most strongly correlate to our performance projection) and not the relatively extraneous ones (similar players, matchups). I am more than happy, however, to be equally transparent about all of my algorithms if critical mass dictates that level of transparency.

    Any other comments or questions, please do feel free to reach out to me personally. nF is a one-man band, so I can assure you that it won't be lost in the ether. :)

    Nik

  • http://www.darrenherman.com dherman76

    Nik, thanks for stopping bye! As you can see, I'm a huge fan. I'll be in touch.