ESPN FPI

This topic has 11 replies, 11 voices, and was last updated 7 years, 1 month ago by Chidojuan.

Viewing 4 reply threads

Author

Posts
- 09/24/2017 at 10:36 PM #42433
  6
  
  Sweetness
  Participant
  
  I’m curious if any of you know much about, or have looked into ESPN FPI at all? I tend to be someone that will put more stock in stats and numbers than anecdotes and observations, but I can’t figure out why FPI is so terrible every year. Not only that it is so terrible, but that ESPN continues to use it as its main analytical predictor.
  
  Every year there are going to be outliers, but it feels like with FPI there are just so many obvious flaws that its credibility is completely undermined.
  
  A few examples I looked at today.
  
  Arizona is still ranked four spots ahead of us. This could make sense if we both had performed similarly leading up to the game and played similar schedules, but they are now 2-2, having played us, Houston, and a couple dogs.
  
  Stanford is ranked #9 in the country, despite being 2-2, getting beaten handily by SC, and San Diego State. Those are two good teams, but FPI has USC ranked at #15 and San Diego State ranked #46 (1 spot behind Arizona). So a two loss stanford team who was not even that competitive going up against the #46 and #15 teams in FPI’s own ranking system, comes in at #9?
  
  Ohio state, despite getting beat at home thoroughly by Oklahoma is still #2 in the country, Oklahoma being #3? I get that Oklahoma hasn’t played anyone besides Ohio state which could ding them, but neither has Ohio State!
  
  North Carolina at 1-3, lone win coming against old dominion comes in at #36, UCLA at #29, and Florida State at #12.
  
  If you’re ESPN, how can you continue to publish this garbage? It’s so bad. I’m not saying this to complain that it slights the Utes. We always seem to be low in it every year, but I just don’t get how ESPN, the largest producer of sports content in the world has continued to publish this model.
- 09/24/2017 at 10:54 PM #42434
  3
  
  rbmw263
  Participant
  
  All you need to know about fpi is that it is released before the season. There is so many variables and annual turnover in cfb that this is.
  
  Sagarin and s&p are far better analytics
- 09/24/2017 at 10:56 PM #42435
  
  Utah
  Participant
  
  I don’t get it. It’s supposed to use the last four years, recruiting rankings and QB play.
  
  I think that’s why it hates Utah. Our QB’s suck and our recruiting classes have been poor.
  
  But who knows.
  - 09/24/2017 at 11:58 PM #42442
    
    noneyadb
    Participant
    
    When you lose 16+ players to the NFL, your returning QB wasn’t exactly stellar statistically, and majority of returning starters are on the defensive side of the ball, that’s not a recipe for high expectations or preseason ranking.
  - 09/25/2017 at 6:57 AM #42444
    
    gUrthBrooks
    Participant
    
    Our recruiting classes have been poor? Past recruiting performance is better evaluated on draft day. Ute coaches know what they’re doing.
    - 09/25/2017 at 7:04 AM #42445
      
      Wilson’s Mustache
      Participant
      
      I think the 4 year avg. is somewherein the 50s or high 40. Not spectacular.
      - 09/25/2017 at 7:58 AM #42449
        5
        
        Tony (admin)
        Keymaster
        
        One could argue that we have great coaching then. We are overachieving based on that, right?
        
        09/25/2017 at 9:07 AM #42453
        
        leftyjace
        Participant
        
        That’s exactly how I see it.
        
        Long live the King!
        
        09/25/2017 at 9:41 AM #42457
        
        Wilson’s Mustache
        Participant
        
        I think that’s definitely a fair statement. I think Kyle has done an amazing job developing players, perhaps, with exception to a few postions (QB & WR). Utah over performs according to its recruiting rankings.
  - 09/25/2017 at 8:09 AM #42450
    
    PlainsUte
    Participant
    
    4-year average sounds good on paper to smooth out statistical noise, but especially in college football there is too much turnover from year-to-year of players and even coaches for it to make sense. Using recruit rankings is pretty suspect, also. A lot of quantitative prediction models will use the prior year’s stats in the first several weeks of the season but then drop them once there are enough games to make the stats robust. Even then there are flaws because the early season strength of schedule varies a lot.
    
    Bottom line: they are ESPN, so they can publish anything they want and people will believe it because of the “brand” and it will generate discussion and clicks to their site.
- 09/25/2017 at 9:38 AM #42456
  6
  
  shakeitsugaree
  Participant
  
  FPI formula is proprietary – which is all I need to know (cough bulls**t cough).
  
  I imagine it goes something like this:
  [(# of stars, recruits) * (# of wins, last year) * (1000 if SEC team) * (100 if head coach a high profile name) * (10 if historic football school)] / (1000 if from small television market) = dogs**t 💩
- 09/25/2017 at 10:56 AM #42467
  1
  
  Chidojuan
  Participant
  
  There are better responses than mine on this thread, but something that struck me early in the season was that many commentators and articles were mentioning touchdowns lost (or gained) with certain players, which seems like an attempt to bastardize baseball and football. I think that is at the heart of FPI, it’s a baseball-like formula applied to football. It’s already been wrong at least twice with Utah, not to mention NC State this past weekend. Question is, when does it attract enough scrutiny to be scrapped.
Author

Posts