An Experimental K-Stat

  1. You have chosen to ignore posts from Paul_K. Show Paul_K's posts

    An Experimental K-Stat

    The PaulK K-Stat is a statistical estimate of how many games (out of 16) a team would win, if it played at this level all year.  Example:  At the end of week 5, buffalo was playing at the level of a 6-10 team.  The K-stat works on the following principles.

    1.  It matters who you play and how badly you play against them.  Beating the Oakland Raiders by one point doesn’t count for a lot this year.

    2.  A team plays its best football for three quarters.  In the fourth quarter teams clutch their rosaries.  Sometimes they put their scrubs on the field.  Coaches wager all the chips on the table for their best chance at a win.  One team runs the ball three times and punts.  The other team throws an interception that gets run back for a touchdown on 4th and 10.  That's why the K-stat gives no credit whatsoever for hail marys.  They are fairly lousy predictors of the future.  In my system the score after 3 quarters is the better predictor of future wins.

    3.  A team that dominates time of possession statistically does better.  I throw in a time domination factor.

    4.  Good ratings start to go stale as teams change.  Almost everybody sees season-ending injuries, or sometimes newbie defenses/offenses start to get better.  My current stale rate is 10% per week.

    5.  The home team is expected to do 3 points better.

    Week 5 K-stats:

    buf       6.0 (out of 16 wins)  6-10

    mia       11.2       11-5

    ne         11.0       11-5

    nyj        11.3        11-5

    tennessee  5.5    5-11 or 6-10

    tampa bay 3.7    4-12 at best

    Indianapolis 14.5  14-2 or 15-1, tops in the NFL, I was surprised but those are the numbers.  Congrats to underdogg.

    Pittsburgh 12.5   12-4 or 13-3  also surprisingly high.

    San Diego 5.6     6-10 at best.  Sorry, JBolt, I do not mess with the numbers.

    Denver 7.7       8-8 at best!  Lower than many people would put them, but honestly, this team lollygags for three quarters and then lucks out at the end. 

    Just to show you how things change from week to week, here's

    Week 4

    buf       6.6

    mia       11.1

    ne         10.0

    nyj        11.1

    Week 3

    buf       9.5

    mia       7.4

    ne         9.4

    nyj        10.3

    Week 2

    buf       9.7

    mia       4.6

    ne         8.7

    nyj        9.8


    What stands out for New England is the way they've grown from a 9-7 potential to an 11-5 potential.  More than New England, Miami has exploded up the charts from a 5-11 to an 11-5.

  2. You have chosen to ignore posts from Paul_K. Show Paul_K's posts

    Re: An Experimental K-Stat

    Complete week 5 K-Stats

    buf 6.0
    mia 11.2
    ne 11.0
    nyj 11.3
    bal 8.2
    cin 7.5
    cle 4.7
    pit 12.5
    hou 5.7
    ind 14.5
    jac 5.4
    ten 5.5
    den 7.7
    kc 4.7
    oak 0.5
    sd 5.6
    dal 8.0
    nyg 11.8
    phi 11.1
    was 6.0
    chi 9.7
    det 6.9
    gb 5.7
    min 10.4
    atl 13.9
    car 4.7
    no 12.8
    tb 3.7
    az 9.4
    sf 8.2
    sea 9.9
    stl 1.8

  3. You have chosen to ignore posts from NickC1188. Show NickC1188's posts

    Re: An Experimental K-Stat

    Could you post the formula?  I'm intrigued, but confused...

    So you combined stats, then created a weighted average that emphasized recent weeks?