Wednesday, February 28, 2007

Statistics for late season prospects

After each match I calculate the Geordies' prospects for promotion or relegation. I make the following assumptions:

  • All teams are equally likely to win or lose a match, i.e. the result could just as accurately be predicted by the toss of a coin.
  • There are 16 teams in the division, top 3 promoted & bottom 3 relegated at the end of the season.
  • Promotion is pretty certain if a team scores more than 36 points.
  • Relegation is pretty certain if a team scores less than 26 points.

Early in the season this set of assumptions may be satisfactory. However, immediate past Captain Tony Hall has pointed out that there is additional information from league positions that reflects teams' abilities. Where there are significant variations in teams' abilities, the coin tossing model is likely to give increasingly misleading predictions as the season progresses. So how do I incorporate all the information from the latest league table into my prediction of prospects?

I propose to start with the latest league table, update the points for each team using the results of the games that we have played, and evaluate the points for all teams in the remaining games by assuming that the expectation for each team in each game is 1 point. Visit the Geordies' website for the resulting 'end of season' table based on the league table published on 23 January 2007 and the known results for the Geordies since that time.