M57 wrote:

Has anyone played with this yet?  I've spent and hour or two with it, and my initial observation is that results vary significantly based on chronology of events.  This could very well be because of small sample sizes, but it's quite easy for a player to win 5 of 10 games (with 4 or 5 players) - sit out 10 or so, and come back and lose just 1 or 2 games to see their ranking tank way below the mean - even though they have won 40% of these games overall.   At least with small sample size - rankings seem to fluctuate pretty wildly, with relatively few losses taking big chunks out of the gains of highly ranked players.

Also, and I'm not sure if this is a glitch, but I'm pretty sure I've been able to create end results where the mean score is well below 1000, with for instance one player having barely over 1000 and the rest of the field well into the mid 900s - this with a number of the players in the field having wins or even multiple wins.  I have also achieve the opposite effect with most all players 'in the black.'  It seems to be all about order.

Summary: It's apparent to me that this needs testing with a much larger data base.

Maybe it's from the small sample size, but it does appear to be very volatile and heavily favor the most recent result. Didn't play around with the numbers up at the top though.

I have not had time to play with it much.  I think working the #s at the top will be necessary.  Here's the description again:

• mu – the initial mean of ratings.
• sigma – the initial standard deviation of ratings. The recommended value is a third of mu.
• beta – the distance which guarantees about 75.6% chance of winning. The recommended value is a half of sigma.
• tau – the dynamic factor which restrains a fixation of rating. The recommended value is sigma per cent.
• draw_probability – the draw probability between two teams. It can be a float or function which returns a float by the given two rating (team performance) arguments and the beta value. If it is a float, the game has fixed draw probability. Otherwise, the draw probability will be decided dynamically per each match.

Looks like I didn't let you set a draw possibility & the default is 0.1.  I can add that ability to set draw possibility sometime soon.

I was also thinking of adding a calculation of the GR based on the current 20 * ratio method.  Would be interesting to contrast the two.  Not sure how long that will take me to get to it.  maybe a week.