Arimaa Forum - Print Page


    
      
        Arimaa Forum
        (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
      

        Arimaa >> General Discussion >> Improvement from first game to second
        
(Message started by: Fritzlein on Jan 28^th, 2007, 9:04am)

Title: Improvement from first game to second
Post by Fritzlein on Jan 28^th, 2007, 9:04am

Omar and I were having a discussion about improving the rating system. We both agree that new players are much weaker than 1500 on average, even though we start out everyone with a rating of 1500. Omar concluded that we could make the ratings more accurate by measuring the true playing strength of players joining the server, and starting them at that rating instead of at 1500. I objected, wagering that players on average improve 100 points between their first game and their second. If we want to start folks at a lower rating to be more accurate, then we should also give them 100 free points just for playing a second game, also to be more accurate.

Omar doubted my claim of improvement between first and second game, so here are the stats across all the human players to have joined the server. There are only three bots against whom enough first and second games were played to be meaningful. (Note that due to rating inversions, newcomers are sometimes asked to play ArimaaScoreP2 before ArimaaScoreP1, so there are some stats with ArimaaScoreP2 as first game for a player.)

bot	First Game	Second Game	Improvement
ShallowBlue	26.9%	45.5%	142
ArimaaScoreP1	66.7%	71.7%	41
ArimaaScoreP2	42.9%	64.4%	153

This lends some weight to the notion that players improve 100 points between their first and second game, although I can't explain why the result for AirmaaScoreP1 is so much less than for the other two bots. Anyway, Omar, this might mean you want to take another look at Ryan Cable's suggestion in this thread:
http://arimaa.com/arimaa/forum/cgi/YaBB.cgi?board=talk;action=display;num=1140458845