View Full Version : CC ladder poll (rating system implementation)

25-07-2008, 09:32 PM
While appreciating that Glenno puts the work he does into maintaining the ladder, since he switched over to measuring it by number of wins rather than a Glicko-style system I have not found it meaningful. It is now a hybrid of activity and performance rather than an indicator of performance alone.

I concur , we must adopt a performance based rating system as soon as practicably possible.

In response, a poll has been created. Several rating systems have been found on the Internet (below are links to such systems). Which shall we implement?

American Postal Chess Tournaments (http://www.correspondencechess.com/apct/ratesys.htm)
Association of British Scrabble Players (http://www.absp.org.uk/ratings/ratings_detail.html)
Chess Federation of Canada (http://www.chess.ca/ratings_system.htm)
Free Internet Correspondence Games Server (http://www.ficgs.com/membership.html#rating_chess)
GameKnot (http://gameknot.com/help_ratings.htm)
Glicko-1 (http://math.bu.edu/people/mg/glicko/glicko.doc/glicko.html)
Glicko-2 (http://math.bu.edu/people/mg/glicko/glicko2.doc/example.html)
International Email Chess Group (http://www.iecg.org/rules7.htm#_Toc126766785)
Scrabble Australia (http://www.scrabble.org.au/ratings/system.htm)
United States Chess Federation (http://math.bu.edu/people/mg/ratings/rating.system.pdf)

Kevin Bonham
25-07-2008, 09:54 PM
A lot of the simpler systems there seem quite satisfactory to me; Glicko-2 is very hard work for the maintainer and with relatively limited data to go on the accuracy payoff might not be worth it.

Interesting to see that Scrabble has assessed an ELO-like modelling curve against actual game data and found a straight line does the trick better.

10-08-2008, 10:38 PM
We've got two weeks to go before this poll ends.

As of 10/8/2008, the number of votes were as follows:

Glicko-1: 2 votes.
Glicko-2: 1 vote.
Others: 0 votes.

17-08-2008, 04:37 PM
One week to go before this poll ends.

As of 17/8/2008, the number of votes were as follows:

Glicko-1: 3 votes.
Glicko-2: 2 votes.
IECG: 1 vote.
USCF: 2 votes.
Others: 0 votes.

23-08-2008, 08:43 PM
24 hours to go now.

24-08-2008, 12:57 PM
I voted for the USCF Rating System, mainly for the following reasons:

a) It is the system used on the Internet Chess Club (ICC), and it seems to work very well there.

b) Very high-rated players are not penalised for playing low-rated players (e.g in the first round of a Swiss tournament)

c) It is significantly easier to calculate how many rating points a player has gained or lost in a tournament.

d) Unlike the Glicko system, the USCF rating system does not penalise you for playing actively. Under the Glicko system, a player can not play for a couple of years, then score very highly at a tournament and gain a couple hundred points. Under the USCF system, this would not happen.

The only disadvantage of this system is that it may not work as effectively with a smaller number of rated players (ICC and USCF have hundreds of thousands of members, whereas the number of ACF rated players is about 5000). Still, I think the above four advantages outweigh this possibility.

25-08-2008, 03:20 PM
Poll now closed.

Final tally of votes:

Glicko-1: 5 votes.
Glicko-2: 2 votes.
IECG: 1 vote.
USCF: 3 votes.