Slashdot Mirror

← Back to Stories (view on slashdot.org)

A Look At Competitive Ranking Systems

Posted by Zonk on from the i-win-no-i-win dept.

Christopher Allen writes "Competitive ranking is used in sports, chess, and other games, and by online services, such as the new TrueSkill system used on Microsoft's Xbox 360 Live service. Rankings establish who is best, create fair competitive matches between players, and handicap players with differing skill levels. An article at 'Life with Alacrity' discusses a number of approaches to competitive ranking on the internet, each with different issues and advantages."

3 of 8 comments (clear)

  1. Interesting stuff going on at beatpaths.com by Swinging+Man · · Score: 3, Interesting

    There's an interesting ranking system emerging over at beatpaths.com. It started as a way of ranking NFL teams based on who has beaten whom, perhaps fueled by a Broncos fan frustrated with his team being ranked too low by other systems. There are plans to analyze the NBA and MLB, but it seems generally applicable to most competitions.

    Another cool thing: dig the graphs straight out of graphviz, a nice open source tool for buiding graphs from textual specifications.

  2. Re:Chess is my favorite by ChristopherA · · Score: 2, Interesting
    An interesting comment was posted by F. Randall Farmer about ever escalating ELO (chess-style) rankings:
    A data point on ELO cheating for you: Yahoo! Games uses ELO rankings for several their two-player games. Before recent abuse mitigation changes, some people used robot to accumulate scores in excess of 6,000,000 points. The abuse-the-ranking game had become a totally seperate competition.

    For now, Yahoo! has capped the ELO scores at 3,000 (I think.) This removed most of the cheating incentive.

  3. eGenesis Ranking System by Teppy · · Score: 3, Interesting

    The article talks about the ranking system that I invented and implemented in A Tale in the Desert for use in the Discipline of Conflict. Great to see the coverage, but unfortunately the algorithm didn't work well in practice, and we've since abandoned it.

    The problem was that it took too long to converge. Of course all the parameters can be adjusted for faster convergence, but then it became too easy to metagame! I concluded that any continuous system that collapses the result to a small amount of data (like a rank (ELO), or a rank+confidence (TrueSkill) or a bitvector (eGenesis)) after a match would suffer from this problem.

    "A Tale in the Desert II" replaced the eGenesis Ranking System with an asynchronous king-of-the-hill method. You start at rank 1, and must play someone at rank 1. It's asynchronous because you don't hold anyone up by not playing - the system never assigns a match. Instead, you just walk up to another rank 1 player and challenge them. They must agree to the match. The winner becomes rank 2, and the loser is "out". If you're out, you can reset back to rank 1, but only once/week. You can metagame your way through a few levels, but it takes an exponential number of co-conspirators to attain a given level. (I've simplified the system a bit. The full system is documented here.)

    Unfortunately, the Conflict Discipline was only popular with a very small number of players, and it's being replaced in ATITD 3.

    Note: A related problem is Judging Systems, where players rate in-game works of art. We've tried a number of algorithms there, and just recently have come up with one that