Upgrade to Chess.com Premium!

Chess rating system


  • 10 days ago · Quote · #1921

    Florings

    Hey, guys... I have a problem: my rating was somewhere 1100 like 4 hours ago, i havent played since then and now its suddenly 900... why is that?

  • 10 days ago · Quote · #1922

    SmyslovFan

    In the real world, if you never played, you would be UNRATED. This isn't the real world. 

    Your rating starts at 1200 unless you're a titled player.

  • 10 days ago · Quote · #1923

    SmyslovFan

    Florings, your standard rating is 11xx, your blitz rating is 9xx. The ratings are different for different time controls.

  • 7 days ago · Quote · #1924

    philosoraptorsz

  • 5 days ago · Quote · #1925

    ChesstyMik

    Hi, can anyone tell me 'how many moves must be played before a timeout is counted as victory/loss', on this site? Is it 3? (each)

  • 5 days ago · Quote · #1926

    baddogno

    ChesstyMik wrote:

    Hi, can anyone tell me 'how many moves must be played before a timeout is counted as victory/loss', on this site? Is it 3? (each)

    All I could find was this:

    Timeouts - No effects until four moves

  • 5 days ago · Quote · #1927

    shamsolhoss

    mznor wrote:

    In playing my first game on Chess.com, I received a rating of 1200, before I played. Why, and how was that arrived at?

    hi

  • 5 days ago · Quote · #1928

    shamsolhoss

    mznor wrote:

    In playing my first game on Chess.com, I received a rating of 1200, before I played. Why, and how was that arrived at?

    hi

  • 4 days ago · Quote · #1929

    ChesstyMik

    thx 'dog

  • 4 days ago · Quote · #1930

    Ed_Seedhouse

    Chess.com assigns a "provisional" rating of 1200 to every new entrant to the rating pool.

    This is quite different from the original Elo system as used by, for example, the USCF, which estimates a provisional rating based upon one's results over a few games.

    I don't like this system of assigning a provisional rating a-priori since it means that if you are much stronger than the average player you will increase your rating at the expense of everyone else in the rating pool and establish a very low average rating for the pool as a result.  With modern computers there is no real excuse for taking this shortcut, in my opinon.

    Still this will not actually affect the predictive power of the local rating system since the difference between ratings is what is used to actually predict results, not the absolute rating level.

  • 4 days ago · Quote · #1931

    DavidMertz1

    Ed_Seedhouse wrote:

    I don't like this system of assigning a provisional rating a-priori since it means that if you are much stronger than the average player you will increase your rating at the expense of everyone else in the rating pool and establish a very low average rating for the pool as a result.

    I'm not so sure about that.  If you start everyone at 1200, then the average rating should be 1200, right?  Unless the average strength of a new player becomes stronger, this really isn't a problem.  And even then, as you say, it's the difference in ratings that matters.

    Someone who's only played a couple of games is not going to have an accurate rating no matter what system you use, because there aren't enough data points.  If I beat a 1200 player in my first game, should my rating be 1361 or 1600?  There's no way the ratings system can know.

  • 4 days ago · Quote · #1932

    Ed_Seedhouse

    DavidMertz1 wrote:

    Someone who's only played a couple of games is not going to have an accurate rating no matter what system you use, because there aren't enough data points.  If I beat a 1200 player in my first game, should my rating be 1361 or 1600?  There's no way the ratings system can know.

    That would depend on whether you take a Baysean approach or not.  On pure probability with only one piece of data the "best" estimate is still that bit of data.  If in your first game you perform at 1600 the best probabilistic estimate of your rating is 1600 but with extremely wide error margins.  If you take a Baysean approach your "prior" will be 1200 at chess.com and the resulting rating will be discovered by plugging both figure into Bayes formula and taking the result.  But so far as I know none of the common rating systems use Bayes.


Back to Top

Post your reply: