Please turn on javascript in your browser to play chess.
Help Forum

Help Forum

  1. 06 May '08 19:47
    I've read the FAQ and still can't help but think that there is something wrong, so I would appreciate if someone could check that I am understanding and calculating correctly.
    I am not a provisional player, but my rating has just dropped after beating a higher rated player.
    This is not the first time that I have noticed this happen, but is the first time I have tried to figure out why.
    The FAQ I think could be slightly clearer in that it does not make clear whether new rating = ((old rating + K) x (score - win expectancy)), or, (old rating + (K x (score - win expectancy))).
    Sorry about my excessive brackets, but I just want to be perfectly clear.
    Anyway, I can get neither to agree with what seems to have just happened to my rating.
    I've just won a game ID 4900749, resulting in my rating dropping from 1741 to 1720.
    Looking at my opponents profile page, his rating previous to our game ending was 1754.
    Assuming that the second bracket option above is correct (I hope!), then I calculate that my win expectancy was 0.092789749 and that my rating should have increased to 1770.
    Advice of where I am going wrong would be appreciated.
  2. Standard member John of Reading
    Scotch addict
    06 May '08 19:55
    The last two entries on your rating graph are telling me:

    (1) That you lost game 4718404 to an opponent rated 1622, and your rating dropped from 1741 to 1720. This is the final downward step on the graph.

    (2) That you won game 4900749 against an opponent rated 1756, and your rating rose from 1720 to your current rating, 1738. This doesn't show as an upward step on the graph; your new current rating is only shown in the text area at the top of the profile page.

    Does that make sense?
  3. Standard member Phlabibit
    Mystic Meg
    06 May '08 19:59
    Originally posted by John of Reading
    The last two entries on your rating graph are telling me:

    (1) That you lost game 4718404 to an opponent rated 1622, and your rating dropped from 1741 to 1720. This is the final downward step on the graph.

    (2) That you won game 4900749 against an opponent rated 1756, and your rating rose from 1720 to your current rating, 1738. This doesn't show ...[text shortened]... rating is only shown in the text area at the top of the profile page.

    Does that make sense?
    I think you are right.

    Game id 4900749
    Player's rating 1720
    Opponent's rating 1756
    Game date 05 May '08

    The player was rated 1720 and has gone up. Next game finished should show the spike upwards.

    P-
  4. 06 May '08 21:27
    Ok. So, if I am understanding correctly, then hanging your pointer over a bar on a ratings graph displays, above the graph, the "old rating"s that have been used to calculate the "new rating"s that result from that game.
    However, if that were the case, then, assuming that the 2nd set of brackets in my original post are correct, I calculate that my current rating should be 1749, as follows.
    1720 + {32 x [1 - (1 / 10 ^ {[(1756 - 1720) / 400] + 1})]}
    Am I still not fully understanding, or is there a problem with the explanation in the FAQ?
    In any case, I think the FAQ needs clarifying with more brackets in order to be mathematically correct.
    Alternative explanations are that there is a problem with either my or RHP's maths/math.
    Hope it doesn't appear that I'm just trying to up my rating.
    The stage I am at, I would prefer my rating not to be artificially inflated.
  5. Standard member John of Reading
    Scotch addict
    07 May '08 06:07
    Ah yes, the brackets in your formula aren't correct. I wrote myself a little spreadsheet:

    Column A - the difference in rating eg 36 when playing a stronger opponent, -36 when playing a weaker opponent

    Column B - the win expectancy =1/((10^($A2/400))+1)

    Column C - your rating change on a win =INT(32*(1-$B2))

    Column D - draw =INT(32*(0.5-$B2))

    Column E - loss =INT(-32*$B2)

    For a 36 point rating difference, this calculates the win expectancy as 44.8% and the rating change as +17. If RHP gave you +18 then I've probably got the rounding step wrong.
  6. 07 May '08 11:46
    John,
    Thank you very much.
    I really appreciate it.
    I can now see exactly where I've been going wrong.
    I was adding 1 to the figure that I was applying as the power of ten, rather than adding it after applying the power of ten.
    The silly results I was then getting were making me think that the second bracketing variation in my original post must be corrrect when of course it is not.
    Sorry if I have put you to a lot of unnecessary work, and thanks again.
    Richard