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

Only Chess Forum

  1. Subscriber sonhouse
    Fast and Curious
    02 Jun '16 22:26 / 1 edit
    I just wondered what the average is of predicted win percentage with a roughly 300 point difference, from the POV of the higher rated player.

    Also is the percentage the same if you are say 900 points V 1200
    as it would be for 1900 V 2200?

    I have been playing such a player 300 points lower and I am winning about 12 to 1, 24 games total, 2 losses. Is that about right statistically?
  2. 03 Jun '16 01:49 / 1 edit
    With a 300 point difference an 85% win rate is expected for the higher rated player. 24 wins and 2 loses puts you around 92%. So, you're doing a little better than expected.

    In my experience the higher up you go, the more likely the higher rated player will win.
    I'm rated 2200 and there's no chance in hell I'll beat a 2400 on here. Even though I'm supposed to have a 25% chance of winning, it's more like 0.1% because that's the odds they will leave the site on any given day.

    Edit: http://www.3dkingdoms.com/chess/elo.htm
  3. Donation ketchuplover
    G.O.A.T.
    03 Jun '16 09:34
    Ye gotta believe
  4. 03 Jun '16 13:46
    The odds with grades on RHP will differ..

    The grading percentage odds were calculated based on one game
    being played OTB with a time control of something like 40 moves in two hours.

    Here players are playing many games at once over a period of days.
  5. Subscriber sonhouse
    Fast and Curious
    03 Jun '16 14:28 / 1 edit
    Originally posted by KnightStalker47
    With a 300 point difference an 85% win rate is expected for the higher rated player. 24 wins and 2 loses puts you around 92%. So, you're doing a little better than expected.

    In my experience the higher up you go, the more likely the higher rated player will win.
    I'm rated 2200 and there's no chance in hell I'll beat a 2400 on here. Even though I' ...[text shortened]... odds they will leave the site on any given day.

    Edit: http://www.3dkingdoms.com/chess/elo.htm
    Yes, especially since most of the 2400 set here are just engines. It wouldn't matter whether you were 2500 OTB GM the engines here would still win.

    The player I have been playing, he challenged me so I am not trying to just up my rating artificially, he is around 14 bills and me around 17. Now 2 losses in 25, average 92% exactly it looks like.
  6. 04 Jun '16 12:10
    Here's a calculator, plug in any difference you want for the probability:
    http://www.bobnewell.net/nucleus/bnewell.php?itemid=279
  7. 04 Jun '16 13:37
    Originally posted by FishHead111
    Here's a calculator, plug in any difference you want for the probability:
    http://www.bobnewell.net/nucleus/bnewell.php?itemid=279
    This site states:
    "This form calculates the theoretical probability of winning a game based on ELO ratings".

    Not so.
    It calculates the match probability, not the individual game probability.

    A hundred games between a player rated 1200 and a player rated 900 should indeed result in a final score of 85/100, but this does not mean that the 1200 player has an 85% chance of winning any individual game, because of the possibility of draws.

    There are numerous ways to get a result of 85/100.
    At one extreme, the 1200 player wins 85 games, draws 0 and loses 15.
    At the other extreme the 1200 player wins 70 games, draws 30 and loses 0.

    To find the individual game probabilities, an average can be taken from the two extremes:
    Win probability: (85+70)/2 = 77.5
    Draw probability: (0+30)/2 = 15
    Lose probability: (15+0)/2 = 7.5

    So whilst the match probability between a 1200 and a 900 player is 85%, the actual probability of the 1200 player winning an individual game is 77.5%
  8. Subscriber sonhouse
    Fast and Curious
    04 Jun '16 15:07
    Originally posted by Tommovich
    This site states:
    "This form calculates the theoretical probability of winning a game based on ELO ratings".

    Not so.
    It calculates the match probability, not the individual game probability.

    A hundred games between a player rated 1200 and a player rated 900 should indeed result in a final score of 85/100, but this does not mean that the 1200 playe ...[text shortened]... 900 player is 85%, the actual probability of the 1200 player winning an individual game is 77.5%
    Is that a linear function, the same percentages if they are rated 2200 and 1900? or 2300 V 2000?
  9. 04 Jun '16 19:42
    Originally posted by sonhouse
    Is that a linear function, the same percentages if they are rated 2200 and 1900? or 2300 V 2000?
    Take a peek at the rating calc info in the FAQ.
  10. 04 Jun '16 23:11
    Originally posted by sonhouse
    Is that a linear function, the same percentages if they are rated 2200 and 1900? or 2300 V 2000?
    The match probability would be the same between two players rated 1200 and 900, as it would be between two players rated 2200 and 1900, 85% in both cases. so yes, the individual game percentages would be the same. And the same holds for any difference of 300 points between players.

    A difference of 200 points between players gives a match probability of 76%.
    Play two games against an opponent 200 points lower than you, and you could expect to win one and draw one.

    That's the theory anyway... I find that in practice I rarely manage a draw against anyone 200 points higher than me, and I find it hard to believe I'm achieving 25% overall against players 200 points higher. The next rainy day we get I might just check that.