Ranking calculation General forum

15 replies. Last post: 2004-10-14

Reply to this topic Return to forum

Ranking calculation
  • Jonatan Rydh at 2004-10-13

    Which rankings are used in the calculation of new rankings after a game is finished: the current rankings or the rankings the players had when the game started? If I am about to resign two games, does it matter which one I resign first?

    Another question: Is it possible to get no ranking increase in a won tournament game? I think the common way is to at least get +1 ranking no matter how much more ranking you got, but recently I won a game and got +0 while my opponent got -4. I always beleived you got as much points as the opponent lose and vice versa, but this seems to be a counter-example. Does it matter if opponent lose by time or by resign/ended game?

  • Tasmanian Devil at 2004-10-13

    Which rankings are used in the calculation of new rankings after a game is finished?

    The current ranking.

    If I am about to resign two games, does it matter which one I resign first?

    Possibly. If one of your opponents has a ranking close to yours and the other one much lower ranking, you may save one or two points by resigning to the lower ranked player first. Otherwise there is probably little or no difference.

    (…) recently I won a game and got +0 while my opponent got -4.

    If one player has finished many ranked games and the other one hasn't, the ranking will change more for the “newbie”. This prevents “painful” ranking losses for established players when new, but strong players register and start playing.

    Does it matter if opponent lose by time or by resign/ended game?

    Yes, I think there is an exception if the new player loses on time before making any moves - the established player then gains ranking points. Otherwise it doesn't matter how the game finishes.

  • Jonatan Rydh at 2004-10-13

    Thanks for the info. I find the idea that you get different final ranking depending on order of resigning and winning games a bit disturbing. Say you have 15 or 20 games going, some winning and some losing and some you are not sure whether you winning or not. Theoretically you could get much more ranking if you wait out the games you are sure of the outcome and play the unknown until all games are either won or lost (but not ended). After this you resign and win the games in correct order to get optimal final ranking! Maybe we are talking 40-50 points in difference.

    Now there are many reasons why this is absurd:

    1) Ranking isn't *that* important.

    2) It's rude to not resign games you lost.

    Still I find it a bit disturbing, there shouldn' be any difference.

  • Tasmanian Devil at 2004-10-13

    Ranking can be important in the qualification for championships. I dragged out a lost Reversi 10x10 game in order to qualify for 1st league. ;-D

    There's not much one can do about such “ranking speculation” I think. It's a consequence of turn-based play.

  • Jonatan Rydh at 2004-10-13

    Just out of morbid curiosity I calculated some figures how big the difference can be depending on when you end your games. Suppose you have 12 games going. Six of which you are winning with the opponents ranking 1500, 1600, 1700, 1800, 1900 and 2000 and six which you are losing with same six ranking levels. It is supposed you start with ranking 2000. It turned out the difference in this example is around 40 ranking points.

    Good final ranking with following order:

    • Win against 1500: (2000 -> 2001)

    • Lose against 1500: (2001 -> 1972)

    • Lose against 1600: (1972 -> 1943)

    • Lose against 1700: (1943 -> 1919)

    • Lose against 1800: (1919 -> 1900)

    • Lose against 1900: (1900 -> 1885)

    • Lose against 2000: (1885 -> 1874)

    • Win against 1600: (1874 -> 1879)

    • Win against 1700: (1879 -> 1887)

    • Win against 1800: (1887 -> 1899)

    • Win against 2000: (1899 -> 1918)

    • Win against 1900: (1918 -> 1933)

    Final ranking 1933.

    Not as good final ranking with this order:

    • Win against 1500: (2000 -> 2001)

    • Win against 1600: (2001 -> 2002)

    • Win against 1900: (2002 -> 2013)

    • Win against 2000: (2013 -> 2028)

    • Win against 1700: (2028 -> 2030)

    • Win against 1800: (2030 -> 2036)

    • Lose against 1800: (2036 -> 2012)

    • Lose against 1900: (2012 -> 1993)

    • Lose against 2000: (1993 -> 1978)

    • Lose against 1700: (1978 -> 1952)

    • Lose against 1600: (1952 -> 1923)

    • Lose against 1500: (1923 -> 1894)

    Final ranking 1894.

    The calculation formula for getting ranking is 15 points for win with the additional bonus or penalty of 1 point per 25 points difference between the two players, but no greater bonus or penalty than 14 points. The one losing the game loses equals amount of points the other one gains. I think this is the formula used, correct me if I am wrong.

    A general optimal (almost) strategy to gain maximum ranking seems to be to first resign all lost games starting with the lowest opponent rankings followed by higher rankings. After all lost games is resigned, start winning games beginning with opponents that have low ranking followed by higher rankings.

  • Tasmanian Devil at 2004-10-13

    No, the change in ranking is 32 / (1 + 10^(diff/400)) where diff is the ranking of the winner minus the ranking of the loser. (Or maybe 10 and 400 should be replaced by 2 and 120 respectively.)

  • Jonatan Rydh at 2004-10-13

    Thanks for the correction. I remade the calculations of the above example with the correct formula. Not suprisingly a similar result was gained, the difference between good and bad ordering was 41 ranking points. Also the same general strategy mentioned holds:

    • Lose against low ranked

    • Lose against high ranked

    • Win against low ranked

    • Win against high ranked

  • Abigail at 2004-10-14

    Well, yes, it's generally known that to it's better to first lose, then win, than the win first, then lose. Note however that your suggested tactics (“first play all the games till you can deduce who wins and who loses, then resign the lost games, then win the won games”) only work if your opponents play the game till the end. Your opponent might resign a game you have “won” before you have resigned your lost games.

  • Alan Turing at 2004-10-14

    Additionally, a player with higher ranking will meet tougher opponents and will have to fight harder. If the ranking is undeserved, it will quickly be distributed among the high-ranking players. ;)

  • Jonatan Rydh at 2004-10-14

    Abigail, yes it's true. It will take you need to be at turn at all your games. The opponent cannot resign games when you are at turn. However I do not recommend anyone to hold their games which are obvious won or lost, since it's both rude and annoying.

  • Jonatan Rydh at 2004-10-14

    Alan, would not a higher ranked than deserved, earn his ranking faster when playing against tough competition than a correct ranked which play more within his league? After all playing against higher ranked players than yourself is the fastest way to improve your game skill.

  • Alan Turing at 2004-10-14

    Yes, but there will still be a level where the the luck turns for the player, undeserved ranking or not. My claim is that both kinds of player will reach this level, although probably at different speeds. An honest player has nothing to worry about, since statistical laws will probably even out any boosted rank. A player who deliberately boosts the rating by resigning tactically, will have to stand to defend the higher rating like everyone else. Chances are that the fall will be much further if the player doesn't belong at the ranking level.

    As for who gets to play with who, there's always unranked play where anyone can challenge anyone else. It's a great, often underappreciated, way to learn, and it's not encumbered by ranking meta-strategies.

  • Jonatan Rydh at 2004-10-14

    Agreed. I'm just saying that the mere fact that there might be an advantage of having a “phony” rating (be it so or not) will encourage people to use “meta-strategies”, cheats or other foul play thus resulting in fluctuating the ranking system making it more unreliable than it allready is. Also unrated games might be useful for your own skill but they are not helping much to the ranking system since any play outside LG:s tournaments (or absence of play) will change your actual skill but not your ranking (directly).

  • alkosan at 2004-10-14

    without doing any calculations i expect the difference between optimal and worst strategy in the long run to be constant and well within “acceptable” distance.

    lets look at the optimal strategy. that would be LLLLLWWWWW at some point in will be inevitable to lose some games so the sequence repeats. LLLLLWWWWWLLLLLWWWWWLLLLLWWWWWLLLLLWWWWW etc… look at the middle of that sequence ….WWWWWLLLLLWWWWWLLLLL….., isn't that the worst strategy? :)

    also, by using the optimal strategy instead of the worst, after the first sequence you will have a higher rating and your opponents a lower rating. this means in the next sequence your wins will earn you less and your losses will cost you more.

    so, by applying the optimal strategy you will temporarily get an higher rating. in the long run however i doubt it's worth the trouble

  • Jonatan Rydh at 2004-10-14

    Yes you have an important point. In the long run the advantage you get will turn to a disadvantage. Easy come easy go… But I still beleive the average will be higher using the strategy mentioned. The reason being the sequence LLLLLLLLLWWWWWWWWLLLLLLLLWWWWWWWLLLLLLLLLLWWWWWWW have three optimal good orderings LLLLLLLLWWWWWWW and only two bad WWWWWWLLLLLLLLLL. Also it will allow you to temporarely have a higher ranking (at a peak) which will allow you to play with higher ranked players in a rated tournament.

Return to forum

Reply to this topic