| |
BGonline.org Forums
Revised chart
Posted By: Tom Keith In Response To: Swiss vs random draw (Bob Koca)
Date: Thursday, 14 March 2013, at 3:59 p.m.
Were much more trials done for the Swiss? The results for the random have many instances of non monotonicity.
I checked this and found a bug. My intention for the Swiss simulation was that if two or more players had the same win-loss record, who-plays-who would be determined at random. And if there was an odd number of such players, the odd man out (to play someone with a better or worse record) would be chosen at random. This randomization wasn't working properly.
The revised (now hopefully correct) chart is shown below. The revised correlations are 0.3849 for the straight random draw versus 0.3853 for the Swiss draw.
Player P(win) Random Swiss 1 .050790 .052823 2 .048250 .050150 3 .046016 .048440 4 .044300 .045386 5 .042116 .043569 6 .040619 .041533 7 .038195 .039404 8 .036812 .037435 9 .035035 .035732 10 .033702 .034094 11 .031666 .032547 12 .030652 .030760 13 .028823 .029071 14 .027336 .027871 15 .025941 .026582 16 .025245 .025499 17 .023885 .023741 18 .022514 .022231 19 .021370 .021361 20 .020212 .020454 21 .019376 .019475 22 .018322 .018311 23 .017312 .017337 24 .016548 .016324 25 .015716 .015339 26 .014959 .014581 27 .014219 .013781 28 .013329 .013015 29 .012851 .012374 30 .012279 .011674 31 .011239 .010949 32 .010921 .010435 33 .010074 .009908 34 .009659 .009135 35 .008954 .008679 36 .008757 .008019 37 .008257 .007698 38 .007582 .007100 39 .007367 .006732 40 .006693 .006341 41 .006463 .005907 42 .006148 .005598 43 .005648 .005208 44 .005376 .004818 45 .004870 .004676 46 .004715 .004406 47 .004449 .004104 48 .004276 .003850 49 .003862 .003511 50 .003565 .003387 51 .003439 .003078 52 .003330 .002912 53 .003065 .002667 54 .002918 .002575 55 .002733 .002351 56 .002505 .002150 57 .002319 .002041 58 .002153 .001883 59 .002119 .001831 60 .001841 .001672 61 .001795 .001499 62 .001570 .001411 63 .001555 .001303 64 .001393 .001272
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.