Author: Bruce Moreland
Date: 23:43:16 01/18/98
Go up one level in this thread
On January 18, 1998 at 17:56:48, Don Dailey wrote: >I discovered 126 "flaky" problems after correcting a bug in >my flake test routine. I recommend these all be taken out >because they imply ambiguous solutions. The ones you guy's >find also should be taken out. No, this is not right. You'll often have a case where the key looks good, shifts away for a while, then comes back for good, with a nice score. Heck, a move can appear to be a winning capture according to a SEE, but really be a losing capture because something is pinned, and as a consequence show up as a correct solution at ply 1 (since it is the first move searched), shift away later in ply 1, and come back when it's apparent that it is OK to go down material in this position. These cases can be handled by running the suite with decent amount of time, and not exiting until the end of the test. Just because a program evaluates a position as +2, then finds another +3 move, when the original move is really a mate in 8, is no reason to discard the problem. Just let it go longer and it should shift back, and you get a crappy solution time on this, and have something to fix, which is the point of this exercise. I don't think it is necessary to look for cooks yet, either. The first pass should simply delete problems that are solved by consensus. It's probably a good idea to just delete everything that all three of you guys solve in under 20 seconds. If you want mine, here they are: 8 11 14 16 18 19 28 29 35 43 49 51 54 60 63 66 68 71 75 76 78 80 81 82 84 87 91 99 109 112 115 117 119 123 128 130 131 135 137 138 142 144 145 146 147 148 153 155 159 160 161 162 180 182 186 187 190 191 196 198 201 202 206 209 211 212 213 215 227 228 231 237 247 250 251 259 262 266 267 270 271 274 277 280 286 287 289 291 292 297 300 302 310 311 312 313 315 316 318 321 327 330 332 335 336 337 338 339 345 348 349 353 354 355 356 357 358 360 361 366 367 370 371 373 375 376 377 379 390 394 395 396 397 398 399 401 402 406 410 412 416 419 420 422 424 427 433 434 438 439 440 443 445 447 459 460 463 464 465 467 468 469 470 472 474 475 476 482 485 488 489 491 492 494 497 498 499 500 501 503 504 505 507 510 512 516 519 525 533 539 544 551 553 557 560 561 562 565 566 569 579 580 582 587 589 592 593 594 595 597 598 600 601 603 605 607 608 610 613 614 615 616 617 621 623 624 627 629 634 635 636 637 640 644 646 647 648 652 653 654 659 663 665 666 667 672 673 674 675 676 677 679 680 681 683 684 685 686 687 688 690 691 692 693 696 697 698 699 701 706 709 712 713 714 715 716 718 719 721 722 725 726 727 728 730 733 734 735 736 740 742 744 745 750 754 755 756 761 762 764 766 767 769 770 771 772 773 774 776 779 780 785 789 790 791 792 795 796 799 801 803 804 805 810 811 814 815 816 818 822 824 825 826 827 829 830 831 832 833 834 836 837 841 842 843 845 846 852 853 854 856 858 860 861 864 866 867 868 872 875 877 878 879 These are all the problems at at least one version of my program has failed to solve in the past few months, at 20 seconds per position. There should be 363 of them. If this is too many, I can generate the same list for my most recent versions. bruce
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.