A couple of questions about handicap and rating

BTW, on this part:

  • Up until a few months ago, OGS ratings (v5) assumed that a handicap of N stones equated to an effective rank adjustment of N
    • For 19x19, was mostly correct, but has the flaw you pointed out.
    • This was grossly incorrect for small boards.
  • Now, OGS ratings (still v5) is told directly what the effective rank adjustment should be (“handicap rank difference”), but it’s always an integer.
    • For 19x19, no change. It’s the number of stones. Still has the flaw you pointed out.
    • For small boards, this is the number in parentheses in the “Game Information” panel. E.g., a handicap 9x9 game might have “Handicap: 2 (Rank: 8)”; this means 2 handicap stones, but overall a handicap rank difference of 8.
    • This was landed as a bugfix to ratings v5 without recomputing historical ratings because it seemed worth doing quickly (to stop actively trashing ratings).
    • It’ll be fixed retroactively once ratings v6 lands.
    • FTR, proposed here and announced here.
  • goratings (v6, in progress) is smarter, and computes the effective rank adjustment from the combination of handicap stones and komi (and ruleset).
    • I.e., fixes the flaw you pointed out.
    • Allows rating system to evaluate non-standard komi in general (so, with this in place, we could have a policy change to allow rated games with arbitrary komi).
    • Will come in on the next major ratings change (v6), whenever that happens.
9 Likes