BGonline.org Forums

Error rate competition proposal

Posted By: Timothy Chow
Date: Friday, 16 October 2009, at 2:19 p.m.

I'm with those who regard error rates as a great tool for self-study, but silly for competition purposes. However, here is a suggestion for how to conduct such a team competition that I think addresses many of the common concerns.

Let n be the number of your unforced checker plays during a game, divided by ten and rounded down. (In the category of "forced plays" I include not just checker plays that can be played in only one way, but also any plays executed after the outcome of the game has been mathematically clinched.) Immediately after the end of a game, your opponent, without consultation or bot assistance of any kind, selects n of your checker plays. These are then subject to full rollouts, according to rollout parameters agreed upon in advance. You, of course, get to do the same to your opponent. The total equity lost through your plays is added to your team's error total.

Since the opening is especially subject to intentional equity losses, each player's first two moves (i.e., the first four moves of the game) are excluded from being selected.

At the end of the competition, the error totals of the two teams are compared. If the higher error total is less than 5% higher than the lower error total, then the error competition is declared a tie. This is to allow for roundoff error, luck, bot imperfection, etc.

I don't have a good solution for cube decision error rates, and suggest that cube decisions be excluded.