| |
BGonline.org Forums
Playing Under a 2.0 in the Near Future
Posted By: Timothy Chow In Response To: Playing Under a 2.0 in the Near Future (Stick)
Date: Friday, 22 February 2013, at 4:16 a.m.
O.K., I guess maybe I misunderstood what you were saying.
1. Suppose one analyzes a match using a procedure that has the following general form. You first analyze on Setting A. If Setting A says that something is not an error, then you let it stand as a non-error. If Setting A says that something is an error, then you appeal to Setting B. If Setting B then says that it's not an error, then you let it stand as a non-error. (XG's analysis settings seem to follow this general form.) Then the result will be a lower error rate than if you analyze every decision on Setting B only and take that as gospel. This will be true for the simple reason that in the former scenario, you get two chances to appeal any error and you get a pass if either judge agrees with you. It doesn't matter whether the game is a low-PR game or whether Setting A is stronger or weaker than Setting B. I think now that this is what you were pointing out, and if so then I agree with you.
2. There is, however, a closely related fallacy that sometimes comes up. If you take a bunch of matches that are analyzed on Setting A, and then pick only the low-PR matches to re-analyze on Setting B, then in general the error rate of those matches will go up. This is a general fact and has nothing to do with whether Setting B is stronger or weaker than Setting A. In particular it does not follow that Setting B will in general yield higher error rates than Setting A. Furthermore, this effect kicks in only if you do all the analysis first and then pick the low ones from the batch for re-analysis. In particular, if you were to take all your money games from your proposed session and re-analyze them on Setting B, one should not expect the error rate to go up, even if your PR is low among the general population. It's only if you take, say, the lowest of your low PR games and re-analyze them that the error rate will tend to go up. If you instead pick your worst games and re-analyze them, then the re-analysis will tend to bring the error rate down.
Having said all that, I admit it's still possible that there are some settings that yield higher error rates in general than other settings. I haven't seen any good evidence of that, though, because the anecdotal evidence I've heard is usually tainted by the above selection bias and therefore tells us nothing.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.