Has anyone tried to objectively measure differences between ranks on OGS?

That’s more or less correct. It was mostly my games, but maybe a third were other people’s games that I grabbed from OGS (all within the 4k - 3d range). Also, I wasn’t paying close attention to the “size” of the mistakes as in the point value, but more the type of mistake.

I don’t have the precise aggregated numbers on hand, because this really wasn’t the intent of my experiment, though that part of my original post appears to have been a lightening rod in the comments. But here’s some rough examples of the insights that I saw:

  • No dan player in my sample misread a single ladder, but the kyus missed about a dozen ladders, including several 1k’s
  • The kyus were about 3x as likely to lose points because they messed up a net (either played a net that doesn’t work, or failed to play one that does) compared to the dans - 1k’s were about 2x as likely as 1d’s
  • The kyus were about 4x as likely to lose points because they played a “puppy go” move in the opening / early midgame compared to the dans, with the 1k’s about 2x as likely as 1d’s

So the pattern is: 1k’s play more like the kyu cohort, but a bit better; 1d’s play more like the dan cohort, but a bit worse. And the gap between the cohorts is quite big.

It could all just be that my sample is too small and biased, or that my methodology is wrong - I don’t know and I don’t really care. I posted this because I’ve learned some interesting stuff about my own play - e.g. my invasion timing is more kyu-like than dan-like - and it would be cool if somebody has designed some automation which makes this analysis easier / faster.

1 Like