For score estimation practice, imagine an app or web page which retrieves filtered* games from OGS archives and shows the final board position, capture count and komi. Users estimate the score manually, perhaps points are awarded based on speed and/or accuracy. I’m geeking out on score estimation currently; I realize it’s not everyone’s idea of fun. When playing on a real board, I’ll thank myself for the practice
I’d be interested to see how I predict my own games if I saw a midgame position a few months after playing it (and not realizing that it was my own). Watching up to 2 dan I feel like I can pretty accurately predict, but my own games I’m out by +/- 30 points due to hubris or despair.
In my experience, dan’s games are much more complicated than mine (4k) and I have problems just stating if a group is dead or not. Very often they “throw” few stones in difficult places and leave them abandoned and I can’t tell if they’re captured or not.
With my games it’s usually easier unless I’m wrong about some tsumego (I think a group is dead but it can be saved or the other way around). That can set me off by some 30 points, but it’s my fault and I eventually recognise it in the endgame.
What I actually do sometimes is to evaluate the score balance and write it in a comment (private or Malko) in the game’s chat. So that later I can compare it to the score estimation by the AI, after the game is finished. That’s good practice. The comments have a reference to the move, so it’s easy to navigate them later.
I usually write something like:
it’s pretty even to me
whoa, I lost at least 20 pts here!
I think I’m behind by about 30 pts now
is that group in the corner alive? Should I put a stone there?
and so on…
It’s quite hilarious and instructive when I compare them to the AI after the game is finished.
It’s very easy to me because I only play correspondence, so I have plenty of time to count and to take notes. Doing that in a quick live game could be difficult though.