5 Comments
User's avatar
Wigan's avatar

Hi - nice analysis and great graphs.

Question on the lower number of mistakes in equal positions - could it be that there are also a large share of forced draws in those positions, which limits the possible scope of best moves? This might particularly be true when it's dead equal, which is an eval I often see when my best continuation was forcing a draw.

Expand full comment
Julian's avatar

I was more thinking about drawn endgames that players usually play on, like 3 vs 2 rook endgames. But forced draws may also play a role.

Expand full comment
Wigan's avatar

Makes sense.

Expand full comment
dboing dboing's avatar

really interesting comparative exploration, and highlight.

question: what are the different sample size behind the 2 histograms for the mistakes per evaluation. I might have to read carefully, and they might be in the text already.

more generally, how about termination depths for same top tier of player ratings strates (from the ceiling or the percentiles from there). That would be comparing statistical measures across pools, but that is what you are doing by saying you chose high level play. sorry.. thinking out loud as usual. I am just curious about no matter the level of play but for same relative strates in the 2 cases about depth of terminations statistics.

Expand full comment
Julian's avatar

I have roughly 3000 classical chess games and 350 chess 960 games. The main problem is that there aren't many chess 960 games with classical time controls.

I only looked at 2500+ games and didn't divide that range further as there were too few games for 960

Expand full comment