English Chess Forum

"I'm slightly surprised no one appears to have let engines loose on the 960 initial positions with a view to discovering any instant wins. Perhaps they have and have kept the results of the analysis under wraps."

There are apparently some positions where white has a significant advantage. There also messy ones, where the Q is on "a" and a B on "h", where 1.g3 makes like difficult for black, who probably has to play a5 to get the Q out, of course black can just go 1....g6.

Kevin Thurlow wrote: ↑

Wed Sep 16, 2020 9:10 am

There are apparently some positions where white has a significant advantage.

This has been stated without proof for years, but nobody with the necessary backup can be bothered to check, or at least to publish their results. You would think AlphaZero could do it in a few hours.

"This has been stated without proof for years, but nobody with the necessary backup can be bothered to check, or at least to publish their results. You would think AlphaZero could do it in a few hours."

That's why I cautiously said, "apparently"! I haven't found any theory for 960, but there must be some...

Maybe when there are early resignations but using engines is cheating.

NickFaulks wrote: ↑

Wed Sep 16, 2020 7:29 pm

Kevin Thurlow wrote: ↑

Wed Sep 16, 2020 9:10 am

There are apparently some positions where white has a significant advantage.

This has been stated without proof for years, but nobody with the necessary backup can be bothered to check, or at least to publish their results. You would think AlphaZero could do it in a few hours.

Probably a bit more: a one minute evaluation per starting position would take more than 15 hours.
A one hour per position would require 40 days of non stop crunching.

Even the latter is probably "doable" by somebody somewhere you might have thought.....

It’s been done - a couple of years ago - by the Sesse supercomputer

Article

Sesse analysis spreadsheet

Matt Fletcher wrote: ↑

Thu Sep 17, 2020 7:51 pm

It’s been done - a couple of years ago - by the Sesse supercomputer

Article

Sesse analysis spreadsheet

Thanks, how had I never seen that? I have been bleating for some time that it really should exist!

I find the spreadsheet more interesting than the chat. The next question is, for the top example, is a +.67 evaluation sufficient to remove this starting position from the 960? I leave this to serious players who understand these things far better than I do, but it is an important point if 960 is, as I hope, to become a serious option.

NickFaulks wrote: ↑

Thu Sep 17, 2020 9:13 pm

The next question is, for the top example, is a +.67 evaluation sufficient to remove this starting position from the 960? I leave this to serious players who understand these things far better than I do, but it is an important point if 960 is, as I hope, to become a serious option.

It might be better expressed the way Alpha Zero did it as a winning percentage. In conventional chess, 0.67 is "White is (much) better".

In the CC event, you play each opponent with both colours in the same starting position, so +0.67 in one game is -0.67 in the other. We are playing different starting positions against each opponent however.