correspondence players wrongly trusted human moves

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Uri Blass
Posts: 8551
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

correspondence players wrongly trusted human moves

Post by Uri Blass » Wed Aug 14, 2019 3:16 pm

15.Nc1 is a human move and probably the first mistake of white in the following game

https://www.iccf.com/event?id=7670


Note that statistics for 15.Nc1 is bad but humans often played that move.

I believe that trusting human moves and ignoring statistics about results that seem to be what part of the correspondence players did is clearly worse than ignoring opening theory and trusting only computer moves more than 10 years ago.

If you use statistics about results in order to decide about the move then it was possible to do it also for engine-engine games in the past so I do not see that human theory was superior to computer theory even more than 10 years ago.



jdart
Posts: 3814
Joined: Fri Mar 10, 2006 4:23 am
Location: http://www.arasanchess.org

Re: correspondence players wrongly trusted human moves

Post by jdart » Wed Aug 14, 2019 6:13 pm

Your point is maybe not as clear as it might be, but I think you are referring to the fact that no previous correspondence players (in my DB, anyway) played Nc1, which is a real good indication that a move is bad or at least suboptimal. Some OTB players are good at openings but it is well-known that correspondence players generally really know openings.

--Jon

Uri Blass
Posts: 8551
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: correspondence players wrongly trusted human moves

Post by Uri Blass » Wed Aug 14, 2019 7:41 pm

jdart wrote:
Wed Aug 14, 2019 6:13 pm
Your point is maybe not as clear as it might be, but I think you are referring to the fact that no previous correspondence players (in my DB, anyway) played Nc1, which is a real good indication that a move is bad or at least suboptimal. Some OTB players are good at openings but it is well-known that correspondence players generally really know openings.

--Jon

I do not think correspondence players generally know openings.

Note that I never had best hardware in correspondence games but inspite of it I could get the ICCF GM title because my opponents probably also did not have top hardware and they also used the computer for less time than me.

my own moves were usually results of analysis with engines and only in rare cases I chose not to play the computer move and even in that case it was based on computer analysis.

jdart
Posts: 3814
Joined: Fri Mar 10, 2006 4:23 am
Location: http://www.arasanchess.org

Re: correspondence players wrongly trusted human moves

Post by jdart » Thu Aug 15, 2019 1:32 am

Ok, I get what you are saying. I do think some people rely on computer analysis too much: they expect it to find the "truth" of a position when the resolution of any issues in it is many plies away, and they aren't using enough search time to get an accurate score. But you are right: human move statistics are in many cases even less reliable.

--Jon

Peter Berger
Posts: 368
Joined: Thu Mar 09, 2006 1:56 pm

Re: correspondence players wrongly trusted human moves

Post by Peter Berger » Sat Aug 17, 2019 2:57 pm

This is relevant knowledge, Uri, but it is outdated.

A few months ago I talked with Matthias Kribben about "Fernschach" by accident. I know him in person because of a completely different context and the topic just so happened to come up in discussion. For once I felt pretty knowledgeable and eager to give my opinion, that is +somehow+ similar to yours in general.

I still feel that I know a little more about computerchess than him in general in hindsight, but I came to the conclusion that I'd probably just be able to draw with him at best under reasonable game conditions. People just don't do this anymore, follow random human analysis that never gets challenged seriously. To the contrary, people have probably become just a little too much believers in the verdict of computers, but I haven't been able to test this in practice yet.

Peter

Post Reply