When will the chess programmers write an engine that plans ?

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
User avatar
mclane
Posts: 17050
Joined: Thu Mar 09, 2006 5:40 pm
Location: US of Europe, germany
Full name: Thorsten Czub
Contact:

Re: When will the chess programmers write an engine that plans ?

Post by mclane » Sat Mar 21, 2020 10:38 pm

acase wrote:
Sat Mar 21, 2020 9:19 pm
mclane wrote:
Fri Mar 20, 2020 10:22 pm
I mean that creates a plan and develops a main line that leads to something.
Not the usual engines we have today. That play chess within a Horizont of search depth.
Because Tactics are FAR more important than plans and strategy. If you (or a computer) make a strategical mistake against an opponent you could possibly save the game and not lose, however if you make one tactical mistake it could lose the game for you immediately. This is absolutely why tactics must be the main focus of all computer chess.
A strategical mistake also leads into a loss. This is nonsense.
Take the games of tactical weak LC0 versus tactical strong Stockfish and you see that tactics is not so important.
How else could LC0 beat Stockfish.
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....

User avatar
mclane
Posts: 17050
Joined: Thu Mar 09, 2006 5:40 pm
Location: US of Europe, germany
Full name: Thorsten Czub
Contact:

Re: When will the chess programmers write an engine that plans ?

Post by mclane » Sat Mar 21, 2020 10:40 pm

Yes they give a score. But how much is the score of an AB program worth when there is no tactics involved ?
You could ask your dog how much he would evaluate. It would be more appropriate.
When there is no tactics in a position, the AB programs guess a score. But how much is this true when the program is calculating
Millions NPS. Cannot be a big evaluation or analyse if it happens so fast, or ?
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....

acase
Posts: 728
Joined: Wed Mar 15, 2006 12:14 am
Location: Columbus, Ohio USA
Full name: Andrew R Case

Re: When will the chess programmers write an engine that plans ?

Post by acase » Sat Mar 21, 2020 11:01 pm

mclane wrote:
Sat Mar 21, 2020 10:38 pm
acase wrote:
Sat Mar 21, 2020 9:19 pm
mclane wrote:
Fri Mar 20, 2020 10:22 pm
I mean that creates a plan and develops a main line that leads to something.
Not the usual engines we have today. That play chess within a Horizont of search depth.
Because Tactics are FAR more important than plans and strategy. If you (or a computer) make a strategical mistake against an opponent you could possibly save the game and not lose, however if you make one tactical mistake it could lose the game for you immediately. This is absolutely why tactics must be the main focus of all computer chess.
A strategical mistake also leads into a loss. This is nonsense.
Take the games of tactical weak LC0 versus tactical strong Stockfish and you see that tactics is not so important.
How else could LC0 beat Stockfish.

LC0 is not tactically weak as you assert here. That may have been true in 2018 or even some of 2019, but when it made TACTICAL oversights back then it lost the games quickly.

Dann Corbit
Posts: 10765
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: When will the chess programmers write an engine that plans ?

Post by Dann Corbit » Sat Mar 21, 2020 11:26 pm

We may say that LC0 is weak in tactics, and it is true, compared to Stockfish.
But LC0 is at least as strong as Tal, tactically. (Analyze some Tal games, and it will find Tal's moves and furthermore it will find many mistakes that Tal made. Tal would stand no chance playing a game against LC0, and LC0 would also surprise Tal with some strikingly good moves).

Some may say that computers don't plan. But you can give it a game between two super GMs to analyze.
For more than half of the moves, the strong programs agree and would make the same move. For the remainder, it is usually the human who is wrong about it.

I can get to 50 pies in an hour. That's twenty five full moves forward. Imagine ten years from now what computers will do, since both hardware and software advances exponentially.

At some point, computers will see 100 plies ahead. No matter how you look at it, that's strategic.

I am at the opposite pole from McLaine. I think that we have, at our beck and call, the equivalent of Fischer, Tal, Morphy, Capablanca, Alekhine, etc. to analyze and explain chess to us, and do it tirelessly day and night. What a great time to be alive
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

jp
Posts: 1320
Joined: Mon Apr 23, 2018 5:54 am

Re: When will the chess programmers write an engine that plans ?

Post by jp » Sat Mar 21, 2020 11:37 pm

Dann Corbit wrote:
Sat Mar 21, 2020 11:26 pm
At some point, computers will see 100 plies ahead. No matter how you look at it, that's strategic.
It's not strategy or tactics. It's calculation.

mclane wrote:
Sat Mar 21, 2020 10:40 pm
You could ask your dog how much he would evaluate. It would be more appropriate.
If you really want what you ask for, the solution is to (get someone to) sponsor a prize for people to do that. A big enough prize will achieve far more than 100 frustrated threads asking the same thing.
Last edited by jp on Sat Mar 21, 2020 11:41 pm, edited 1 time in total.

carldaman
Posts: 1832
Joined: Sat Jun 02, 2012 12:13 am

Re: When will the chess programmers write an engine that plans ?

Post by carldaman » Sat Mar 21, 2020 11:38 pm

Check out this game that occurred as recently as yesterday, while I was typing up my previous post. :)



Notice the smooth flow of the game, with Nezh aiming gradually for the King by building up
the attacking position first, with logical human-like maneuvers (Rook lift, K-side pawn storm) then
followed by a spectacular trap with Rf8-f3!!? which may or may not be the best move, objectively,
but one that poses immediate problems, and then the smashing blow Nxd4 that finished off
poor Phalanx (a fine engine of GM strength, otherwise).
[Edit: this game was played entirely without a book]

Engines may not actually plan, but a combination of high-end search and a really good evaluation can produce what looks a lot like plans, for all intents and purposes. Hide the names of the players and show this game to a real GM, asking him to try to guess who had Black. :lol:

What impresses me about Nezh, besides the dazzling sacs, is this ability to build up powerful menacing positions. Without the preexistence of these positions, these sacs could never arrive at all. It reminds of what Spielmann said about Alekhine - (paraphrasing) that he could see the same sacrifices or combinations as well as Alekhine did, but that he was unable to bring about the underlying positions in the first place!

jp
Posts: 1320
Joined: Mon Apr 23, 2018 5:54 am

Re: When will the chess programmers write an engine that plans ?

Post by jp » Sat Mar 21, 2020 11:45 pm

carldaman wrote:
Sat Mar 21, 2020 11:38 pm
Engines may not actually plan, but a combination of high-end search and a really good evaluation can produce what looks a lot like plans, for all intents and purposes.
And the more one-sided you make the contest (e.g. give one engine 10 times the computer power of the other), the more it produces what superficially looks like "positional" play.

Dann Corbit
Posts: 10765
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: When will the chess programmers write an engine that plans ?

Post by Dann Corbit » Sun Mar 22, 2020 12:02 am

jp wrote:
Sat Mar 21, 2020 11:37 pm
Dann Corbit wrote:
Sat Mar 21, 2020 11:26 pm
At some point, computers will see 100 plies ahead. No matter how you look at it, that's strategic.
It's not strategy or tactics. It's calculation.
Both strategy and tactics are calculation.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

User avatar
Ovyron
Posts: 3958
Joined: Tue Jul 03, 2007 2:30 am

Re: When will the chess programmers write an engine that plans ?

Post by Ovyron » Sun Mar 22, 2020 12:46 am

Maybe some really smart human can check out an engine's tree in full and tell us about its plans. Perhaps the plans are there but we're just too dumb to see them.

carldaman
Posts: 1832
Joined: Sat Jun 02, 2012 12:13 am

Re: When will the chess programmers write an engine that plans ?

Post by carldaman » Sun Mar 22, 2020 1:04 am

jp wrote:
Sat Mar 21, 2020 11:45 pm
carldaman wrote:
Sat Mar 21, 2020 11:38 pm
Engines may not actually plan, but a combination of high-end search and a really good evaluation can produce what looks a lot like plans, for all intents and purposes.
And the more one-sided you make the contest (e.g. give one engine 10 times the computer power of the other), the more it produces what superficially looks like "positional" play.
My convention is to test with 4 cores if the engine supports it. If I had more available, I probably would not use them all because of the diminishing returns. Phalanx is a single-core engine, so it gets ... just one core. I don't test for a rating list, so I'm mostly interested in the stylistic content of the games - or the way the engines conduct their games, in other words.

But, to offer up a game played on equal hardware, here's an encounter with The King (v3.50) that concluded just minutes ago, as if on demand!! :D As a bonus, CyberNezh is slightly handicapped by the well-known issue involving The King's use of one core on the opponent's time, but no big deal there. :)

One rarely gets to witness such a thorough dismantling of the loathed Berlin Defense. [No book was used]
Enjoy.



OK, but now someone may object that such a contest is still lopsided. However, CyberNezh is far from being a top engine. Its moves are often too risky, bordering on downright unsound - but boy, how they can pay off against a bedazzled opponent. Besides, the most instructive games are often the master vs amateur, or SuperGM vs mere GM affairs.

Post Reply