Sting SF 25

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

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Vinvin
Posts: 4848
Joined: Thu Mar 09, 2006 8:40 am
Full name: Vincent Lejeune

Re: Sting SF 25

Post by Vinvin » Sat Dec 19, 2020 5:39 pm

lech wrote:
Sat Dec 19, 2020 5:04 pm
Moreover, it is very important to know and understand that many positions (long solutions) should be solving by playing (auto-playing) not searching. :)
Don't you think the "auto-playing" algorithm can be implemented in the search ?

lech
Posts: 1100
Joined: Sun Feb 14, 2010 9:02 pm

Re: Sting SF 25

Post by lech » Sat Dec 19, 2020 7:37 pm

Vinvin wrote:
Sat Dec 19, 2020 5:39 pm
lech wrote:
Sat Dec 19, 2020 5:04 pm
Moreover, it is very important to know and understand that many positions (long solutions) should be solving by playing (auto-playing) not searching. :)
Don't you think the "auto-playing" algorithm can be implemented in the search ?
Of course, you are absolutelly right. It costs only one thread.

BTW. In JCER rating Sting 25 is + c.a. 150 ELO. A joke, but very nice! :shock: :D
Maybe, I can't be friendly, but let me be useful.

lech
Posts: 1100
Joined: Sun Feb 14, 2010 9:02 pm

Re: Sting SF 25

Post by lech » Sun Dec 20, 2020 11:52 am

But the main question is: for what it is to be done?
Sting is not a toy for machines, but is to be useful for chess funs.
I think that they use searching, auto-playing and skipping some obvious first moves instantly, when they see that it can be a proper solution.

Moreover, if anyone in a mechanical way reads solutions and can't understand (or skips) what happens on chessboard, is rather a lover of machines, only.
Maybe, I can't be friendly, but let me be useful.

jp
Posts: 1438
Joined: Mon Apr 23, 2018 5:54 am

Re: Sting SF 25

Post by jp » Wed Dec 23, 2020 2:43 am

lech wrote:
Sat Dec 19, 2020 5:04 pm
The main goal of my work is to prove that a chess (alpha-beta) engine is able, without some dirty tricks or external tools or selective code or artifical intelligence, to solve all positions from the above-mentioned "IQ" test.
I believe that it is possible to do (implement) it without any ELO loss too.
Moreover, it is very important to know and understand that many positions (long solutions) should be solving by playing (auto-playing) not searching. :)
But it solves your chosen test positions without resorting to auto-play, right?

lech
Posts: 1100
Joined: Sun Feb 14, 2010 9:02 pm

Re: Sting SF 25

Post by lech » Wed Dec 23, 2020 10:36 am

jp wrote:
Wed Dec 23, 2020 2:43 am
lech wrote:
Sat Dec 19, 2020 5:04 pm
The main goal of my work is to prove that a chess (alpha-beta) engine is able, without some dirty tricks or external tools or selective code or artifical intelligence, to solve all positions from the above-mentioned "IQ" test.
I believe that it is possible to do (implement) it without any ELO loss too.
Moreover, it is very important to know and understand that many positions (long solutions) should be solving by playing (auto-playing) not searching. :)
But it solves your chosen test positions without resorting to auto-play, right?
For the above-mentioned "IQ" test I had selected positions containing possible problems for some "classic" alpha-beta ("ELO") engines, and next I grouped it by reasons. Sting is a try to solve the positions (problems). In my opinion an engine solving all (100/100) the positions (of course without tricks) gets intelligent, in fact without useless here learning. Moreover the groups of problems are typical for creating an artifical intelligence, generally. The auto-playing way is mainly dedicated some very long solution positions. They are not selected to the
"IQ" test.
Maybe, I can't be friendly, but let me be useful.

jp
Posts: 1438
Joined: Mon Apr 23, 2018 5:54 am

Re: Sting SF 25

Post by jp » Thu Dec 24, 2020 6:47 am

lech wrote:
Wed Dec 23, 2020 10:36 am
For the above-mentioned "IQ" test I had selected positions containing possible problems for some "classic" alpha-beta ("ELO") engines, and next I grouped it by reasons. Sting is a try to solve the positions (problems). In my opinion an engine solving all (100/100) the positions (of course without tricks) gets intelligent, in fact without useless here learning. Moreover the groups of problems are typical for creating an artifical intelligence, generally. The auto-playing way is mainly dedicated some very long solution positions. They are not selected to the "IQ" test.
Okay, good (because auto-play probably counts as a "trick" too).

lech
Posts: 1100
Joined: Sun Feb 14, 2010 9:02 pm

Re: Sting SF 25

Post by lech » Thu Dec 24, 2020 12:18 pm

jp wrote:
Thu Dec 24, 2020 6:47 am
Okay, good (because auto-play probably counts as a "trick" too).
You are wrong. If you start the auto-play(ing), you don't change anything. It is only the other way (alternative) to get a solution, especially if it is very long and a "horizon effect" appears, and search(ing) is not able to do it. Sting does NOT contain any code dedicated this alternaive possibility.
The problem for user is probably that their GUIs don't have such a simple option. If you play match, you are selecting two engines. In this case you choose only one engine and commands go to this only one which plays as White and Black on the same "Hash table". If the authors of GUIs will add such an obvious and simple possibility, each engine can use auto-play(ing) instantly. They are ready to do it now!
It is not a trick. It is only a legal existing alternative possibility. :)
Maybe, I can't be friendly, but let me be useful.

jp
Posts: 1438
Joined: Mon Apr 23, 2018 5:54 am

Re: Sting SF 25

Post by jp » Tue Jan 12, 2021 6:54 am

lech wrote:
Thu Dec 24, 2020 12:18 pm
If the authors of GUIs will add such an obvious and simple possibility, each engine can use auto-play(ing) instantly. They are ready to do it now!
But we can always play engine vs. engine games from the position, and it can be the same engine playing against itself.

lech wrote:
Thu Dec 24, 2020 12:18 pm
If you start the auto-play(ing), you don't change anything. It is only the other way (alternative) to get a solution, especially if it is very long and a "horizon effect" appears, and search(ing) is not able to do it. Sting does NOT contain any code dedicated this alternaive possibility.
<snip>
It is not a trick. It is only a legal existing alternative possibility. :)
The issue is that the human has to start it auto-playing. That's human interference ("coaching")! Without human interference it would only know to try auto-playing with extra dedicated code, and you call extra code a "trick"! :)

lech
Posts: 1100
Joined: Sun Feb 14, 2010 9:02 pm

Re: Sting SF 25

Post by lech » Wed Jan 13, 2021 4:43 pm

The hashtable is not common for two (not one) engines and it can sometimes make a difference.
It is possible to make an auto-play(ing) manually. Start a search(ing), stop it, skip one move (it is possible to skip more moves too) and start search(ing) again (don't clear the hash table!), stop it, skip one move ....

Who starts and stops a serch(ing)? Machine? Not, a human!
An auto-play(ing) extends or replaces (alternative) a search(ing) only. It is not a trick!
Search(ing), as a rule, is not able to return the result (mate, stalemate, draw by the 50-rule).
The auto-play(ing) gets it.

Sting SF 26 is here: https://www.mediafire.com/file/e3efogxp ... 6.zip/file
Maybe, I can't be friendly, but let me be useful.

Post Reply