Sting SF Black Hole 8 (AI: "scepticism & intuition")

[lech]

My Black Hole idea (“scepticism and intuition”) is to solve two possible and serious AI problems (nonlinear relations): stopping a bad and overestimated decision or getting a sensible “something from nothing” when there is no good decision.
Sting is also a “one brain in one head” engine (to play and solve in the same way).

download: https://www.mediafire.com/file/3qtgzp33 ... 8.zip/file

[Dann Corbit]

Thanks.
There are certain categories of problems where Sting really does excel.

[lech]

Now Sting is able to find checkmate in cyclic mechanisms (e.g. Troitsky's study - https://www.talkchess.com/viewtopic.php ... 38#p953438), directly. This is the next problematic and very important area for AI.

[lech]

Let me explain what makes Sting unique.
Imagine an ellipse divided into 3 areas (A, B, C) representing a logical (non-linear) problem. Area A is reserved for solved trials. Area C is for abstract problems. In chess this means: positions practically impossible to get in the game or worthless studies. Area B contains unsolved (bad choices) attempts according to the main algorithm. Sometimes this can be a very important (and dangerous) area for some logical problems.
The Black Hole idea works in this area and is mainly intended for chess studies. But the whole idea can be used elsewhere (other logical problems) as well.

[lech]

In reference to my previous text, the Black Hole idea can work unnecessarily when the main algorithm is able to solve a position on its own. For this reason, I added a very useful UCI option: AI (enable or disable the Black Hole idea), which is enabled by default.
Try the following position with both values of this option (clear hash!) to see the difference.
[pgn]
[Event "1st HM Katsnelson MT"]
[Date "2024"]
[White "V. Hergert"]
[Black ""]
[Result "1-0"]
[SetUp "1"]
[FEN "8/1k1n4/8/KP6/5p1p/6bP/4P1P1/3N1B2 w - - 0 1"]
[PlyCount "15"]
1. Kb4 Nf6 2. Kc4 Ne4 3. Kd3 {and f.e.} Nf2+ 4. Nxf2 Bxf2 5. e4 fxe3 6. g4 hxg3 7. Bg2+
Kb6 8. Ke2 1-0
[/pgn]

[Ajedrecista]

Hello:

In reference to my previous text, the Black Hole idea can work unnecessarily when the main algorithm is able to solve a position on its own. For this reason, I added a very useful UCI option: AI (enable or disable the Black Hole idea), which is enabled by default.
Try the following position with both values of this option (clear hash!) to see the difference.
[pgn]
[Event "1st HM Katsnelson MT"]
[Date "2024"]
[White "V. Hergert"]
[Black ""]
[Result "1-0"]
[SetUp "1"]
[FEN "8/1k1n4/8/KP6/5p1p/6bP/4P1P1/3N1B2 w - - 0 1"]
[PlyCount "15"]
1. Kb4 Nf6 2. Kc4 Ne4 3. Kd3 {and f.e.} Nf2+ 4. Nxf2 Bxf2 5. e4 fxe3 6. g4 hxg3 7. Bg2+
Kb6 8. Ke2 1-0
[/pgn]

Congratulations for your continuous work.

Excellent problem! Source:

https://www.arves.org/arves/images/PDF- ... T-2024.pdf

(Problem Nº 3). The PDF is in Russian language. Translation of the anotation of the problem with the help of Google Translator:

[...]

The idea of ​​locking the black bishop after two pawn sacrifices for en passant captures is excellent. The initial position is unpleasant, the white bishop on f1 is already locked, the bishop and two black pawns are standing in their places instead of successively moving to them. The introduction is boring, the king and knight monotonously approach the center of events. In addition, if black does not exchange knights, then after 3. … Nd6 4. e4 fxe4 e.p. 5. Kxe3 Kb6 there is still a lot of play to play before white wins. With a more lively opening game and a better initial position, the study could have won a prize.

[...]

Regards from Spain.

Ajedrecista.

[lech]

In reference to my previous text, the Black Hole idea can work unnecessarily when the main algorithm is able to solve a position on its own. For this reason, I added a very useful UCI option: AI (enable or disable the Black Hole idea), which is enabled by default.

Of course, the main idea of my work is to avoid any options. Only when the main algorithm can fail, Sting should run the Black Hole idea by itself. And it seems that this is the main problem of many AI projects. It is not always obvious, so it is possible that in these cases some tricks (parameters, mechanical methods, random hits) are proposed by programmers. However, I am far from saying that all AI projects are scams.