Sting SF Black Hole 10

[lech]

Another attempt to achieve an independent AI chess engine (many improvements).
The Black Hole idea is described here: viewtopic.php?p=974329#p974329 .
This idea refers to positions such as in the "IQ" test: (http://talkchess.com/forum3/viewtopic.p ... 40#p941510). As a rule, such positions are very difficult (out of range) for classical heuristic and NN algorithms and need versions, parameters, tricks (human help).
download: https://www.mediafire.com/file/8z99q5gd ... 0.zip/file

[lech]

The "IQ" test is not part of Sting Black Hole idea. I only joined (selected) difficult positions posted here (Talkchess) by many users.

If someone writes "Sting IQ" test, it means he is a troll.

[lech]

In addition, I was far from adding here (Talkchess) any data (authors) to the items of this test ("IQ"). Many of these items are modified studies, and the idea of using them here is different from the authors.
Under pressure, I added a few names and I regret it.

[Dann Corbit]

Thanks. I got my copy

[lech]

There is a bug and it needs to be fixed: Sting SF Black Hole 10x.
This concerns only the commands: forward and auto-play. They are very important for autonomous intelligence, to prove (eventually) that a decision (first move) is good.

This concludes the serious part of my work. Now Sting (10) solves all the elements (of this type) from the "IQ" test.
I remind you that Black Hole is limited by the material and this is my subjective decision (no flaw).

download: https://www.mediafire.com/file/lf76ia96 ... x.zip/file

[lech]

I think it is the end of Sting. It proves (main goal) that heuristic algorithms supported by the Black Hole idea ("skepticism and intuition") can be considered AI (autonomous intelligence). For such AI it is very
important to solve as many (soon) items as possible from "area B" ( viewtopic.php?p=974421#p974421 ).
Of course, it would be nice to get a clear (independent of the source of the problem, here chess) algorithm of the Black Hole idea. But this is a problem for scientists, not for me, I can only be useful. By the way, it seems to be possible.

[lech]

I remind you that Black Hole is limited by the material and this is my subjective decision (no flaw).

The Black Hole idea works with the material: (<=) 2 * queen + 2 * rook + 2 * bishop.

[lech]

I remind you that Black Hole is limited by the material and this is my subjective decision (no flaw).

The Black Hole idea works with the material: (<=) 2 * queen + 2 * rook + 2 * bishop.

This applies to initial (starting) positions. Of course, it can be increased, but it makes no sense. Such positions (like in the "IQ" test) in the middle-game occur very rarely (statistically not at all), and mainly in studies (endgames), where the economy of pieces is very important.
Therefore, such a unique position as this: viewtopic.php?p=978138#p978138 will not be solved by Sting, what could be a problem for local trolls, mainly.

[lech]

Another valuable piece of information is that the idea of ​​Black Hole as an autonomous intelligence tool does not work with MultiPV parameter. Sorry, if it is a problem.

[lech]

I recommend (how_to_use.txt file) to use exactly 1024M HASH. I don't know how Sting (the Black Hole idea) works when you use more or less. 1024 M HASH seems to be a perfect value and can make a big difference.