[d]r4rk1/2nqbppp/1nb1p3/p2pP3/PppP1B2/1P3N1P/2PQBPP1/R2NR1K1 b 0 1
Above is a position I posted 2 years ago.
At that point, no engine was able to offer a solution.
So, please, engine programmers, and enthusiasts, switch on your engines and tell me how black wins this?
PS. Please note, that again, as the position was composed by myself, or actually stems from a game of mine, I am not 100% certain black is winning this, but close to 98%.
OK, I will anticipate things a bit, as I have now some free time to post, but will still await intriguing engine output.
The point is that, if black does not find quick counterplay, and the position on the queen side gets blocked, white can actually win by attacking on the king side, where it has a substantial advantage.
So black must do something quickly, and the most reasonable thing is to sacrifice a piece on a4, getting 2 connected passers for it. I think the right way to sacrifice is Ba4, and not Na4, but still the position is sufficiently complicated to be completely certain.
At some point, after Ba4, you might get the following position:
[d]r4rk1/2nqbppp/4p3/p2pP3/n1pP1B2/1pP2N1P/3QBPP1/R2NR1K1 w - - 0 1
which I think should strongly favour black.
And the key to it all is the black passer on b3 - this is the strongest possible passer in chess.
What is so peculiar about the b3 passer?
Well, most engines would give passer bonus for b3, and also connected passer bonus additionally, but that is not enough. The b3 passer is due significantly more bonus, I would say some 50cps above all usual bonus points.
Why so?
Well, the reason is simple: b3 connects to a larger own pawn chain, f7-e6-d5-c4, which considerably decreases the enemy pieces' chances to detain this passer, as well as the own pieces' chances to organize quickly and bring up reserves, trying to promote it.
As said, and as supported by Carl, the longer chain, although this time pointed away from the enemy king, has the power to neutralise the enemy pieces' efficiency and boost the own pieces' efficiency in the process.
That means, black has time to improve, while white can not possibly do so.
In terms of specific eval, the b3 passer, as you see, is a passer, defended by an own pawn, c4, which own pawn is in turn defended by another own pawn, d5. That means, b3 is a passer, connecting to a larger own chain, which is sufficient evidence to assign an additional 50cps bonus.
Do you think 50cps is something to sneeze at?
Hardly, if it changes a lost position to a draw, and a draw to a win.
Again, this is one of the subtler positional terms engines hate to implement.
So again, I would give 50cps bonus for any passer defended by a pawn that is in turn defended.
A bit strange definition, but I think not difficult to implement, moreover that such positions are relatively frequent, if you know how to use them.
The bonus would be due both for the middlegame and the endgame, supposedly equal values, but not larger in the endgame, as longer chains are valid mainly in the middlegame. So contrary to most passers, this type of passer might be stronger in the middlegame.
So, for the time being, I leave it at that, very interested in any engine output, even if just for the sake of statistics, and especially in high-end machines output with stronger engines.
Any engine seeing a win for black after Ba4 or Na4, or at least in the second posted diagram?
After Bxa4 SF thinks that White is about half a pawn ahead, with many fail-lows and fail-highs though. White generally avoids the position with the b3 passer.
lantonov wrote:After Bxa4 SF thinks that White is about half a pawn ahead, with many fail-lows and fail-highs though. White generally avoids the position with the b3 passer.
[d]r4rk1/2n1bppp/1n2p3/3pP3/qppP1B1P/p4N2/2PQBPP1/R2NR1K1 w - - 0 1
It is not clear to me how white is going to stop the 2 black passers, as b3 threatens.
It is also not clear to me what can white achieve on the king side, I think nothing, as the valuable d3 square, where Be2 might go to attack the black king, is under the control of the d5 pawn, part of the longer chain.
It is not an easy position, but I think black should win this.
It might be interesting to see a longer search with Louis' 16-core...