This is where top Engines fail compared to human............

[Lyudmil Tsvetkov]

[d]6k1/p7/8/4q3/8/P4R2/6P1/6K1 w - - 0 1

But this one is a draw, is not it?

only thing being changes is black has a pawn instead of b pawn.

What we are looking for is to find an implementable algorithm to solve fortresses.

Yes it's a draw.

Thanks, then what would be your algorithm to successfully solve R+2 pawns vs Q + pawn fortresses, with no exceptions?

[Lyudmil Tsvetkov]

As I said, I already thought so too, maybe I should have asked more clearly what I wanted to ask, hoping I wouldn't have to read each and every posting having lead to the position of the dia.

What's the position in question, if there still is any, unclear about fortress or not?

Sorry for my lazyness and yet joining so late in the very long thread, if it doesn't matter any more anyhow, you just need not answer again and I leave it for good too, still having spared a lot of time not reading all the rest.

It is not a position, we are discussing if SF is buggy or not.

[peter]

It is not a position, we are discussing if SF is buggy or not.

Thanks. Lyudmil.
I thought it would still be the one or the other postion questionable as for the objective outcome.
If not so, I'm out of that again, cause I already know for sure that SF is buggy, even if not as buggy as my own evals and outputs are at most of the postions in question.

[Lyudmil Tsvetkov]

OK, my algorithm for Q vs R, after the feedback, would be the following:

R+pawn vs Q:

- pawn on b2,c2,d2,e2,f2,g2
- pawn defending rook
- king adjacent to pawn on the 1st or 2nd ranks

R+pawn vs Q + pawn

- pawn on b2,c2,d2,e2,f2,g2
- pawn defending rook
- king adjacent to pawn on the 1st or 2nd ranks
- queen side pawn on the file adjacent to the rook side pawn not past its 5th rank

R+ 2 pawns vs Q+pawn

case 1 - one rook side pawn on the king side and one on the queen side

- pawn on g2 or b2
- pawn defending rook
- king adjacent to pawn on the 1st or 2nd ranks
- rook defending pawn on b3 or g3 respectively that are blocked by enemy pawns on b4 or g4 or rook defending pawn on a3 or h3 respectively that are opposed by the queen side pawn

case 2 - both rook side pawns on one and the same side of the board

- pawns on f2 and g3
- g3 pawn defending rook
- king adjacent to f2 pawn on the 1st or 2nd ranks
- queen side pawn on the h or f files not past its 4th rank

same for pawns on c2 and b3, with black queen pawn on a or c files

Again, this looks like a crazy thing to implement, and no one knows how many exceptions would still be found.

Adding more pawns becomes even crazier enterprise, no straight programmer would ever implement such knowledge.

And this concerns only by far the easiest eg fortress algorithm.

That is why I think only way to proceed is relying on search.

On the other hand, R+minor vs Q is easy: if queen side has no passers and candidate passers or an obvious mating net/significant other advantages, it is a draw.

At least, you can safely scale any score that is lower than the material difference Q- (R+minor), meaning queen side has no other major relative advantages.

[Lyudmil Tsvetkov]

Ok, we know this is lost:

[d]6k1/5p2/7p/5R2/3q2P1/5P2/6K1/8 w - - 0 1

Problem is, if this is lost:

[d]6k1/5p2/8/5R2/3q2P1/5P2/6K1/8 w - - 0 1

and also this one

[d]6k1/5p2/7p/8/5R2/3q2P1/5PK1/8 w - - 0 1

[Lyudmil Tsvetkov]

I guess this one is also a draw:

[d]8/8/3k4/p3q3/Pp6/1P4R1/5PK1/8 w - - 0 1

How do you specify for this?

Anyone with 9-men tbs?

Louis, what does SF say about this?

[BBauer]

Lyudmil Tsvetkov wrote:

It is not a position, we are discussing if SF is buggy or not.

And we are discussing the failture rate in difficult positions of the lomonomoronosow table bases.
And we discuss who has the 8-piece table bases.
Feel free not to read anything here.
You save a lot of time.

Kind regards
Bernhard[/quote]

[peter]

And we are discussing the failture rate in difficult positions of the lomonomoronosow table bases.
And we discuss who has the 8-piece table bases.
Feel free not to read anything here.
You save a lot of time.

I see Bernhard, you confirm my very first suspicion, but now it's too late, I already did read too much.
The first position Lyudmil gave in his last posting is a lost and the third one a drawn SF+Peter 8men position and the second one is a
drawn 7men lomonomoronosow tb.

[Lyudmil Tsvetkov]

And we are discussing the failture rate in difficult positions of the lomonomoronosow table bases.
And we discuss who has the 8-piece table bases.
Feel free not to read anything here.
You save a lot of time.

I see Bernhard, you confirm my very first suspicion, but now it's too late, I already did read too much.
The first position Lyudmil gave in his last posting is a lost and the third one a drawn SF+Peter 8men position and the second one is a
drawn 7men lomonomoronosow tb.

Thanks Peter.

That is what I supposed.

Problem is, what is this one:

[d]6k1/6p1/8/6R1/3q1P2/6P1/6K1/8 w - - 0 1

and this one:

[d]8/5k2/4p3/3q4/4R3/5P2/4PK2/8 w - - 0 1

I suppose they are both draws.

In such case, if they are indeed draws, we have a simple rule for Q+pawn vs R+ 2 pawns:

- pawn defends another pawn that defends rook
- rook on 4th or 5th ranks
- king adjacent to less advanced own pawn on ranks 1 or 2
- queen side pawn on a file adjacent to the file where the pawn defending rook is

= draw in every case

If both of the above positions are indeed confirmed draws, it is unbelievable what defence among pawns and pieces can actually do!

[Lyudmil Tsvetkov]

I suppose this one is also a draw:

[d]8/3pk3/2q5/3R4/4P3/3PK3/8/8 w - - 0 1

if this is the case, which I am almost certain, then indeed this is a very good rule, which could easily be specified, but will unfortunately reflect just a small portion of all rook fortresses:

-pawn defending pawn which in turn defends rook, which rook is on rank 4 or 5
- enemy pawn on a file adjacent to the file where the pawn defending rook is

= simple draw

Absolutely unbelievable what defence among pieces can do!

And engines do not score piece defence almost in any way.

Most of imbalance evaluation has to do with piece defence.