Assorted tweaks

[Lyudmil Tsvetkov]

Same as dead knights, but adressed to bishops.

A dead bishop is a bishop on h2/g1 with own f2 and g3 pawns blocked by enemy pawns, or a2/b1 with own c2 and b3 pawns blocked by enemy pawns, or g2/h1/f1/h3 with own pawns on e2, f3 and g4 blocked by enemy pawns, or b2/a1/c1/a3 with own pawns on d2, c3 and b4 blocked by enemy pawns.

[d]4k3/b1p5/1pP5/1P4p1/5pP1/4pP2/4P1BK/8 w - - 0 1

Above, Bg2 and Ba7 are dead bishops.

As with knights, the penalty should be huge, some 80 to 200cps.

Again, this is different from both low mobility bishops on the edge and trapped bishops, following the same logic as with knights, as the condition for a dead knight is simply much more permanent.

In actual case, you are playing without a full bishop.

No, do not tell me this is stupid, because I have seen former versions of SF playing like this with very long TC.

And indeed, who would suppose Bg2 above is fully worthless, as it has 3! mobile squares, it is even not a low mobility piece, but is still worth nothing.

[Lyudmil Tsvetkov]

Well, everybody knows what a trapped bishop is: a bishop on a7 with enemy defended pawn on b6, or a bishop on h7 with enemy defended pawn on g6

[d]4k3/B1p2p1B/1p4p1/8/8/8/8/4K3 w - - 0 1

above, Ba7 and Bh7 are trapped bishops, some 100cps penalty, as obviously the bishop has already captured one pawn and will capture one more before it perishes

Penalty should be less than for a dead bishop, of course

To tell you the truth, in the last years I have not trapped even a single strong engine's bishop in this way, it simply does not happen, engines see it early in their search as the lines are shallow. Last bishop of an engine I managed to trap was many many years ago.

Not so with knights.
I manage to trap a strong engine's knight every 5th game or so.

What is a trapped knight?

A knight on g1 with own pawns on f2 and g3 that are blocked by enemy pawns or a knight on b1 with own pawns on c2 and b3 that are blocked by enemy pawns.

Penalty should be some 120cps or so, as the knight might be sacrificed for 2 enemy pawns at some point; still, that side will remain without a piece.

[d]rn2k1n1/2p2p2/1pP2Pp1/pP4P1/P7/8/8/4K3 w q - 0 1

Above, Ng1 and Nb1 are trapped knights.
As you see, Nb1 also traps its own rook on top of that. So this is a really bad situation that should be avoided.

As said, engines fully do not understand that, as the lines leading to winning the knight or other gains are deep. I manage to win every 5th game or so against strong engines like that - I simply build long chains towards the queen side, the engine is very active on the other side of the board and completely forgets about its knight.

I do not know which engine avoids that, of the top, I have won even games against SF like that, although I must admit that not all winning positions with that feature against SF actually result in wins.

One way or another, engines, even top engines, do not understand that.

[Lyudmil Tsvetkov]

Now, just a quick suggestion on how one could possibly fineevaluate positions with bishops vs knights.

In general, I think the most important rule is penalty for bishop for any own pawn on square the colour of the bishop, higher penalty for blocked central pawns on square the colour of the bishop.
This is what everyone knows and does.

Apart from that, I think there are only 2 possible working rules with much smaller, but still some real implications:

- give 5cps bonus for any own unopposed pawn the bishop side has; bishops like asymmetrical pawn structures, knights like symmetrical pawn structures

- give 10cps bonus for any blocked defended pawn after the 3rd to the knight side; these will more or less denote the position is closed or is about to close, where knights are better. It makes absolutely no sense to give bonus to the knight for any blocked pawn, blocked pawns must be chain/defended, and only when you have already 3 such blocked pawns, you give 10cps bonus for 4th, 5th such pawn, etc. Giving the bonus for any blocked defended pawn makes absolutely no sense, as there is no rule that says bishops are worse with less than 3 blocked defended pawns, but they definitely start being so, when the number of pawns increases.

Well, I do not know if this works, but it makes sense trying it. Knight vs bishop is such a difficult problem, that I think almost nothing apart from bishop same square colour pawns will work, but the above 2 rules are still pretty reasonable for finetuning purposes.

[Lyudmil Tsvetkov]

Well, there are also such kinds of pawns.

Any isolated pawn the square in front of which is controlled by 2 enemy pawns at the same time is an isolated backward/backwardfated pawn.

5cps penalty in such a case, small, as the condition is not very forceful, but still matters.

[d]6k1/8/8/3p4/8/2P1P3/8/6K1 w - - 0 1
d5 above is isolated backward pawn; obviously, it can not move without being lost

[Lyudmil Tsvetkov]

Well, the availability of at least a single shelter pawn on its 2nd rank should carry some bonus with it, as pawns on the 2nd rank are the best defenders.

20cps bonus in case you have at least one 2nd rank shelter pawn, 30cps in case this is a g2/b2 shelter pawn, g2/b2 shelter pawns are the optimal shelter pawns, but unfortunately, they are also one of the best storming pawns...

[Lyudmil Tsvetkov]

Well, any 5th rank pawn blocking a doubled defended enemy pawn is due some nice bonus, maybe 15cps or so.

In the first instance, such a pawn blocks/restricts 2 enemy doubled pawns early on, which is a good feature, and in the second, it makes backward the pawn defending the doubled pawn.

[d]6k1/1pp2pp1/1p4p1/1P4P1/8/8/8/6K1 w - - 0 1
g5 and b5 deserve the bonus above; g5 blocks g6 enemy pawn that is doubled and defended by f7 at the same time. Similarly for b5.

I think this is a very useful feature that could hardly fail.