What is a backward pawn?

[michiguel]

[d]6k1/8/p1p5/P1P5/1P6/8/8/6K1 w - - 0 1

How could we define b4 above?

Well, it seems simple. A backward-fated pawn would be a pawn that:

- is not opposed
- whose stop square is controlled by 2 enemy pawns
- and both of which enemy pawns are blocked by friendly pawns
- and there are no other friendly pawns on the same rank or behind on adjacent files

4 necessary conditions, a condition less simply will not do.

I would assign backward-fated pawns some 50cps penalty, way above the 30cps penalty for a normal backward pawn. And would not count ranks, as those pawns are very specific. It should be mentioned also that just adding the penalties for 2 normal backward pawns would not quite do it, as there is a quantitative difference.

If SF or some other engine applied the definition correctly, they would see it really makes sense to introduce such a peculiar pawn term. If engines do not understand backward pawns, they are fully clueless of backward-fated ones.

People have been laughing at me every time I mentioned backward-fated pawns, but do not you see they are real?

the b4 pawn is simply VERY weak. It can't move, period, until something is done about either the black a or c pawn.

I don't like the "backward-fated" term. What does it mean? That right NOW the white pawn can't advance even with piece support? What if a or c black pawn is removed? No longer "backward fated".

The "weak pawn" concept is easier. There are degrees of weakness. The above is an extreme example. But by recognizing that black having the a/c pawns present makes the white pawn very weak, black will try to maintain those two pawns and prevent white from any sort of liquidation/exchange that removes one.

Another case.

white pawns on a4/c4. black pawns on a6/b6. What about that b6 pawn? In Crafty, it is weak. Can't be defended by a pawn since all friendly pawns on adjacent files are either missing or advanced far enough they can not defend it. The pawn can advance but will be lost.

Now remove white c4 pawn. black pawn at b6 is weak, because it can't be defended by a pawn, and the square in front of it is attacked by pawns as many times as it is defended, which will leave black with an isolated (and passed) pawn on b5.

I like the general idea of mobility rather than trying to create multiple categories of pawns that are weak and then name each one something different (and often something confusing).

Right, it means that it can not advance even with the support of pieces.
Let is leave dynamic features to search, it is not clear if a6 and c6 will ever be removed, for the time being b4 can not budge.

[d][d]3r1n2/2p3k1/1p1q1p2/p1pPp1p1/P1P1P1P1/4BP2/5K1R/7Q b - - 0 32

Did you see that diagram? It is from the Komodo-Critter game in TCEC. White is winning this with a pawn less. And one of the main reasons for this is that b6 is backward-fated, or whatever you would like to call it. 3 white pawns on the queen side hold 4 black. So that b6 is due an enormous penalty, almost a full pawn. Do you want that engines do not consider features worth a full pawn? Why should we neglect such features and instead go after phantoms?

I bet Crafty does not consider b6 as weak. And does not penalise it with almost a full pawn. I also bet Crafty does not see white is winning here with a pawn less, just as Critter. Am I right?

But the irony is that white won advancing f4...

Miguel

[BeyondCritics]

My take on this is that you are somewhat mixing terms.

A pawn that is defended by a pawn is not considered backward in any book I have (g6 in your first example as it is defended by h7). The fact that it can not safely advance is not so important when it is perfectly safe where it is. If you want to talk about it being immobile, that is a different issue.

the "unopposed" requirement stems from the rook, in the books I have read over the years. A pawn that is backward (can not be safely defended by a friendly pawn) is generally considered to be worse if it is (a) on a half-open file (no enemy pawn in front of it AND (b) the enemy has rooks to attack it.

IMHO, a pawn that is backward and unopposed is not any weaker or stronger than a backward pawn that is opposed, if there are no enemy rooks to contend with. Any backward pawn is weak, and inviting to enemy kings in the endgame. backward pawns on half-open files make inviting targets even in the middle game if you have rooks to attack on the file.

I've seen games decided by two types of backward pawns, repeatedly.

The first is the classic backward pawn on a half-open file, which is just as bad as an isolated pawn on a half-open file, when the enemy has rooks. Pile up on the pawn, tying up the opponent by making him defend, and either win the pawn outright, or switch to some other idea once he is tied down and not coordinated very well.

The second is a pawn that is backward, but the file is not half-open. If you have just one, you might manage to defend it with your king, depending on the position. If you have two, the enemy king can make you commit to saving one and then to eat the other pawn chain. Doesn't matter whether the file is half open or not.

[d]6k1/5p2/4p1p1/3p2P1/1p1P4/p7/P7/6K1 w - - 0 1

I tend to concur with Bob's definition. In the diagram above the e6 pawn is still defended, and thus not vulnerable to direct attack, so that's an argument against calling it backward.

However, since it cannot safely advance, there is an element of backwardness about it, and the square in front of it (e5) can serve as a good outpost for White pieces. These considerations should justify some sort of a penalty for such a pawn, but a lesser one that a fully backward [undefendable by a fellow pawn] pawn on an open file.

The Black b-pawn, even though backward, could be pushed in some cases creating a dangerous advanced passed pawn. It's not as clear whether a backward pawn that sits in the opponent's half of the board deserves any significant penalty.

CL

Crafty calls that a3 pawn a "hidden passed pawn". GM Dzhindi used to call this a protected passed pawn himself, because a3 is certainly passed here when you think about it and it is defended by b4. after b3 black ends up with a passed pawn for certain.

You really do not read this thread.

When you add pieces, there is no danger of a passed pawn.

[d][d]6k1/5p2/4p1p1/n2p2P1/1p1P4/p7/P7/2N3K1 w - - 0 1
One knight each side added; where is the danger of the hidden passed a3 pawn?
a3 is not an immediate danger, but b4 is certainly backward, like it or not, a2 stops both a3 and b4 pawns by blocking a3 and attacking the stop square of b4.

Very simple. White has to avoid any move that lets me force the exchange of knights. Because allowing such is an instant loss. Unless white chooses to brink his king over to the a-file area, which ties it down and restricts white's ability to defend elsewhere.

Adding the knights does NOT eliminate the advantage of a3-b4, it just makes it harder to use that advantage. In this case, as black I would try to get the knight to c3 and take on a2. White knight will have great trouble with the a/b pawns after that

We are talking here about a static feature, and the static feature is that one white pawn, a2, holds 2 black pawns, a3 and b4, simultaneously.

Sorry, but you can't have it BOTH ways.

remove the knights.

Is b4 weak, or is a3 overwhelmingly strong? a3 is very strong so long as black can get enough pieces to attack b3 to make pushing b3 safe. That's not "static" at all, that is dynamic. So leave the pieces out of it. b4 is weak, as it can't safely advance, but there is a stinger in that b4 CAN advance because a3 will promote before white's pawn that takes on b3. In this position, the white pawn on a3 is not really holding back anything whatsoever. And, in fact, my program considers this a significant pawn advantage, not a weakness. Why? Ask GM Dzhindi. We had that discussion almost 20 years ago and he convinced me it is correct. Yes, Crafty still gets a penalty for b4 being weak, but it gets a significantly larger bonus for a3 being so strong.

Professor Hyatt, you taught me that it is better to leave dynamic issues to the search, so let us concentrate on the static features then.

Knights present are a natural feature, as in 95% of all situations there are pieces on the board instead of only pawns. The big limitation of the hidden passer concept is that it is constructed for and could be applied only in pawn endgames. Even one piece added already makes the rule non-functional. So this is an ad-hoc rule, bad at that, as it is valid in only 5% of cases.

Backward pawns, on the other hand, including giving penalty to b4, are valid in all situations where there are pieces apart from the pawns, i.e. in 95% of cases. This already seems to me like a good generalised rule, and not an ad-hoc rule.

Giving a3 a very nice bonus for its advanced status, either in psqt or otherwise, is another matter.

Not so fast. The "hidden passer" idea works in any reduced material endgame. If you can prevent b3 with a piece, fine. But that piece is tied down. By a lowly pawn. And if the piece moves away, particularly knights, the hidden passer can become a real passer that wins...

What we try to capture in pawn structure evaluation is an answer to the question "if all pieces are removed, who stands better?" Because the side that stands better is going to try to trade pieces to reach that situation, and the side that stands worse has to struggle to avoid the trades. So the pawn structure becomes an integral part of the game. In other positions, the opposite happens. A bishop vs knight endgame, where the pawn structure restricts the bishop but not the knight. So the side with the bishop wants to trade, the side with the knight would rather trade other pieces but not the knights.

Trying to come up with a general-purpose pawn structure evaluation is actually pretty difficult, because most of the terms have to be general-purpose, while humans don't think that way, we look at specific features. The more general-purpose the eval becomes, the more errors it will make in specific circumstances. Of course if you omit some of those general-purpose rules, it makes even more errors because it has no clue at all about that specific feature.

The evaluation of a chess engine will likely NEVER come even close to what a human does. But the computer is so fast, and searches so deeply, that it can get by with an inferior evaluation because the human has a very inferior tactical analyzer.

[d]6k1/7p/8/n7/1p6/p7/P5PP/2N3K1 w - - 0 1
Well, here is a reduced material endgame where it does not work. White does not have a hidden passer, just a humble candidate. But who is better? Where is the power of the hidden passer? The backward b4 pawn however is a valid deficiency even in this extremely reduced, but not pieceless endgame. It leaves white with an efficient pawn majority on the king side.

This is anything but a correct proof position:
i) You have granted yourself the right to move. Due to that alone white will conquer the fourth rank. This advantage alone can be decisive in knight endgames.
ii)You have misplaced the black knight to the edge.
iii) You put your knight in contact with b3-a2, preventing any breakthrough.

So if you are allowed to place your knight, i am allowed to place my knight. I just move my knight to b5:
[d]6k1/7p/8/1n6/1p6/p7/P5PP/2N3K1 w - - 0 1
1. Kf2 Nc3 2.Ke3 Nxa2! and white fights for the draw.
Due to advantage (i) this is still possible (according to stockfish)

What happens if i grant me the first move?

According to to stockfish, this gives me an decisive advantage (-246 cp)

[bob]

[d]6k1/8/p1p5/P1P5/1P6/8/8/6K1 w - - 0 1

How could we define b4 above?

Well, it seems simple. A backward-fated pawn would be a pawn that:

- is not opposed
- whose stop square is controlled by 2 enemy pawns
- and both of which enemy pawns are blocked by friendly pawns
- and there are no other friendly pawns on the same rank or behind on adjacent files

4 necessary conditions, a condition less simply will not do.

I would assign backward-fated pawns some 50cps penalty, way above the 30cps penalty for a normal backward pawn. And would not count ranks, as those pawns are very specific. It should be mentioned also that just adding the penalties for 2 normal backward pawns would not quite do it, as there is a quantitative difference.

If SF or some other engine applied the definition correctly, they would see it really makes sense to introduce such a peculiar pawn term. If engines do not understand backward pawns, they are fully clueless of backward-fated ones.

People have been laughing at me every time I mentioned backward-fated pawns, but do not you see they are real?

the b4 pawn is simply VERY weak. It can't move, period, until something is done about either the black a or c pawn.

I don't like the "backward-fated" term. What does it mean? That right NOW the white pawn can't advance even with piece support? What if a or c black pawn is removed? No longer "backward fated".

The "weak pawn" concept is easier. There are degrees of weakness. The above is an extreme example. But by recognizing that black having the a/c pawns present makes the white pawn very weak, black will try to maintain those two pawns and prevent white from any sort of liquidation/exchange that removes one.

Another case.

white pawns on a4/c4. black pawns on a6/b6. What about that b6 pawn? In Crafty, it is weak. Can't be defended by a pawn since all friendly pawns on adjacent files are either missing or advanced far enough they can not defend it. The pawn can advance but will be lost.

Now remove white c4 pawn. black pawn at b6 is weak, because it can't be defended by a pawn, and the square in front of it is attacked by pawns as many times as it is defended, which will leave black with an isolated (and passed) pawn on b5.

I like the general idea of mobility rather than trying to create multiple categories of pawns that are weak and then name each one something different (and often something confusing).

Right, it means that it can not advance even with the support of pieces.
Let is leave dynamic features to search, it is not clear if a6 and c6 will ever be removed, for the time being b4 can not budge.

[d][d]3r1n2/2p3k1/1p1q1p2/p1pPp1p1/P1P1P1P1/4BP2/5K1R/7Q b - - 0 32

Did you see that diagram? It is from the Komodo-Critter game in TCEC. White is winning this with a pawn less. And one of the main reasons for this is that b6 is backward-fated, or whatever you would like to call it. 3 white pawns on the queen side hold 4 black. So that b6 is due an enormous penalty, almost a full pawn. Do you want that engines do not consider features worth a full pawn? Why should we neglect such features and instead go after phantoms?

I bet Crafty does not consider b6 as weak. And does not penalise it with almost a full pawn. I also bet Crafty does not see white is winning here with a pawn less, just as Critter. Am I right?

Crafty does not consider b6 as weak. Neither do I. I just consider it immobile. I don't have to dedicate resources to defend it, but I can't move it which limits my options to the kingside. I'm not convinced this pawn is the reason black lost, however.

[Lyudmil Tsvetkov]

[d]6k1/8/p1p5/P1P5/1P6/8/8/6K1 w - - 0 1

How could we define b4 above?

Well, it seems simple. A backward-fated pawn would be a pawn that:

- is not opposed
- whose stop square is controlled by 2 enemy pawns
- and both of which enemy pawns are blocked by friendly pawns
- and there are no other friendly pawns on the same rank or behind on adjacent files

4 necessary conditions, a condition less simply will not do.

I would assign backward-fated pawns some 50cps penalty, way above the 30cps penalty for a normal backward pawn. And would not count ranks, as those pawns are very specific. It should be mentioned also that just adding the penalties for 2 normal backward pawns would not quite do it, as there is a quantitative difference.

If SF or some other engine applied the definition correctly, they would see it really makes sense to introduce such a peculiar pawn term. If engines do not understand backward pawns, they are fully clueless of backward-fated ones.

People have been laughing at me every time I mentioned backward-fated pawns, but do not you see they are real?

the b4 pawn is simply VERY weak. It can't move, period, until something is done about either the black a or c pawn.

I don't like the "backward-fated" term. What does it mean? That right NOW the white pawn can't advance even with piece support? What if a or c black pawn is removed? No longer "backward fated".

The "weak pawn" concept is easier. There are degrees of weakness. The above is an extreme example. But by recognizing that black having the a/c pawns present makes the white pawn very weak, black will try to maintain those two pawns and prevent white from any sort of liquidation/exchange that removes one.

Another case.

white pawns on a4/c4. black pawns on a6/b6. What about that b6 pawn? In Crafty, it is weak. Can't be defended by a pawn since all friendly pawns on adjacent files are either missing or advanced far enough they can not defend it. The pawn can advance but will be lost.

Now remove white c4 pawn. black pawn at b6 is weak, because it can't be defended by a pawn, and the square in front of it is attacked by pawns as many times as it is defended, which will leave black with an isolated (and passed) pawn on b5.

I like the general idea of mobility rather than trying to create multiple categories of pawns that are weak and then name each one something different (and often something confusing).

Right, it means that it can not advance even with the support of pieces.
Let is leave dynamic features to search, it is not clear if a6 and c6 will ever be removed, for the time being b4 can not budge.

[d][d]3r1n2/2p3k1/1p1q1p2/p1pPp1p1/P1P1P1P1/4BP2/5K1R/7Q b - - 0 32

Did you see that diagram? It is from the Komodo-Critter game in TCEC. White is winning this with a pawn less. And one of the main reasons for this is that b6 is backward-fated, or whatever you would like to call it. 3 white pawns on the queen side hold 4 black. So that b6 is due an enormous penalty, almost a full pawn. Do you want that engines do not consider features worth a full pawn? Why should we neglect such features and instead go after phantoms?

I bet Crafty does not consider b6 as weak. And does not penalise it with almost a full pawn. I also bet Crafty does not see white is winning here with a pawn less, just as Critter. Am I right?

But the irony is that white won advancing f4...

Miguel

That is easy to see.
It is difficult to assess the main position as winning for white though.

[Lyudmil Tsvetkov]

[d]6k1/8/p1p5/P1P5/1P6/8/8/6K1 w - - 0 1

How could we define b4 above?

Well, it seems simple. A backward-fated pawn would be a pawn that:

- is not opposed
- whose stop square is controlled by 2 enemy pawns
- and both of which enemy pawns are blocked by friendly pawns
- and there are no other friendly pawns on the same rank or behind on adjacent files

4 necessary conditions, a condition less simply will not do.

I would assign backward-fated pawns some 50cps penalty, way above the 30cps penalty for a normal backward pawn. And would not count ranks, as those pawns are very specific. It should be mentioned also that just adding the penalties for 2 normal backward pawns would not quite do it, as there is a quantitative difference.

If SF or some other engine applied the definition correctly, they would see it really makes sense to introduce such a peculiar pawn term. If engines do not understand backward pawns, they are fully clueless of backward-fated ones.

People have been laughing at me every time I mentioned backward-fated pawns, but do not you see they are real?

the b4 pawn is simply VERY weak. It can't move, period, until something is done about either the black a or c pawn.

I don't like the "backward-fated" term. What does it mean? That right NOW the white pawn can't advance even with piece support? What if a or c black pawn is removed? No longer "backward fated".

The "weak pawn" concept is easier. There are degrees of weakness. The above is an extreme example. But by recognizing that black having the a/c pawns present makes the white pawn very weak, black will try to maintain those two pawns and prevent white from any sort of liquidation/exchange that removes one.

Another case.

white pawns on a4/c4. black pawns on a6/b6. What about that b6 pawn? In Crafty, it is weak. Can't be defended by a pawn since all friendly pawns on adjacent files are either missing or advanced far enough they can not defend it. The pawn can advance but will be lost.

Now remove white c4 pawn. black pawn at b6 is weak, because it can't be defended by a pawn, and the square in front of it is attacked by pawns as many times as it is defended, which will leave black with an isolated (and passed) pawn on b5.

I like the general idea of mobility rather than trying to create multiple categories of pawns that are weak and then name each one something different (and often something confusing).

Right, it means that it can not advance even with the support of pieces.
Let is leave dynamic features to search, it is not clear if a6 and c6 will ever be removed, for the time being b4 can not budge.

[d][d]3r1n2/2p3k1/1p1q1p2/p1pPp1p1/P1P1P1P1/4BP2/5K1R/7Q b - - 0 32

Did you see that diagram? It is from the Komodo-Critter game in TCEC. White is winning this with a pawn less. And one of the main reasons for this is that b6 is backward-fated, or whatever you would like to call it. 3 white pawns on the queen side hold 4 black. So that b6 is due an enormous penalty, almost a full pawn. Do you want that engines do not consider features worth a full pawn? Why should we neglect such features and instead go after phantoms?

I bet Crafty does not consider b6 as weak. And does not penalise it with almost a full pawn. I also bet Crafty does not see white is winning here with a pawn less, just as Critter. Am I right?

Crafty does not consider b6 as weak. Neither do I. I just consider it immobile. I don't have to dedicate resources to defend it, but I can't move it which limits my options to the kingside. I'm not convinced this pawn is the reason black lost, however.

In any case, you have to come up with a reasonable explanation how 6 white pawns are able to contain 7 black ones. None of black pawns is actually able to move.

I think, although funnily nicknamed in this thread, I think this pawn feature is the last eval feature still deserving a very large penalty that has not been applied as a rule in chess engines. Having a pawn feature deserving 50cps penalty in mg/80-90cps in eg, and on many occasions even more, and that on the standard scale, that is still not a part of modern top engines is indeed astounding. Worth more than rook on open file, considerably more than rook on 7th and isolated, passed pawns, etc., but not featuring in any way in engine eval. Well, that is really a paradox.

I think this pawn has a right of existence on its own, no matter how you call it.

[Lyudmil Tsvetkov]

My take on this is that you are somewhat mixing terms.

A pawn that is defended by a pawn is not considered backward in any book I have (g6 in your first example as it is defended by h7). The fact that it can not safely advance is not so important when it is perfectly safe where it is. If you want to talk about it being immobile, that is a different issue.

the "unopposed" requirement stems from the rook, in the books I have read over the years. A pawn that is backward (can not be safely defended by a friendly pawn) is generally considered to be worse if it is (a) on a half-open file (no enemy pawn in front of it AND (b) the enemy has rooks to attack it.

IMHO, a pawn that is backward and unopposed is not any weaker or stronger than a backward pawn that is opposed, if there are no enemy rooks to contend with. Any backward pawn is weak, and inviting to enemy kings in the endgame. backward pawns on half-open files make inviting targets even in the middle game if you have rooks to attack on the file.

I've seen games decided by two types of backward pawns, repeatedly.

The first is the classic backward pawn on a half-open file, which is just as bad as an isolated pawn on a half-open file, when the enemy has rooks. Pile up on the pawn, tying up the opponent by making him defend, and either win the pawn outright, or switch to some other idea once he is tied down and not coordinated very well.

The second is a pawn that is backward, but the file is not half-open. If you have just one, you might manage to defend it with your king, depending on the position. If you have two, the enemy king can make you commit to saving one and then to eat the other pawn chain. Doesn't matter whether the file is half open or not.

[d]6k1/5p2/4p1p1/3p2P1/1p1P4/p7/P7/6K1 w - - 0 1

I tend to concur with Bob's definition. In the diagram above the e6 pawn is still defended, and thus not vulnerable to direct attack, so that's an argument against calling it backward.

However, since it cannot safely advance, there is an element of backwardness about it, and the square in front of it (e5) can serve as a good outpost for White pieces. These considerations should justify some sort of a penalty for such a pawn, but a lesser one that a fully backward [undefendable by a fellow pawn] pawn on an open file.

The Black b-pawn, even though backward, could be pushed in some cases creating a dangerous advanced passed pawn. It's not as clear whether a backward pawn that sits in the opponent's half of the board deserves any significant penalty.

CL

Crafty calls that a3 pawn a "hidden passed pawn". GM Dzhindi used to call this a protected passed pawn himself, because a3 is certainly passed here when you think about it and it is defended by b4. after b3 black ends up with a passed pawn for certain.

You really do not read this thread.

When you add pieces, there is no danger of a passed pawn.

[d][d]6k1/5p2/4p1p1/n2p2P1/1p1P4/p7/P7/2N3K1 w - - 0 1
One knight each side added; where is the danger of the hidden passed a3 pawn?
a3 is not an immediate danger, but b4 is certainly backward, like it or not, a2 stops both a3 and b4 pawns by blocking a3 and attacking the stop square of b4.

Very simple. White has to avoid any move that lets me force the exchange of knights. Because allowing such is an instant loss. Unless white chooses to brink his king over to the a-file area, which ties it down and restricts white's ability to defend elsewhere.

Adding the knights does NOT eliminate the advantage of a3-b4, it just makes it harder to use that advantage. In this case, as black I would try to get the knight to c3 and take on a2. White knight will have great trouble with the a/b pawns after that

We are talking here about a static feature, and the static feature is that one white pawn, a2, holds 2 black pawns, a3 and b4, simultaneously.

Sorry, but you can't have it BOTH ways.

remove the knights.

Is b4 weak, or is a3 overwhelmingly strong? a3 is very strong so long as black can get enough pieces to attack b3 to make pushing b3 safe. That's not "static" at all, that is dynamic. So leave the pieces out of it. b4 is weak, as it can't safely advance, but there is a stinger in that b4 CAN advance because a3 will promote before white's pawn that takes on b3. In this position, the white pawn on a3 is not really holding back anything whatsoever. And, in fact, my program considers this a significant pawn advantage, not a weakness. Why? Ask GM Dzhindi. We had that discussion almost 20 years ago and he convinced me it is correct. Yes, Crafty still gets a penalty for b4 being weak, but it gets a significantly larger bonus for a3 being so strong.

Professor Hyatt, you taught me that it is better to leave dynamic issues to the search, so let us concentrate on the static features then.

Knights present are a natural feature, as in 95% of all situations there are pieces on the board instead of only pawns. The big limitation of the hidden passer concept is that it is constructed for and could be applied only in pawn endgames. Even one piece added already makes the rule non-functional. So this is an ad-hoc rule, bad at that, as it is valid in only 5% of cases.

Backward pawns, on the other hand, including giving penalty to b4, are valid in all situations where there are pieces apart from the pawns, i.e. in 95% of cases. This already seems to me like a good generalised rule, and not an ad-hoc rule.

Giving a3 a very nice bonus for its advanced status, either in psqt or otherwise, is another matter.

Not so fast. The "hidden passer" idea works in any reduced material endgame. If you can prevent b3 with a piece, fine. But that piece is tied down. By a lowly pawn. And if the piece moves away, particularly knights, the hidden passer can become a real passer that wins...

What we try to capture in pawn structure evaluation is an answer to the question "if all pieces are removed, who stands better?" Because the side that stands better is going to try to trade pieces to reach that situation, and the side that stands worse has to struggle to avoid the trades. So the pawn structure becomes an integral part of the game. In other positions, the opposite happens. A bishop vs knight endgame, where the pawn structure restricts the bishop but not the knight. So the side with the bishop wants to trade, the side with the knight would rather trade other pieces but not the knights.

Trying to come up with a general-purpose pawn structure evaluation is actually pretty difficult, because most of the terms have to be general-purpose, while humans don't think that way, we look at specific features. The more general-purpose the eval becomes, the more errors it will make in specific circumstances. Of course if you omit some of those general-purpose rules, it makes even more errors because it has no clue at all about that specific feature.

The evaluation of a chess engine will likely NEVER come even close to what a human does. But the computer is so fast, and searches so deeply, that it can get by with an inferior evaluation because the human has a very inferior tactical analyzer.

[d]6k1/7p/8/n7/1p6/p7/P5PP/2N3K1 w - - 0 1
Well, here is a reduced material endgame where it does not work. White does not have a hidden passer, just a humble candidate. But who is better? Where is the power of the hidden passer? The backward b4 pawn however is a valid deficiency even in this extremely reduced, but not pieceless endgame. It leaves white with an efficient pawn majority on the king side.

This is anything but a correct proof position:
i) You have granted yourself the right to move. Due to that alone white will conquer the fourth rank. This advantage alone can be decisive in knight endgames.
ii)You have misplaced the black knight to the edge.
iii) You put your knight in contact with b3-a2, preventing any breakthrough.

So if you are allowed to place your knight, i am allowed to place my knight. I just move my knight to b5:
[d]6k1/7p/8/1n6/1p6/p7/P5PP/2N3K1 w - - 0 1
1. Kf2 Nc3 2.Ke3 Nxa2! and white fights for the draw.
Due to advantage (i) this is still possible (according to stockfish)

What happens if i grant me the first move?

According to to stockfish, this gives me an decisive advantage (-246 cp)

This is already of tactical nature.

The first position with the knights on c1 and a5, both controlling the b3 square, is much more to the point and unbiased, as our aim was to see if indeed a3 represents a real danger in positions with pieces on the board, or only in pure pawn endgames. To decide that, the control over the b3 square, the ability to play b4-b3 is relevant, but it seems, at least here, b4-b3 is no real danger.

[Ralph Stoesser]

My take on this is that you are somewhat mixing terms.

A pawn that is defended by a pawn is not considered backward in any book I have (g6 in your first example as it is defended by h7). The fact that it can not safely advance is not so important when it is perfectly safe where it is. If you want to talk about it being immobile, that is a different issue.

the "unopposed" requirement stems from the rook, in the books I have read over the years. A pawn that is backward (can not be safely defended by a friendly pawn) is generally considered to be worse if it is (a) on a half-open file (no enemy pawn in front of it AND (b) the enemy has rooks to attack it.

IMHO, a pawn that is backward and unopposed is not any weaker or stronger than a backward pawn that is opposed, if there are no enemy rooks to contend with. Any backward pawn is weak, and inviting to enemy kings in the endgame. backward pawns on half-open files make inviting targets even in the middle game if you have rooks to attack on the file.

I've seen games decided by two types of backward pawns, repeatedly.

The first is the classic backward pawn on a half-open file, which is just as bad as an isolated pawn on a half-open file, when the enemy has rooks. Pile up on the pawn, tying up the opponent by making him defend, and either win the pawn outright, or switch to some other idea once he is tied down and not coordinated very well.

The second is a pawn that is backward, but the file is not half-open. If you have just one, you might manage to defend it with your king, depending on the position. If you have two, the enemy king can make you commit to saving one and then to eat the other pawn chain. Doesn't matter whether the file is half open or not.

[d]6k1/5p2/4p1p1/3p2P1/1p1P4/p7/P7/6K1 w - - 0 1

I tend to concur with Bob's definition. In the diagram above the e6 pawn is still defended, and thus not vulnerable to direct attack, so that's an argument against calling it backward.

However, since it cannot safely advance, there is an element of backwardness about it, and the square in front of it (e5) can serve as a good outpost for White pieces. These considerations should justify some sort of a penalty for such a pawn, but a lesser one that a fully backward [undefendable by a fellow pawn] pawn on an open file.

The Black b-pawn, even though backward, could be pushed in some cases creating a dangerous advanced passed pawn. It's not as clear whether a backward pawn that sits in the opponent's half of the board deserves any significant penalty.

CL

Crafty calls that a3 pawn a "hidden passed pawn". GM Dzhindi used to call this a protected passed pawn himself, because a3 is certainly passed here when you think about it and it is defended by b4. after b3 black ends up with a passed pawn for certain.

You really do not read this thread.

When you add pieces, there is no danger of a passed pawn.

[d][d]6k1/5p2/4p1p1/n2p2P1/1p1P4/p7/P7/2N3K1 w - - 0 1
One knight each side added; where is the danger of the hidden passed a3 pawn?
a3 is not an immediate danger, but b4 is certainly backward, like it or not, a2 stops both a3 and b4 pawns by blocking a3 and attacking the stop square of b4.

Very simple. White has to avoid any move that lets me force the exchange of knights. Because allowing such is an instant loss. Unless white chooses to brink his king over to the a-file area, which ties it down and restricts white's ability to defend elsewhere.

Adding the knights does NOT eliminate the advantage of a3-b4, it just makes it harder to use that advantage. In this case, as black I would try to get the knight to c3 and take on a2. White knight will have great trouble with the a/b pawns after that

We are talking here about a static feature, and the static feature is that one white pawn, a2, holds 2 black pawns, a3 and b4, simultaneously.

Sorry, but you can't have it BOTH ways.

remove the knights.

Is b4 weak, or is a3 overwhelmingly strong? a3 is very strong so long as black can get enough pieces to attack b3 to make pushing b3 safe. That's not "static" at all, that is dynamic. So leave the pieces out of it. b4 is weak, as it can't safely advance, but there is a stinger in that b4 CAN advance because a3 will promote before white's pawn that takes on b3. In this position, the white pawn on a3 is not really holding back anything whatsoever. And, in fact, my program considers this a significant pawn advantage, not a weakness. Why? Ask GM Dzhindi. We had that discussion almost 20 years ago and he convinced me it is correct. Yes, Crafty still gets a penalty for b4 being weak, but it gets a significantly larger bonus for a3 being so strong.

Professor Hyatt, you taught me that it is better to leave dynamic issues to the search, so let us concentrate on the static features then.

Knights present are a natural feature, as in 95% of all situations there are pieces on the board instead of only pawns. The big limitation of the hidden passer concept is that it is constructed for and could be applied only in pawn endgames. Even one piece added already makes the rule non-functional. So this is an ad-hoc rule, bad at that, as it is valid in only 5% of cases.

Backward pawns, on the other hand, including giving penalty to b4, are valid in all situations where there are pieces apart from the pawns, i.e. in 95% of cases. This already seems to me like a good generalised rule, and not an ad-hoc rule.

Giving a3 a very nice bonus for its advanced status, either in psqt or otherwise, is another matter.

Not so fast. The "hidden passer" idea works in any reduced material endgame. If you can prevent b3 with a piece, fine. But that piece is tied down. By a lowly pawn. And if the piece moves away, particularly knights, the hidden passer can become a real passer that wins...

What we try to capture in pawn structure evaluation is an answer to the question "if all pieces are removed, who stands better?" Because the side that stands better is going to try to trade pieces to reach that situation, and the side that stands worse has to struggle to avoid the trades. So the pawn structure becomes an integral part of the game. In other positions, the opposite happens. A bishop vs knight endgame, where the pawn structure restricts the bishop but not the knight. So the side with the bishop wants to trade, the side with the knight would rather trade other pieces but not the knights.

Trying to come up with a general-purpose pawn structure evaluation is actually pretty difficult, because most of the terms have to be general-purpose, while humans don't think that way, we look at specific features. The more general-purpose the eval becomes, the more errors it will make in specific circumstances. Of course if you omit some of those general-purpose rules, it makes even more errors because it has no clue at all about that specific feature.

The evaluation of a chess engine will likely NEVER come even close to what a human does. But the computer is so fast, and searches so deeply, that it can get by with an inferior evaluation because the human has a very inferior tactical analyzer.

[d]6k1/7p/8/n7/1p6/p7/P5PP/2N3K1 w - - 0 1
Well, here is a reduced material endgame where it does not work. White does not have a hidden passer, just a humble candidate. But who is better? Where is the power of the hidden passer? The backward b4 pawn however is a valid deficiency even in this extremely reduced, but not pieceless endgame. It leaves white with an efficient pawn majority on the king side.

This is anything but a correct proof position:
i) You have granted yourself the right to move. Due to that alone white will conquer the fourth rank. This advantage alone can be decisive in knight endgames.
ii)You have misplaced the black knight to the edge.
iii) You put your knight in contact with b3-a2, preventing any breakthrough.

So if you are allowed to place your knight, i am allowed to place my knight. I just move my knight to b5:
[d]6k1/7p/8/1n6/1p6/p7/P5PP/2N3K1 w - - 0 1
1. Kf2 Nc3 2.Ke3 Nxa2! and white fights for the draw.
Due to advantage (i) this is still possible (according to stockfish)

What happens if i grant me the first move?

According to to stockfish, this gives me an decisive advantage (-246 cp)

This is already of tactical nature.

The first position with the knights on c1 and a5, both controlling the b3 square, is much more to the point and unbiased, as our aim was to see if indeed a3 represents a real danger in positions with pieces on the board, or only in pure pawn endgames. To decide that, the control over the b3 square, the ability to play b4-b3 is relevant, but it seems, at least here, b4-b3 is no real danger.

No danger only because you placed the white knight on c1. If you place the white knight on the more unbiased square d2, white is lost due to the breakthrough b4-b3.

It's not correct to only evaluate static features of a position. A good evaluation function should also estimate dynamic aspects of the position. Passed pawn evaluation is a good example. It is worth many ELO points.

[Lyudmil Tsvetkov]

My take on this is that you are somewhat mixing terms.

A pawn that is defended by a pawn is not considered backward in any book I have (g6 in your first example as it is defended by h7). The fact that it can not safely advance is not so important when it is perfectly safe where it is. If you want to talk about it being immobile, that is a different issue.

the "unopposed" requirement stems from the rook, in the books I have read over the years. A pawn that is backward (can not be safely defended by a friendly pawn) is generally considered to be worse if it is (a) on a half-open file (no enemy pawn in front of it AND (b) the enemy has rooks to attack it.

IMHO, a pawn that is backward and unopposed is not any weaker or stronger than a backward pawn that is opposed, if there are no enemy rooks to contend with. Any backward pawn is weak, and inviting to enemy kings in the endgame. backward pawns on half-open files make inviting targets even in the middle game if you have rooks to attack on the file.

I've seen games decided by two types of backward pawns, repeatedly.

The first is the classic backward pawn on a half-open file, which is just as bad as an isolated pawn on a half-open file, when the enemy has rooks. Pile up on the pawn, tying up the opponent by making him defend, and either win the pawn outright, or switch to some other idea once he is tied down and not coordinated very well.

The second is a pawn that is backward, but the file is not half-open. If you have just one, you might manage to defend it with your king, depending on the position. If you have two, the enemy king can make you commit to saving one and then to eat the other pawn chain. Doesn't matter whether the file is half open or not.

[d]6k1/5p2/4p1p1/3p2P1/1p1P4/p7/P7/6K1 w - - 0 1

I tend to concur with Bob's definition. In the diagram above the e6 pawn is still defended, and thus not vulnerable to direct attack, so that's an argument against calling it backward.

However, since it cannot safely advance, there is an element of backwardness about it, and the square in front of it (e5) can serve as a good outpost for White pieces. These considerations should justify some sort of a penalty for such a pawn, but a lesser one that a fully backward [undefendable by a fellow pawn] pawn on an open file.

The Black b-pawn, even though backward, could be pushed in some cases creating a dangerous advanced passed pawn. It's not as clear whether a backward pawn that sits in the opponent's half of the board deserves any significant penalty.

CL

Crafty calls that a3 pawn a "hidden passed pawn". GM Dzhindi used to call this a protected passed pawn himself, because a3 is certainly passed here when you think about it and it is defended by b4. after b3 black ends up with a passed pawn for certain.

You really do not read this thread.

When you add pieces, there is no danger of a passed pawn.

[d][d]6k1/5p2/4p1p1/n2p2P1/1p1P4/p7/P7/2N3K1 w - - 0 1
One knight each side added; where is the danger of the hidden passed a3 pawn?
a3 is not an immediate danger, but b4 is certainly backward, like it or not, a2 stops both a3 and b4 pawns by blocking a3 and attacking the stop square of b4.

Very simple. White has to avoid any move that lets me force the exchange of knights. Because allowing such is an instant loss. Unless white chooses to brink his king over to the a-file area, which ties it down and restricts white's ability to defend elsewhere.

Adding the knights does NOT eliminate the advantage of a3-b4, it just makes it harder to use that advantage. In this case, as black I would try to get the knight to c3 and take on a2. White knight will have great trouble with the a/b pawns after that

We are talking here about a static feature, and the static feature is that one white pawn, a2, holds 2 black pawns, a3 and b4, simultaneously.

Sorry, but you can't have it BOTH ways.

remove the knights.

Is b4 weak, or is a3 overwhelmingly strong? a3 is very strong so long as black can get enough pieces to attack b3 to make pushing b3 safe. That's not "static" at all, that is dynamic. So leave the pieces out of it. b4 is weak, as it can't safely advance, but there is a stinger in that b4 CAN advance because a3 will promote before white's pawn that takes on b3. In this position, the white pawn on a3 is not really holding back anything whatsoever. And, in fact, my program considers this a significant pawn advantage, not a weakness. Why? Ask GM Dzhindi. We had that discussion almost 20 years ago and he convinced me it is correct. Yes, Crafty still gets a penalty for b4 being weak, but it gets a significantly larger bonus for a3 being so strong.

Professor Hyatt, you taught me that it is better to leave dynamic issues to the search, so let us concentrate on the static features then.

Knights present are a natural feature, as in 95% of all situations there are pieces on the board instead of only pawns. The big limitation of the hidden passer concept is that it is constructed for and could be applied only in pawn endgames. Even one piece added already makes the rule non-functional. So this is an ad-hoc rule, bad at that, as it is valid in only 5% of cases.

Backward pawns, on the other hand, including giving penalty to b4, are valid in all situations where there are pieces apart from the pawns, i.e. in 95% of cases. This already seems to me like a good generalised rule, and not an ad-hoc rule.

Giving a3 a very nice bonus for its advanced status, either in psqt or otherwise, is another matter.

Not so fast. The "hidden passer" idea works in any reduced material endgame. If you can prevent b3 with a piece, fine. But that piece is tied down. By a lowly pawn. And if the piece moves away, particularly knights, the hidden passer can become a real passer that wins...

What we try to capture in pawn structure evaluation is an answer to the question "if all pieces are removed, who stands better?" Because the side that stands better is going to try to trade pieces to reach that situation, and the side that stands worse has to struggle to avoid the trades. So the pawn structure becomes an integral part of the game. In other positions, the opposite happens. A bishop vs knight endgame, where the pawn structure restricts the bishop but not the knight. So the side with the bishop wants to trade, the side with the knight would rather trade other pieces but not the knights.

Trying to come up with a general-purpose pawn structure evaluation is actually pretty difficult, because most of the terms have to be general-purpose, while humans don't think that way, we look at specific features. The more general-purpose the eval becomes, the more errors it will make in specific circumstances. Of course if you omit some of those general-purpose rules, it makes even more errors because it has no clue at all about that specific feature.

The evaluation of a chess engine will likely NEVER come even close to what a human does. But the computer is so fast, and searches so deeply, that it can get by with an inferior evaluation because the human has a very inferior tactical analyzer.

[d]6k1/7p/8/n7/1p6/p7/P5PP/2N3K1 w - - 0 1
Well, here is a reduced material endgame where it does not work. White does not have a hidden passer, just a humble candidate. But who is better? Where is the power of the hidden passer? The backward b4 pawn however is a valid deficiency even in this extremely reduced, but not pieceless endgame. It leaves white with an efficient pawn majority on the king side.

This is anything but a correct proof position:
i) You have granted yourself the right to move. Due to that alone white will conquer the fourth rank. This advantage alone can be decisive in knight endgames.
ii)You have misplaced the black knight to the edge.
iii) You put your knight in contact with b3-a2, preventing any breakthrough.

So if you are allowed to place your knight, i am allowed to place my knight. I just move my knight to b5:
[d]6k1/7p/8/1n6/1p6/p7/P5PP/2N3K1 w - - 0 1
1. Kf2 Nc3 2.Ke3 Nxa2! and white fights for the draw.
Due to advantage (i) this is still possible (according to stockfish)

What happens if i grant me the first move?

According to to stockfish, this gives me an decisive advantage (-246 cp)

This is already of tactical nature.

The first position with the knights on c1 and a5, both controlling the b3 square, is much more to the point and unbiased, as our aim was to see if indeed a3 represents a real danger in positions with pieces on the board, or only in pure pawn endgames. To decide that, the control over the b3 square, the ability to play b4-b3 is relevant, but it seems, at least here, b4-b3 is no real danger.

No danger only because you placed the white knight on c1. If you place the white knight on the more unbiased square d2, white is lost due to the breakthrough b4-b3.

It's not correct to only evaluate static features of a position. A good evaluation function should also estimate dynamic aspects of the position. Passed pawn evaluation is a good example. It is worth many ELO points.

Hi Ralph.
Are you still around?

The framework is completely empty, it needs someone to revive it...

By static I meant excluding search, just as Mr. Hyatt understands. Of course, elements like control of defended/unsafe squares in front of passers or piece psqt would be included into static eval.

What concerns your example to put the knight on d2 however, I think this already gives too much edge to black, that already has the enormous edge of 2 very advanced pawns on the 5th and 6th ranks. The point was, that even with this enormous black psqt edge, white is still able to hold the position and perform well in some circumstances, precisely because of the backward feature. Many engines will not think a pawn could be backward on the 4th rank as b4, but it could. The difference is in the values, as b4 backward will get much much less penalty than b7 black backward.

I think, if you have to change backward definition and apply through ranks, you should also very carefully tune the rank penalties.

Happy to see you again around Ralph.