Maybe I have mentioned this in the past, but now I will be as specific as possible.
Being specific in computer chess is a good thing. People who know this will understand me, others will think I am nuts again.
Anyway, backward pawns, that are part of the king shelter, are due bigger penalty in general - absolutely no doubt about that.
However, there are only 2 backward pawns, part of the king shelter, that are due bigger penalty: f7 and g7 pawns. (respectively c7 and b7 backward pawns with the black king castled long)
So, those are pawns on the 2nd rank. But, please note, h7/a7 backward pawn part of the shelter is excluded. Also excluded are backward pawns on the 3rd rank, like h6,g6 and f6, or a6,b6 and c6 with the black king castled short and long respectively.
For those backward pawns part of the shelter it simply does not make any sense to score an overpenalty, they might be just like any other backward pawn.
But, f7 and g7 pawns do deserve a sufficient overpenalty, as they make the king shelter very inflexible.
[d]6k1/5p2/4p3/4P3/8/8/8/6K1 w - - 0 1
f7 backward pawn part of the king shelter, due bigger penalty
[d]1k4k1/2p3p1/1p3p2/1P3P2/8/8/8/1K4K1 w - - 0 1
g7 backward pawn part of the king shelter; c7 backward pawn part of the shelter
The overpenalty might be increasing by 1/2 or even double the respective backward penalty.
Is not it just possible to tune to the whole and not special-case?
Well, not, as those 2 pawns are penalised as a whole, this will include also same backward pawns that are not part of the shelter, and for engines that do not make distinction for backward pawns in terms of ranks the mess will be complete, as you want to overpenalise just 2 out of the existing at least 6 or 8 backward pawns for the specific files. So that special-casing is obligatory.
How important is this?
Extremely important. You can gain huge amount of elo on that, no doubt about it, as this is a king safety related feature, and for king safety you need good eval, as search is very deep there.
Problem?
Even if you give this penalty, engines will still not be able to recognise the benefit of it quickly, as it requires deep search. SF and Komodo do not usually see the benefit of such penalty, even at very long TC, so how weaker engines will make use of it I am completely at a loss.
One thing is certain - this is an extremely useful rule, related to king safety.
King safety has a very wide range of features engines do not understand, and that is just one of those.
Assorted tweaks
Moderator: Ras
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: Backward f7 and g7 pawns part of the shelter
Of course, this would be true only for the mg, or at least where king safety applies.
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: Main doubled rule - symmetrical doubled opposed
This is the famous rule 1 pawn stops 2 enemy pawns, or 2 pawns stop 3 enemy pawns, or 3 pawns stop 4 enemy pawns.
With other doubled pawns, this is not true.
With other doubled pawns, this is not true.
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Single shelter pawn on the 3rd rank
This is also related to king safety.
No pawn shelter - no pawns on the 2nd and 3rd ranks within the shelter - is due some 20-30cps penalty, but, although a single own pawn on the 2nd or 3rd rank will generally suffice to keep the king safe, a small penalty is also due in the case when there is a single own pawn in the shelter, but it is on the 3rd, and not the 2nd rank.
Penalty might be some 5-10cps additionally, which might prove important in a range of cases.
[d]6k1/6p1/8/8/8/6P1/8/6K1 w - - 0 1
white has a single own pawn within the shelter, on the 3rd rank, and therefore some 5-10cps penalty is due; although there is still a pawn on the 2nd or 3rd rank close to the king, a single pawn on the 3rd rank is the worst case scenario after no pawn shelter
In distinction, black has also a single pawn within the shelter on the 2nd and 3rd ranks, but that pawn is on the 2nd rank, so no additional penalty here - the king is safe with a single pawn on the 2nd rank
[d]6k1/8/5p2/6p1/8/5PP1/8/6K1 w - - 0 1
here already the black king is due the overpenalty
white has 2 pawns on the 3rd rank within the shelter on ranks 2 and 3, so no penalty here
black has just a single pawn on the 3rd rank within the shelter, f6, so an overpenalty is due here
And indeed, look at thousands of positions, you will see that this is true - no pawn shelter is bad for king safety, having a single shelter pawn on the 3rd rank is also not good, but much more acceptable, still this is somewhat dangerous and due a penalty, while having a single shelter pawn on the 2nd rank is just fine.
No pawn shelter - no pawns on the 2nd and 3rd ranks within the shelter - is due some 20-30cps penalty, but, although a single own pawn on the 2nd or 3rd rank will generally suffice to keep the king safe, a small penalty is also due in the case when there is a single own pawn in the shelter, but it is on the 3rd, and not the 2nd rank.
Penalty might be some 5-10cps additionally, which might prove important in a range of cases.
[d]6k1/6p1/8/8/8/6P1/8/6K1 w - - 0 1
white has a single own pawn within the shelter, on the 3rd rank, and therefore some 5-10cps penalty is due; although there is still a pawn on the 2nd or 3rd rank close to the king, a single pawn on the 3rd rank is the worst case scenario after no pawn shelter
In distinction, black has also a single pawn within the shelter on the 2nd and 3rd ranks, but that pawn is on the 2nd rank, so no additional penalty here - the king is safe with a single pawn on the 2nd rank
[d]6k1/8/5p2/6p1/8/5PP1/8/6K1 w - - 0 1
here already the black king is due the overpenalty
white has 2 pawns on the 3rd rank within the shelter on ranks 2 and 3, so no penalty here
black has just a single pawn on the 3rd rank within the shelter, f6, so an overpenalty is due here
And indeed, look at thousands of positions, you will see that this is true - no pawn shelter is bad for king safety, having a single shelter pawn on the 3rd rank is also not good, but much more acceptable, still this is somewhat dangerous and due a penalty, while having a single shelter pawn on the 2nd rank is just fine.
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Targeting the center
Well, attacking the 4 centralmost squares of the board - d4,e4,d5 and e5 - is of course important, but one should know in what particular cases.
Attacking the empty centralmost squares will be tentamount to psqt.
Attacking enemy pieces on the 4 centralmost squares is irrelevant, as purely tactical, so no rules could be formulated here.
Capturing on the 4 centralmost squares is even more devoid of any logic, as this is even more pronouncedly tactically related. (btw., this is the reason why capturing towards the center does not work - it is difficult or impossible to formulate strict rules for tactics, tactics likes to have it its own way)
However, there is one rule of targeting the center that works - and that will be attacking enemy pawns on any of the 4 centralmost squares. This is already a good rule, as it has long-term effect, pawns are slow in changing positions, so sooner or later such targeted central pawns might pay off in some way, after otherwise incalculable very deep complications, or even in the endgame.
Any own piece attacking an enemy pawn on e4,d4,e5 or d5 is due some very small bonus, maybe some 2-3cps.
[d]6k1/8/1q6/3pR3/3Pp3/8/6B1/6K1 w - - 0 1
Bg2 is due bonus for attacking e4, Re5 is due bonus for attacking e4 and d5, while Qb6 bonus for attacking d4.
Thus, white is better here in terms of central attacks.
[d]6k1/8/2n5/4p3/3pP3/3P1N2/8/6K1 w - - 0 1
Nf3 is due bonus for attacking both e5 and d4; black is due no bonus here
Of course, if you already score attacking pawns, this will be just an overbonus for attacking the center.
Attacking the empty centralmost squares will be tentamount to psqt.
Attacking enemy pieces on the 4 centralmost squares is irrelevant, as purely tactical, so no rules could be formulated here.
Capturing on the 4 centralmost squares is even more devoid of any logic, as this is even more pronouncedly tactically related. (btw., this is the reason why capturing towards the center does not work - it is difficult or impossible to formulate strict rules for tactics, tactics likes to have it its own way)
However, there is one rule of targeting the center that works - and that will be attacking enemy pawns on any of the 4 centralmost squares. This is already a good rule, as it has long-term effect, pawns are slow in changing positions, so sooner or later such targeted central pawns might pay off in some way, after otherwise incalculable very deep complications, or even in the endgame.
Any own piece attacking an enemy pawn on e4,d4,e5 or d5 is due some very small bonus, maybe some 2-3cps.
[d]6k1/8/1q6/3pR3/3Pp3/8/6B1/6K1 w - - 0 1
Bg2 is due bonus for attacking e4, Re5 is due bonus for attacking e4 and d5, while Qb6 bonus for attacking d4.
Thus, white is better here in terms of central attacks.
[d]6k1/8/2n5/4p3/3pP3/3P1N2/8/6K1 w - - 0 1
Nf3 is due bonus for attacking both e5 and d4; black is due no bonus here
Of course, if you already score attacking pawns, this will be just an overbonus for attacking the center.
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Attackable pawns of the shelter
I hope Roger is already preparing my new thread, as this starts getting too short.
What is an attackable pawn of the king shelter?
Well, any pawn of the shelter, which could be attacked by a lever created by an enemy pawn when it moves forward.
Does it make sense to score all attackable pawns of the shelter?
Well, of course not. Only pawns that it makes sense to consider additionally from stroming pawns and levers are attackable shelter pawns on the 3rd rank. Attackable shelter pawns on the 4th rank are not important, while attackable shelter pawns on the 2nd rank are very easily seen by search, as at this point the attack will be already rolling.
However, attackable shelter pawns on the 3rd rank are not easily seen by search, as the respective lines are deep, and besides scoring such attackable pawns might help you to start an attack early, or, for the defending side, not prematurely push its shelter pawns from the 2nd to the 3rd rank.
[d]6k1/6p1/7p/8/6P1/8/8/6K1 w - - 0 1
h6 is an attackable pawn of the shelter on the 3rd rank; so, in order to trigger that condition, you need to have a shelter pawn on the 3rd rank, an enemy storming pawn on its 4th rank on adjacent file, plus necessarily the square in front of the enemy storming pawn being empty.
above, h6 is attackable shelter pawn, as enemy has g4 storming pawn on adjacent file, plus the square in front of g4 is empty; without the empty square condition it does not make sense to score attackable pawns, as otherwise this square might be blocked by an own pawn for example
[d]6k1/7p/6p1/8/5PP1/8/8/6K1 w - - 0 1
attackable g6 shelter pawn - it is on its 3rd rank, enemy has storming pawn on its 4th rank on adjacent file, f4, plus the square in front of f4 is empty
I think such a rule might be quite useful for a range of engines, of course, you should be lucky with the size of the penalty; I think some 8-12cps penalty might be a good indication where to start.
Of course, this will be considered only for the mg, as this is strictly king safety related.
As you see, there is no end to king safety features, unfortunately engines consider only a very small portion of those.
What is an attackable pawn of the king shelter?
Well, any pawn of the shelter, which could be attacked by a lever created by an enemy pawn when it moves forward.
Does it make sense to score all attackable pawns of the shelter?
Well, of course not. Only pawns that it makes sense to consider additionally from stroming pawns and levers are attackable shelter pawns on the 3rd rank. Attackable shelter pawns on the 4th rank are not important, while attackable shelter pawns on the 2nd rank are very easily seen by search, as at this point the attack will be already rolling.
However, attackable shelter pawns on the 3rd rank are not easily seen by search, as the respective lines are deep, and besides scoring such attackable pawns might help you to start an attack early, or, for the defending side, not prematurely push its shelter pawns from the 2nd to the 3rd rank.
[d]6k1/6p1/7p/8/6P1/8/8/6K1 w - - 0 1
h6 is an attackable pawn of the shelter on the 3rd rank; so, in order to trigger that condition, you need to have a shelter pawn on the 3rd rank, an enemy storming pawn on its 4th rank on adjacent file, plus necessarily the square in front of the enemy storming pawn being empty.
above, h6 is attackable shelter pawn, as enemy has g4 storming pawn on adjacent file, plus the square in front of g4 is empty; without the empty square condition it does not make sense to score attackable pawns, as otherwise this square might be blocked by an own pawn for example
[d]6k1/7p/6p1/8/5PP1/8/8/6K1 w - - 0 1
attackable g6 shelter pawn - it is on its 3rd rank, enemy has storming pawn on its 4th rank on adjacent file, f4, plus the square in front of f4 is empty
I think such a rule might be quite useful for a range of engines, of course, you should be lucky with the size of the penalty; I think some 8-12cps penalty might be a good indication where to start.
Of course, this will be considered only for the mg, as this is strictly king safety related.
As you see, there is no end to king safety features, unfortunately engines consider only a very small portion of those.
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Space advantage on both sides of the board
Well, having space advantage on both sides of the board - the queen and the king side - is generally a good thing, as this will mean you have the initiative, are active on many fronts, which obviously makes the opponent's defence more difficult.
Space advantage is generally determined by the presence of advanced own pawns on their 5th or 6th ranks. However, when scoring space advantage on both sides of the board, you need not so much to assess the different own pawns on advanced ranks on the queen and king side, which sometimes might already lead you to score such pawns when the game is already decided, but rather capture the moment when one player starts getting the initiative/space advantage on both sides of the board. This will be determined of course by the presence of own pawns on the 5th rank.
There are different configurations of own pawns on their 5th rank on both the queen and king side, however, most of them happen extremely rarely and are not that forceful. There is one combination of own pawns on the 5th rank that is relatively frequent and deserves to be scored.
Those will be pawns on their 5th rank, that are 2 files apart, for example c5 and f5. For all such pawns, some bonus is due.
So, whenever you have pawns on a5 and d5, or b5 and e5, or c5 and f5, or d5 and g5, or e5 and h5 simultaneously, you give some 10cps overbonus.
Space advantage on both sides of the board should be scored only in the mg, as space is much more important in the mg, and besides in the eg this would already clash with passer code.
All of the above tandems of pawns control at least one square on the king side and at least one square on the queen side, meaning their influence is widespread.
[d]6k1/1pp5/p2p4/P2P4/8/8/8/6K1 w - - 0 1
a5 and d5 pawns gaining space advantage on both sides of the board
[d]6k1/8/2p2p2/2PppP2/8/8/8/6K1 w - - 0 1
c5 and f5 pawns gaining space advantage on both sides of the board
I still am not aware of an engine that successfully evaluates space advantage in terms of pawns in all its incarnations.
Space advantage is generally determined by the presence of advanced own pawns on their 5th or 6th ranks. However, when scoring space advantage on both sides of the board, you need not so much to assess the different own pawns on advanced ranks on the queen and king side, which sometimes might already lead you to score such pawns when the game is already decided, but rather capture the moment when one player starts getting the initiative/space advantage on both sides of the board. This will be determined of course by the presence of own pawns on the 5th rank.
There are different configurations of own pawns on their 5th rank on both the queen and king side, however, most of them happen extremely rarely and are not that forceful. There is one combination of own pawns on the 5th rank that is relatively frequent and deserves to be scored.
Those will be pawns on their 5th rank, that are 2 files apart, for example c5 and f5. For all such pawns, some bonus is due.
So, whenever you have pawns on a5 and d5, or b5 and e5, or c5 and f5, or d5 and g5, or e5 and h5 simultaneously, you give some 10cps overbonus.
Space advantage on both sides of the board should be scored only in the mg, as space is much more important in the mg, and besides in the eg this would already clash with passer code.
All of the above tandems of pawns control at least one square on the king side and at least one square on the queen side, meaning their influence is widespread.
[d]6k1/1pp5/p2p4/P2P4/8/8/8/6K1 w - - 0 1
a5 and d5 pawns gaining space advantage on both sides of the board
[d]6k1/8/2p2p2/2PppP2/8/8/8/6K1 w - - 0 1
c5 and f5 pawns gaining space advantage on both sides of the board
I still am not aware of an engine that successfully evaluates space advantage in terms of pawns in all its incarnations.
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Defending duos
What is a defending duo?
2 own pawns side by side on the 2nd or 3rd rank, when the enemy has 2 pawns side by side on the same files 2 ranks apart.
[d]6k1/8/8/pp6/6pp/PP6/6PP/6K1 w - - 0 1
a3-b3 is a defending duo for white, as it is on the 3rd rank; a5-b5 is a normal duo, as it is on its 4th rank
h2-g2 is another defending duo for white, as it is on the 2nd rank; h4-g4 is not a defending duo, but a normal duo, as it is on its 5th rank
So, whenever you find this symmetry, you can give some small bonus, 5cps for duos on the 3rd rank, and 10cps for duos on the 2nd rank. Thus, a3-b3 above will get 5cps overbonus, while h2-g2 10cps overbonus. As you see, in sharp distinction to most other cases, here you give the bonus for the pawns that are less advanced, and not the more advanced pawns.
Why is this bonus due?
Well, because such symmetrical structures help to blunt the enemy's attack, are able to close files and entire sides, etc. It is more difficult for the attacker to break through, when you have such defending duos.
For example, look above. When black advances one of its pawns on the queen side, a4 or b4, white will be able to close the game there, on a4 follows b4, while on b4 a4, and closing the game usually means blunting the enemy attack.
Same for the duos on the king side: no matter what pawn black advances, white is able to close the game, on h3 follows g3, while on g3 h3. As you see, in many cases this is strictly related to king safety: black would like to open files on the king side to attack the white king, but, when white has such a defending duo, part of the king shelter, chances are instead of opening files, the game will be closed there and the attack blunted.
Of course, the more advanced enemy duos get their nice bonus for being advanced, this is just a small overbonus for the ability to close the game.
Please note, that there is one further condition, in order for the bonus to work and be applied: the symmetry in terms of duo pawns must be full, i.e. no other duo pawns for either side on adjacent files, as this already breaks the symmetry and creates ample opportunities for efficient attack and breaking through. No other pawns for either side on adjacent files means none of the players has a double duo pawn, a pawn adjacent to 2 own pawns on the same rank on adjacent files.
[d]6k1/8/8/pp6/5ppp/PPP5/5PPP/6K1 w - - 0 1
the symmetry does not work on the queen side, black breaks through; white has a double duo pawn on b3
same for the king side, the symmetry does not work here, as both players have double duo pawns, g2 and g4 respectively. Black breaks through.
So, the condition that none of the players has a double duo pawn is absolutely necessary. You need full symmetry for the rule to work, just 2 pawns side by side for each player, with no additional duo pawns on adjacent files.
Again, the bonus is given for 2 duo pawns on the 2nd or 3rd rank side by side, when the enemy has fully symmetrical 2 duo pawns side by side on its 5th or 4th rank respectively.
Bonus might be considered just for the mg, but you might check what happens in the eg too.
2 own pawns side by side on the 2nd or 3rd rank, when the enemy has 2 pawns side by side on the same files 2 ranks apart.
[d]6k1/8/8/pp6/6pp/PP6/6PP/6K1 w - - 0 1
a3-b3 is a defending duo for white, as it is on the 3rd rank; a5-b5 is a normal duo, as it is on its 4th rank
h2-g2 is another defending duo for white, as it is on the 2nd rank; h4-g4 is not a defending duo, but a normal duo, as it is on its 5th rank
So, whenever you find this symmetry, you can give some small bonus, 5cps for duos on the 3rd rank, and 10cps for duos on the 2nd rank. Thus, a3-b3 above will get 5cps overbonus, while h2-g2 10cps overbonus. As you see, in sharp distinction to most other cases, here you give the bonus for the pawns that are less advanced, and not the more advanced pawns.
Why is this bonus due?
Well, because such symmetrical structures help to blunt the enemy's attack, are able to close files and entire sides, etc. It is more difficult for the attacker to break through, when you have such defending duos.
For example, look above. When black advances one of its pawns on the queen side, a4 or b4, white will be able to close the game there, on a4 follows b4, while on b4 a4, and closing the game usually means blunting the enemy attack.
Same for the duos on the king side: no matter what pawn black advances, white is able to close the game, on h3 follows g3, while on g3 h3. As you see, in many cases this is strictly related to king safety: black would like to open files on the king side to attack the white king, but, when white has such a defending duo, part of the king shelter, chances are instead of opening files, the game will be closed there and the attack blunted.
Of course, the more advanced enemy duos get their nice bonus for being advanced, this is just a small overbonus for the ability to close the game.
Please note, that there is one further condition, in order for the bonus to work and be applied: the symmetry in terms of duo pawns must be full, i.e. no other duo pawns for either side on adjacent files, as this already breaks the symmetry and creates ample opportunities for efficient attack and breaking through. No other pawns for either side on adjacent files means none of the players has a double duo pawn, a pawn adjacent to 2 own pawns on the same rank on adjacent files.
[d]6k1/8/8/pp6/5ppp/PPP5/5PPP/6K1 w - - 0 1
the symmetry does not work on the queen side, black breaks through; white has a double duo pawn on b3
same for the king side, the symmetry does not work here, as both players have double duo pawns, g2 and g4 respectively. Black breaks through.
So, the condition that none of the players has a double duo pawn is absolutely necessary. You need full symmetry for the rule to work, just 2 pawns side by side for each player, with no additional duo pawns on adjacent files.
Again, the bonus is given for 2 duo pawns on the 2nd or 3rd rank side by side, when the enemy has fully symmetrical 2 duo pawns side by side on its 5th or 4th rank respectively.
Bonus might be considered just for the mg, but you might check what happens in the eg too.
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Blocked spearheads
What is a spearhead?
The most advanced pawn of a long chain.
e5 is the spearhead of the white c3-d4-e5 long chain; h3 is the spearhead of the black e6-f5-g4-h3 long chain
The term stems from Mr. Carl Langan, as far as I am aware, and I like everything Mr. Langan suggests.
Computer chess wiki does not seem to make a distinction between the most advanced pawn of a long chain and any most advanced defended pawn, even if of a chain of only 2 pawns, for it all such more advanced defended pawns are called heads/peaks of a chain, but there certainly is a distinction.
[d]6k1/8/7p/3P2p1/2P5/1P6/8/6K1 w - - 0 1
d5 is a spearhead of a long chain; g5 is just a random head/peak/lead pawn
if g5 is strong, there is nothing stronger than a spearhead.
Spearheads are the strongest pawns on the board.
As you see, while g5 is just defended by a single own pawn, d5 is defended by an own pawn, which is in turn defended by another own pawn, which of course makes the whole structure and especially the spearhead much more enduring and eval-relevant.
Spearheads happen on ranks 4 through 7, files a through h.
Now, as we try to assess long chains in terms of their inner pawns, which seems reasonable, there is no need to give additional bonus to spearheads of long chains. However, there is a certain number of elements, for which spearheads would come quite in handy. One such element is whether the spearhead is blocked by an enemy pawn or not.
A blocked spearhead is a spearhead of a long chain that is blocked by an enemy pawn.
[d]6k1/8/3p3p/3P2pP/2P2pP1/1P6/8/6K1 w - - 0 1
d5 is a blocked spearhead of a long chain, d5 is blocked by the enemy d6 pawn; f4 is a spearhead of a long chain that is not blocked by enemy pawn
How should we assess blocked spearheads?
Well, any element thats is blocked is generally more enduring and this raises its value. As spearheads have mostly achieved what they strive, to get to a good square and be excellently defended, they do not need freedom of movement, such freedom is much more urgently needed by other chain pawns and other pawns in general. On the other hand, as said, the condition of being blocked means permanence of the feature. Therefore the bonus.
Any blocked spearhead is due some 10cps overbonus, both for the mg and eg.
Blocked spearheads happen on all files, ranks 4 through 6.
You can define spearheads as the most advanced pawn of a long chain, or a pawn that is defended by an inner chain pawn and is not an inner pawn itself.
Chain evaluation is extremely rich. And apart from chain evaluation, there is of course also blocked chain evaluation, which is a different and supposedly more advanced branch of chain evaluation. There is an awful lot to do in eval in terms of blocked/unblocked chains.
But, I ask myself, if engines have only a very superficial understanding of chain eval, how are they going to cope with blocked chain eval?
The most advanced pawn of a long chain.
e5 is the spearhead of the white c3-d4-e5 long chain; h3 is the spearhead of the black e6-f5-g4-h3 long chain
The term stems from Mr. Carl Langan, as far as I am aware, and I like everything Mr. Langan suggests.
Computer chess wiki does not seem to make a distinction between the most advanced pawn of a long chain and any most advanced defended pawn, even if of a chain of only 2 pawns, for it all such more advanced defended pawns are called heads/peaks of a chain, but there certainly is a distinction.
[d]6k1/8/7p/3P2p1/2P5/1P6/8/6K1 w - - 0 1
d5 is a spearhead of a long chain; g5 is just a random head/peak/lead pawn
if g5 is strong, there is nothing stronger than a spearhead.
Spearheads are the strongest pawns on the board.
As you see, while g5 is just defended by a single own pawn, d5 is defended by an own pawn, which is in turn defended by another own pawn, which of course makes the whole structure and especially the spearhead much more enduring and eval-relevant.
Spearheads happen on ranks 4 through 7, files a through h.
Now, as we try to assess long chains in terms of their inner pawns, which seems reasonable, there is no need to give additional bonus to spearheads of long chains. However, there is a certain number of elements, for which spearheads would come quite in handy. One such element is whether the spearhead is blocked by an enemy pawn or not.
A blocked spearhead is a spearhead of a long chain that is blocked by an enemy pawn.
[d]6k1/8/3p3p/3P2pP/2P2pP1/1P6/8/6K1 w - - 0 1
d5 is a blocked spearhead of a long chain, d5 is blocked by the enemy d6 pawn; f4 is a spearhead of a long chain that is not blocked by enemy pawn
How should we assess blocked spearheads?
Well, any element thats is blocked is generally more enduring and this raises its value. As spearheads have mostly achieved what they strive, to get to a good square and be excellently defended, they do not need freedom of movement, such freedom is much more urgently needed by other chain pawns and other pawns in general. On the other hand, as said, the condition of being blocked means permanence of the feature. Therefore the bonus.
Any blocked spearhead is due some 10cps overbonus, both for the mg and eg.
Blocked spearheads happen on all files, ranks 4 through 6.
You can define spearheads as the most advanced pawn of a long chain, or a pawn that is defended by an inner chain pawn and is not an inner pawn itself.
Chain evaluation is extremely rich. And apart from chain evaluation, there is of course also blocked chain evaluation, which is a different and supposedly more advanced branch of chain evaluation. There is an awful lot to do in eval in terms of blocked/unblocked chains.
But, I ask myself, if engines have only a very superficial understanding of chain eval, how are they going to cope with blocked chain eval?
-
carldaman
- Posts: 2287
- Joined: Sat Jun 02, 2012 2:13 am
Re: Blocked spearheads
Thanks for the kind flattering words, Mr. Lyudmil.Lyudmil Tsvetkov wrote:What is a spearhead?
The most advanced pawn of a long chain.
e5 is the spearhead of the white c3-d4-e5 long chain; h3 is the spearhead of the black e6-f5-g4-h3 long chain
The term stems from Mr. Carl Langan, as far as I am aware, and I like everything Mr. Langan suggests.
Computer chess wiki does not seem to make a distinction between the most advanced pawn of a long chain and any most advanced defended pawn, even if of a chain of only 2 pawns, for it all such more advanced defended pawns are called heads/peaks of a chain, but there certainly is a distinction.
[d]6k1/8/7p/3P2p1/2P5/1P6/8/6K1 w - - 0 1
d5 is a spearhead of a long chain; g5 is just a random head/peak/lead pawn
if g5 is strong, there is nothing stronger than a spearhead.
Spearheads are the strongest pawns on the board.
As you see, while g5 is just defended by a single own pawn, d5 is defended by an own pawn, which is in turn defended by another own pawn, which of course makes the whole structure and especially the spearhead much more enduring and eval-relevant.
Spearheads happen on ranks 4 through 7, files a through h.
Now, as we try to assess long chains in terms of their inner pawns, which seems reasonable, there is no need to give additional bonus to spearheads of long chains. However, there is a certain number of elements, for which spearheads would come quite in handy. One such element is whether the spearhead is blocked by an enemy pawn or not.
A blocked spearhead is a spearhead of a long chain that is blocked by an enemy pawn.
[d]6k1/8/3p3p/3P2pP/2P2pP1/1P6/8/6K1 w - - 0 1
d5 is a blocked spearhead of a long chain, d5 is blocked by the enemy d6 pawn; f4 is a spearhead of a long chain that is not blocked by enemy pawn
How should we assess blocked spearheads?
Well, any element thats is blocked is generally more enduring and this raises its value. As spearheads have mostly achieved what they strive, to get to a good square and be excellently defended, they do not need freedom of movement, such freedom is much more urgently needed by other chain pawns and other pawns in general. On the other hand, as said, the condition of being blocked means permanence of the feature. Therefore the bonus.
Any blocked spearhead is due some 10cps overbonus, both for the mg and eg.
Blocked spearheads happen on all files, ranks 4 through 6.
You can define spearheads as the most advanced pawn of a long chain, or a pawn that is defended by an inner chain pawn and is not an inner pawn itself.
Chain evaluation is extremely rich. And apart from chain evaluation, there is of course also blocked chain evaluation, which is a different and supposedly more advanced branch of chain evaluation. There is an awful lot to do in eval in terms of blocked/unblocked chains.
But, I ask myself, if engines have only a very superficial understanding of chain eval, how are they going to cope with blocked chain eval?
Nice of you to continue posting here. I have hardly had the time or energy to read, let alone post much, of late. 
If I read you correctly, I kind of like the blocked spearhead bonus, as the blockage actually causes congestion for the opponent and that often restricts enemy piece mobility, and in a more general case it is certainly useful for the other chain pawns to be free to advance.
In a more strict case, involving a chain pointed at the enemy king, I would prefer that as many chain pawns as possible be blocked, to form a full barrier and cause maximum congestion for the opponent. In such circumstances, it is the flank pawns on the g- and h-files (or a and b, if the King is castled there) that have to be mobile, for storming purposes.
Regards,
CL