pointy chains redux (part 1)

[cdani]

I think it is one of the problems of the evaluation, we "make score" when the position has occurred.
But this does not require engine to play it.
So we have to tell "the engine" this is good with another method.
There is much work to do.
Thinking that it was before "the chicken or the egg".

Have you a specific idea?

[carldaman]

I tried with more restrictive prerequisites, but also without luck.

Abstract:

If Number_of_pices_that_attack_the_king > 1 and oponent_piece_value > 2500 then ...

OK, Daniel, I'm glad you have been experimenting with the pawn chain code.

Could you check to see if the modified Andscacs at least makes the correct f5-f4 push in the following position:

[d]r1bq1rk1/pppnn1bp/3p2p1/3Ppp2/2P1P3/2N1BP2/PP2B1PP/R2QNRK1 b - - 0 11

And, could you also check to see if the reported eval score after f5-f4 is less lopsided than the unmodified Andscacs?

At the very least, it would be quite interesting to see if the changes actually result in improved play and eval in such positions, EVEN THOUGH no Elo is gained from it overall.

Thanks,
Carl

[cdani]

Could you check to see if the modified Andscacs at least makes the correct f5-f4 push in the following position:

[d]r1bq1rk1/pppnn1bp/3p2p1/3Ppp2/2P1P3/2N1BP2/PP2B1PP/R2QNRK1 b - - 0 11

And, could you also check to see if the reported eval score after f5-f4 is less lopsided than the unmodified Andscacs?

Two comparable versions play f4 most of the iterations. The one with the bonus (73171) gives as expected better eval to black than the other (73174).

73171:

Code: Select all

info depth 22 seldepth 35 score cp -59 nodes 16100383 nps 999030 time 16116 pv f
5f4 e3f2 g6g5 e1d3 c7c5 b2b4 b7b6 a2a3 d7f6 b4c5 b6c5 a1b1 e7g6 d1c2 d8e7 c3b5 f
8f7 f2e1 e7d8 e1d2 c8b7 b5c3 a7a6 f1e1
info depth 22 currmove a7a5 currmovenumber 2
info depth 22 currmove d7f6 currmovenumber 3
info depth 22 currmove c7c5 currmovenumber 5
info depth 22 currmove d7b6 currmovenumber 6
info depth 22 currmove d7c5 currmovenumber 9
info depth 22 currmove f8f6 currmovenumber 11
info depth 22 currmove b7b6 currmovenumber 12
info depth 22 currmove b7b5 currmovenumber 16
info depth 22 currmove g7h8 currmovenumber 22
info depth 23 currmove f5f4 currmovenumber 1
info depth 23 seldepth 36 score cp -59 nodes 28983762 nps 998992 time 29013 pv f
5f4 e3f2 g6g5 e1d3 c7c5 b2b4 b7b6 a2a3 c8a6 c3b5 a6b5 c4b5 f8f6 b4c5 b6c5 a3a4 f
6h6 d1d2 e7g6 d3b2 g7f6 b2c4 d7b6 d2c2 b6c4 e2c4 a8b8 a4a5 b8b7
info depth 23 currmove f8f6 currmovenumber 2
info depth 23 currmove b7b6 currmovenumber 3
info depth 23 currmove a7a5 currmovenumber 4
info depth 23 currmove d7f6 currmovenumber 5
info depth 23 currmove c7c5 currmovenumber 6
info depth 23 currmove d7b6 currmovenumber 7
info depth 23 currmove d7c5 currmovenumber 8
info depth 23 currmove c7c6 currmovenumber 9
info depth 23 currmove f5e4 currmovenumber 12
info depth 23 currmove b7b5 currmovenumber 15
info depth 23 currmove a7a6 currmovenumber 17
info depth 23 currmove h7h5 currmovenumber 21
info depth 24 currmove f5f4 currmovenumber 1
info depth 24 seldepth 35 score cp -62 nodes 41926320 nps 1000699 time 41897 pv
f5f4 e3f2 g6g5 e1d3 c7c5 b2b4 b7b6 a2a3 c8a6 c3b5 a6b5 c4b5 f8f6 b4c5 b6c5 a3a4
f6h6 d1d2 a8b8 d3b2 d8e8 b2c4 d7b6 d2c3 e8h5 h2h3 b6c4 e2c4 e7g6 a4a5
info depth 24 currmove b7b6 currmovenumber 2
info depth 24 currmove f8f6 currmovenumber 3
info depth 24 currmove d7f6 currmovenumber 4
info depth 24 currmove c7c5 currmovenumber 5
info depth 24 currmove a7a6 currmovenumber 6
info depth 24 currmove a7a5 currmovenumber 7
info depth 24 currmove d7b6 currmovenumber 8
info depth 24 currmove d7c5 currmovenumber 9
info depth 24 currmove f5e4 currmovenumber 10
info depth 24 currmove d8e8 currmovenumber 11
info depth 24 currmove c7c6 currmovenumber 12
info depth 24 currmove g8h8 currmovenumber 13
info depth 24 currmove g6g5 currmovenumber 14
info depth 24 currmove b7b5 currmovenumber 15
info depth 24 currmove e7c6 currmovenumber 17
info depth 24 currmove g7f6 currmovenumber 18
info depth 24 currmove f8f7 currmovenumber 19
info depth 24 currmove a8b8 currmovenumber 20
info depth 24 currmove h7h6 currmovenumber 21
info depth 24 currmove h7h5 currmovenumber 22

73174:

Code: Select all

info depth 22 seldepth 35 score cp -52 nodes 16058632 nps 1045688 time 15357 pv
f5f4 e3f2 g6g5 e1d3 c7c5 d5c6 b7c6 h2h4 h7h6 b2b4 c8b7 h4g5 h6g5 c4c5 d6d5 d1b3
f8f7 e4d5 e7d5 c3e4 d7f8 e4d6 f7d7 d6b7 d7b7 b3a4 e5e4 f3e4
info depth 22 currmove d7f6 currmovenumber 2
info depth 22 currmove c7c5 currmovenumber 3
info depth 22 currmove g6g5 currmovenumber 4
info depth 22 currmove d7b6 currmovenumber 5
info depth 22 currmove f8f7 currmovenumber 8
info depth 22 currmove d7c5 currmovenumber 11
info depth 22 currmove a7a5 currmovenumber 17
info depth 22 currmove f8f6 currmovenumber 19
info depth 23 currmove f5f4 currmovenumber 1
info depth 23 seldepth 38 score cp -54 nodes 21756931 nps 1042647 time 20867 pv
f5f4 e3f2 g6g5 e1d3 b7b6 b2b4 a7a5 a2a3 c8a6 d1c2 e7g6 b4b5 a6b7 a1d1 d8e7 f2e1
d7f6 f1f2 g6h4 e1d2 a8b8 d1b1 e7f7
info depth 23 currmove d7f6 currmovenumber 2
info depth 23 currmove d7b6 currmovenumber 3
info depth 23 currmove a7a5 currmovenumber 4
info depth 23 currmove g6g5 currmovenumber 5
info depth 23 currmove f8f6 currmovenumber 6
info depth 23 currmove c7c5 currmovenumber 7
info depth 23 currmove b7b5 currmovenumber 8
info depth 23 currmove f5e4 currmovenumber 9
info depth 23 currmove b7b6 currmovenumber 10
info depth 23 currmove g7f6 currmovenumber 13
info depth 23 currmove f8f7 currmovenumber 15
info depth 23 currmove a8b8 currmovenumber 17
info depth 23 currmove h7h6 currmovenumber 23
info depth 24 currmove f5f4 currmovenumber 1
info depth 24 seldepth 41 score cp -55 nodes 50233972 nps 989539 time 50765 pv f
5f4 e3f2 g6g5 e1d3 c7c5 b2b4 b7b6 a1b1 f8f6 d1c2 f6h6 a2a3 c8a6 c3b5 a6b5 c4b5 e
7g6 b4c5 b6c5 a3a4 d8c7 f2e1 a8c8 e1c3 c5c4 d3f2 d7f6 c2b2
info depth 24 currmove b7b6 currmovenumber 2
info depth 24 currmove d7f6 currmovenumber 3
info depth 24 currmove c7c5 currmovenumber 4
info depth 24 currmove a7a5 currmovenumber 5
info depth 24 currmove d7b6 currmovenumber 6
info depth 24 currmove g6g5 currmovenumber 7
info depth 24 currmove b7b5 currmovenumber 8
info depth 24 currmove f8f6 currmovenumber 9
info depth 24 currmove f5e4 currmovenumber 10
info depth 24 currmove c7c6 currmovenumber 11
info depth 24 currmove a8b8 currmovenumber 12
info depth 24 currmove f8f7 currmovenumber 13
info depth 24 currmove g7f6 currmovenumber 15
info depth 24 currmove d7c5 currmovenumber 16
info depth 24 currmove a7a6 currmovenumber 17
info depth 24 currmove g7h8 currmovenumber 20
info depth 24 currmove h7h6 currmovenumber 23

May be this modification is not good for Andscacs (in fact it loses a few elo with it), simply because somehow it's already understanding relatively enough those types of positions. Really difficult to say.

[Lyudmil Tsvetkov]

[d][d]r1bqk1nr/1pp3bp/n2p4/p2Pp3/2P2p2/2NB2N1/PP3PPP/R1BQK2R w KQkq - 0 10

Very briefly from me.

I still prefer black here.

A pointed chain/triade is a triade, and other factors are other factors.

Above, the black pointed chain/d6,e5,f4 triade of pawns alone compensates for the big lag in development of black, and the unsafe position of the king.
So really a major factor here, even though most of the pawns of the chain are not blocked.

I also do not quite agree with the necessity to requite the presence of a bishop sitting on a colour opposite the colour of the triade pawns: bishop pawns would already take this into account.
Besides, a triade is useful in all situations in the middlegame, regardless of bishops, presence of queens, etc.; the important hing is that you are still into the middlegame, so a lot of material present.

Another hint might be to possibly make a distinction in terms of the different triades: but maybe in some engines a c3,d4,e5 triade is worth more than a d3,e4,f5 triade, or vice-versa?
Maybe it might make sense to distinguish in terms of the size of the bonus, being more specific and avoiding possible redundancies. Who knows?

No surprise there, Lyudmil, I know what to expect by now.

I'll say this, though - Black suffers in the above diagram for tactical reasons; f5-f4 was basically forced, and objectively, Qh5+ is now very annoying, but only IF White plays it (and he did not). Black was forced into this line because he had no better options. The position was by no means easy to play for White, either, and he in fact did go wrong allowing Black to win, meaning that Black did have some real practical chances as well.

However, in all fairness, I posted a rather extreme example, just to illustrate that when congestion points are not present, it's anybody's game, and the pointy chain is no guarantee of an advantage under those less-than-ideal circumstances.

In the next installment, I'm going to show more balanced examples of what happens when the congestion point is missing.

Objective regards,
CL

You are looking at it subjectively, Carl. (I mean the above diagram)

Waiting for the new installment.

[cdani]

I tried with more restrictive prerequisites, but also without luck.

Abstract:

If Number_of_pices_that_attack_the_king > 1 and oponent_piece_value > 2500 then ...

I thought a little more about it. It's possible that it does not work simply because it already evaluates as good the number of squares attacked near the king in different ways. So a consequence of achieving this type of pawn chains tends to produce an increase in this bonus thanks to obtaining more space for own pieces, and less space for defending ones, so it does not needs the extra bonus of pawn chains to favor the attack.

[velmarin]

I thought a little more about it. It's possible that it does not work simply because it already evaluates as good the number of squares attacked near the king in different ways. So a consequence of achieving this type of pawn chains tends to produce an increase in this bonus thanks to obtaining more space for own pieces, and less space for defending ones, so it does not needs the extra bonus of pawn chains to favor the attack.

It is clear Daniel,
only in special circumstances this is well, how not always to start a storm of pawns, for example.
From there try to see these circumstances in a different way.
For example, previously evaluate favourable conditions without scores,
and if this condition is positive,
evaluate with score, even in search if necessary.
An idea.

[carldaman]

I tried with more restrictive prerequisites, but also without luck.

Abstract:

If Number_of_pices_that_attack_the_king > 1 and oponent_piece_value > 2500 then ...

I thought a little more about it. It's possible that it does not work simply because it already evaluates as good the number of squares attacked near the king in different ways. So a consequence of achieving this type of pawn chains tends to produce an increase in this bonus thanks to obtaining more space for own pieces, and less space for defending ones, so it does not needs the extra bonus of pawn chains to favor the attack.

Hi Daniel,

Your new version of Andscacs (0.72) plays an impressive kind of chess, so I'm not that surprised it may not need as much help in this area as most other engines.

There's more to chess than just Elo, and I think Andscacs can go far - both stylistically and strength-wise.

Regards,
CL

[cdani]

[Lyudmil Tsvetkov]

I tried with more restrictive prerequisites, but also without luck.

Abstract:

If Number_of_pices_that_attack_the_king > 1 and oponent_piece_value > 2500 then ...

I thought a little more about it. It's possible that it does not work simply because it already evaluates as good the number of squares attacked near the king in different ways. So a consequence of achieving this type of pawn chains tends to produce an increase in this bonus thanks to obtaining more space for own pieces, and less space for defending ones, so it does not needs the extra bonus of pawn chains to favor the attack.

Actually, it is usually the other way: you have good pointed pawn chain, but, at least until a certain point in time, less space for your pieces, meaning worse mobility, psqt and attacks.

The pointed chain bonus compensates precisely for that.

Another attempt is to make it valid just for closed positions, i.e. when all of the pawns of the triade are blocked by enemy pawns.

[cdani]

Another attempt is to make it valid just for closed positions, i.e. when all of the pawns of the triade are blocked by enemy pawns.

Ok!