Improving SF passer code
Moderators: hgm, Rebel, chrisw
-
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: A backward proposal for SF
f5 is the best move, but it creates a strong advanced protected pawn, being a storming pawn at the same time, and further limits the black knight.
-
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: A backward proposal for SF
You surprised me with this.michiguel wrote:You are 4 years lateLyudmil Tsvetkov wrote:[d]r2qk2r/1b3ppp/p2p1b2/2nNp3/1R2P3/2P5/1PN2PPP/3QKB1R w Kkq - 3 17
As Gaviota is a backward pawn expert even without having backward pawns eval in its code, does Gaviota find Rb7 above, with substantial white advantage?
[d]r1bqkb1r/1p3ppp/p1nppn2/8/2P1P3/N1N5/PP3PPP/R1BQKB1R b KQkq - 1 8
Why do you think Kasparov plays here d6-d5, sacrificing a pawn, just to get rid of the bind on the central d5 square?
And this, in the heat of the World championship match?
As a prominent bind/backward pawns expert, does Gaviota find d6-d5 above as the best move for black?
http://www.talkchess.com/forum/viewtopi ... 28&start=0
And this has nothing to do with backward pawns. It is about color weaknesses and outposts. The backward pawn is there with or without the sacrifice.
Miguel
I presume you played the rook sacrifice for Gaviota in her place.
So, Gaviota is the only engine seeing the sacrifice and giving white advantage. You explained what makes Gaviota do so: big bonus for knight outpost on d5, penalty for low mobility piece on b7, bad black dark-coloured bishop.
I think most strong programs have those terms.
Why is it then that only Gaviota sees the sacrifice? Maybe the size of the bonus?
Until you give reasonable explanation, I am willing to accept conspiration theories.
The sacrifice has all to do with backward pawns, as, if there was not such a pawn on d6, which is the main feature of the position, the sacrifice would have been altogether impossible, as the black knight on b7 would have had good escape square on d6; but now, the d6 pawn prevents that.
-
- Posts: 3232
- Joined: Mon May 31, 2010 1:29 pm
- Full name: lucasart
Re: A backward proposal for SF
It's not that simple. There is a reason why SF considers connected pawns not backward. I'm pretty sure changing that would be a regression:Lyudmil Tsvetkov wrote:Now you start to realise.lucasart wrote:Here's a more interesting example:
[d]8/1pp5/p7/P1P5/8/8/8/8 w - - 0 1
Here, SF will give connected bonus to all 3 black pawns (2 phalanx and 1 supported), and will penalize both white pawns for being isolated. In reality, the two white pawns form an important bind on the b6 square.
When will you push a patch with this?
* what if c5 isn't there ?
* what if there's a pawn on a7 ?
* what if the weak pawn is defended by a pawn (move everything down one rank and have a black pawn defend b6 in a7 and/or c7) ?
Theory and practice sometimes clash. And when that happens, theory loses. Every single time.
-
- Posts: 6401
- Joined: Thu Mar 09, 2006 8:30 pm
- Location: Chicago, Illinois, USA
Re: A backward proposal for SF
You read it already. It is in the thread.Lyudmil Tsvetkov wrote:You surprised me with this.michiguel wrote:You are 4 years lateLyudmil Tsvetkov wrote:[d]r2qk2r/1b3ppp/p2p1b2/2nNp3/1R2P3/2P5/1PN2PPP/3QKB1R w Kkq - 3 17
As Gaviota is a backward pawn expert even without having backward pawns eval in its code, does Gaviota find Rb7 above, with substantial white advantage?
[d]r1bqkb1r/1p3ppp/p1nppn2/8/2P1P3/N1N5/PP3PPP/R1BQKB1R b KQkq - 1 8
Why do you think Kasparov plays here d6-d5, sacrificing a pawn, just to get rid of the bind on the central d5 square?
And this, in the heat of the World championship match?
As a prominent bind/backward pawns expert, does Gaviota find d6-d5 above as the best move for black?
http://www.talkchess.com/forum/viewtopi ... 28&start=0
And this has nothing to do with backward pawns. It is about color weaknesses and outposts. The backward pawn is there with or without the sacrifice.
Miguel
I presume you played the rook sacrifice for Gaviota in her place.
So, Gaviota is the only engine seeing the sacrifice and giving white advantage. You explained what makes Gaviota do so: big bonus for knight outpost on d5, penalty for low mobility piece on b7, bad black dark-coloured bishop.
I think most strong programs have those terms.
Why is it then that only Gaviota sees the sacrifice? Maybe the size of the bonus?
Until you give reasonable explanation, I am willing to accept conspiration theories.
No, the backward pawn is BEFORE and AFTER the sacrifice. So, whatever the penalty is, it won't matter to decide if the sacrifice is good or not. It cancels out.
The sacrifice has all to do with backward pawns, as, if there was not such a pawn on d6, which is the main feature of the position, the sacrifice would have been altogether impossible, as the black knight on b7 would have had good escape square on d6; but now, the d6 pawn prevents that.
Miguel
-
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: A backward proposal for SF
It will not be a regression.lucasart wrote:It's not that simple. There is a reason why SF considers connected pawns not backward. I'm pretty sure changing that would be a regression:Lyudmil Tsvetkov wrote:Now you start to realise.lucasart wrote:Here's a more interesting example:
[d]8/1pp5/p7/P1P5/8/8/8/8 w - - 0 1
Here, SF will give connected bonus to all 3 black pawns (2 phalanx and 1 supported), and will penalize both white pawns for being isolated. In reality, the two white pawns form an important bind on the b6 square.
When will you push a patch with this?
* what if c5 isn't there ?
* what if there's a pawn on a7 ?
* what if the weak pawn is defended by a pawn (move everything down one rank and have a black pawn defend b6 in a7 and/or c7) ?
c5 is not necessary, the penalty is due just with a single white pawn, because usually, statistically if you want, the square in front of such a backward pawn on the 2nd rank will be controlled by enemy pieces rather than own ones. That is why the penalty is due.
As said, you want to consider such backward connected pawns just for the 2nd rank.
They might also be relevant for 3rd rank, with lower values, but chances are current SF will see liability only in the case of the 2nd rank.
I know what I am talking about, watching many SF games.
-
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: A backward proposal for SF
You might just want to push a patch with my suggested values for the backward connected array for the 2nd rank, please see the subthread in this thread.
If it does not pass, this is my last message on the forum.
If it does not pass, this is my last message on the forum.
-
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: A backward proposal for SF
If there is one salient feature about the above position, it is the backward d6 pawn.michiguel wrote:No, the backward pawn is BEFORE and AFTER the sacrifice. So, whatever the penalty is, it won't matter to decide if the sacrifice is good or not. It cancels out.
Miguel
Such structures are known to be played along those lines.
You still did not acknowledge if you specifically tuned Gaviota eval terms to be able to cope with this position?