Hi,
I have a program which is good tactically but so laughable positionally that even I laugh when playing it or watching it play.
My goal is to *try* to get it to be positionally more solid but I don't have a good path forwards toward that goal.
I would have to say that I *think* the positional term coefficients may be to blame, wildly out of whack, even mis-chosen entirely or omitting important ones.
If anyone has a practical opinion on how to proceed, I'm all ears. Any information can be requested here.
--Stuart
horrid positional play in a solid tactical searcher
Moderators: hgm, Rebel, chrisw
-
- Posts: 737
- Joined: Wed Mar 08, 2006 8:08 pm
- Location: Orange County California
- Full name: Stuart Cracraft
-
- Posts: 4367
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
Re: horrid positional play in a solid tactical searcher
Piece-square tables and mobility terms alone should get you fairly reasonable play. Especially if auto-tuned. You can certainly add more terms but that is a good start IMO.
--Jon
--Jon
-
- Posts: 737
- Joined: Wed Mar 08, 2006 8:08 pm
- Location: Orange County California
- Full name: Stuart Cracraft
-
- Posts: 2488
- Joined: Tue Aug 30, 2016 8:19 pm
- Full name: Rasmus Althoff
Re: horrid positional play in a solid tactical searcher
Do the simple things first. You know, what chess beginners learn regarding positional play:
- pair of bishops is good, except in blocked positions
- isolated double pawns are usually bad
- a backward pawn is bad if on a semi-open file (mark that as target for the rooks)
- rooks belong on open files
- doubling rooks on open files is nice
- rook on the enemy pawn rank is good
- knights are good in the centre or even on the 6th/3rd rank
- a knight on the rim is dim
- early on, get the minor pieces out first
- castling is good if there is pawn cover
- usually, pawns capture towards the centre
You can put most of that in static eval. The usual weights are somewhere between 5 and 40 centipawns.
If nothing tactical is going on, then these positional things will guide the move selection.
- pair of bishops is good, except in blocked positions
- isolated double pawns are usually bad
- a backward pawn is bad if on a semi-open file (mark that as target for the rooks)
- rooks belong on open files
- doubling rooks on open files is nice
- rook on the enemy pawn rank is good
- knights are good in the centre or even on the 6th/3rd rank
- a knight on the rim is dim
- early on, get the minor pieces out first
- castling is good if there is pawn cover
- usually, pawns capture towards the centre
You can put most of that in static eval. The usual weights are somewhere between 5 and 40 centipawns.
If nothing tactical is going on, then these positional things will guide the move selection.
-
- Posts: 1494
- Joined: Thu Mar 30, 2006 2:08 pm
Re: horrid positional play in a solid tactical searcher
On top of this good list I would add King Safety and passed pawn evaluation.Ras wrote:Do the simple things first. You know, what chess beginners learn regarding positional play:
- pair of bishops is good, except in blocked positions
- isolated double pawns are usually bad
- a backward pawn is bad if on a semi-open file (mark that as target for the rooks)
- rooks belong on open files
- doubling rooks on open files is nice
- rook on the enemy pawn rank is good
- knights are good in the centre or even on the 6th/3rd rank
- a knight on the rim is dim
- early on, get the minor pieces out first
- castling is good if there is pawn cover
- usually, pawns capture towards the centre
You can put most of that in static eval. The usual weights are somewhere between 5 and 40 centipawns.
If nothing tactical is going on, then these positional things will guide the move selection.
-
- Posts: 737
- Joined: Wed Mar 08, 2006 8:08 pm
- Location: Orange County California
- Full name: Stuart Cracraft
Re: horrid positional play in a solid tactical searcher
Can you quantify in millipawns?Ras wrote:Do the simple things first. You know, what chess beginners learn regarding positional play:
- pair of bishops is good, except in blocked positions
- isolated double pawns are usually bad
- a backward pawn is bad if on a semi-open file (mark that as target for the rooks)
- rooks belong on open files
- doubling rooks on open files is nice
- rook on the enemy pawn rank is good
- knights are good in the centre or even on the 6th/3rd rank
- a knight on the rim is dim
- early on, get the minor pieces out first
- castling is good if there is pawn cover
- usually, pawns capture towards the centre
You can put most of that in static eval. The usual weights are somewhere between 5 and 40 centipawns.
If nothing tactical is going on, then these positional things will guide the move selection.
-
- Posts: 686
- Joined: Thu Mar 03, 2011 4:57 pm
- Location: Germany
Re: horrid positional play in a solid tactical searcher
Here you can find tons of inspiration, be selective and take the stuff that you like your engine to knowCan you quantify in millipawns?
http://talkchess.com/forum/viewtopic.ph ... 65&t=46166
-
- Posts: 7220
- Joined: Mon May 27, 2013 10:31 am
Re: horrid positional play in a solid tactical searcher
Skipper doesn't have that problem anymore:
[pgn]
[Event "Computer Chess Game"]
[Site "HP"]
[Date "2017.05.12"]
[Round "-"]
[White "Fairy-Max 4.8S"]
[Black "SkipperWinb"]
[Result "1/2-1/2"]
[TimeControl "120"]
[Annotator "1. +0.11 1... +0.07"]
1. c4 {+0.11/8} h5 {+0.07/260 2.1} 2. Qb3 {-0.01/8 4} c6 {+0.07/280 2.0} 3.
d3 {+0.27/9 6} g6 {+0.15/230 2.0} 4. Nc3 {+0.13/8 2.4} h4 {+0.12/220 2.0}
5. Nf3 {+0.28/9 2.0} Bg7 {-0.05/230 1.9} 6. Bg5 {+0.14/8 1.5} Bf6
{-0.04/230 1.9} 7. Be3 {+0.14/9 1.6} Bg7 {+0.00/200 1.9} 8. O-O-O
{+0.09/8 2.4} b6 {-0.09/170 1.8} 9. Bg5 {-0.10/8 3} Bf6 {-0.09/220 1.8} 10.
Bf4 {+0.03/8 1.4} Ba6 {-0.10/210 1.8} 11. g3 {+0.16/7 1.3} hxg3
{-0.18/140 1.8} 12. fxg3 {-0.02/8 1.5} Bb7 {-0.09/200 1.7} 13. c5
{-0.01/8 7} b5 {-0.06/180 1.7} 14. a4 {+0.14/8 1.3} a6 {+0.04/180 1.7} 15.
Ne4 {+0.10/8 1.9} Bg7 {+0.00/200 1.6} 16. Ne5 {-0.02/8 1.5} Nh6
{+0.04/180 1.6} 17. Bg2 {-0.09/7 1.9} Rf8 {-0.04/140 1.6} 18. Bf3
{-0.05/7 1.4} Nf5 {+0.02/160 1.6} 19. d4 {-0.19/7 1.6} Nh6 {+0.05/150 1.6}
20. Qe3 {-0.01/7 1.7} Nf5 {+0.00/200 1.5} 21. Qb3 {+0.00/10 1.4} Qa5
{+0.10/170 1.5} 22. Nc4 {-0.01/8 1.9} Qd8 {+0.00/140 1.5} 23. Ne5
{+0.00/8 1.6}
{XBoard adjudication: repetition draw} 1/2-1/2
[/pgn]
[pgn]
[Event "Computer Chess Game"]
[Site "HP"]
[Date "2017.05.12"]
[Round "-"]
[White "Fairy-Max 4.8S"]
[Black "SkipperWinb"]
[Result "1/2-1/2"]
[TimeControl "120"]
[Annotator "1. +0.11 1... +0.07"]
1. c4 {+0.11/8} h5 {+0.07/260 2.1} 2. Qb3 {-0.01/8 4} c6 {+0.07/280 2.0} 3.
d3 {+0.27/9 6} g6 {+0.15/230 2.0} 4. Nc3 {+0.13/8 2.4} h4 {+0.12/220 2.0}
5. Nf3 {+0.28/9 2.0} Bg7 {-0.05/230 1.9} 6. Bg5 {+0.14/8 1.5} Bf6
{-0.04/230 1.9} 7. Be3 {+0.14/9 1.6} Bg7 {+0.00/200 1.9} 8. O-O-O
{+0.09/8 2.4} b6 {-0.09/170 1.8} 9. Bg5 {-0.10/8 3} Bf6 {-0.09/220 1.8} 10.
Bf4 {+0.03/8 1.4} Ba6 {-0.10/210 1.8} 11. g3 {+0.16/7 1.3} hxg3
{-0.18/140 1.8} 12. fxg3 {-0.02/8 1.5} Bb7 {-0.09/200 1.7} 13. c5
{-0.01/8 7} b5 {-0.06/180 1.7} 14. a4 {+0.14/8 1.3} a6 {+0.04/180 1.7} 15.
Ne4 {+0.10/8 1.9} Bg7 {+0.00/200 1.6} 16. Ne5 {-0.02/8 1.5} Nh6
{+0.04/180 1.6} 17. Bg2 {-0.09/7 1.9} Rf8 {-0.04/140 1.6} 18. Bf3
{-0.05/7 1.4} Nf5 {+0.02/160 1.6} 19. d4 {-0.19/7 1.6} Nh6 {+0.05/150 1.6}
20. Qe3 {-0.01/7 1.7} Nf5 {+0.00/200 1.5} 21. Qb3 {+0.00/10 1.4} Qa5
{+0.10/170 1.5} 22. Nc4 {-0.01/8 1.9} Qd8 {+0.00/140 1.5} 23. Ne5
{+0.00/8 1.6}
{XBoard adjudication: repetition draw} 1/2-1/2
[/pgn]
-
- Posts: 759
- Joined: Fri Jan 04, 2013 4:55 pm
- Location: Nice
Re: horrid positional play in a solid tactical searcher
http://www.talkchess.com/forum/viewtopic.php?t=63870Henk wrote:Skipper doesn't have that problem anymore:
[pgn]
[Event "Computer Chess Game"]
[Site "HP"]
[Date "2017.05.12"]
[Round "-"]
[White "Fairy-Max 4.8S"]
[Black "SkipperWinb"]
[Result "1/2-1/2"]
[TimeControl "120"]
[Annotator "1. +0.11 1... +0.07"]
1. c4 {+0.11/8} h5 {+0.07/260 2.1} 2. Qb3 {-0.01/8 4} c6 {+0.07/280 2.0} 3.
d3 {+0.27/9 6} g6 {+0.15/230 2.0} 4. Nc3 {+0.13/8 2.4} h4 {+0.12/220 2.0}
5. Nf3 {+0.28/9 2.0} Bg7 {-0.05/230 1.9} 6. Bg5 {+0.14/8 1.5} Bf6
{-0.04/230 1.9} 7. Be3 {+0.14/9 1.6} Bg7 {+0.00/200 1.9} 8. O-O-O
{+0.09/8 2.4} b6 {-0.09/170 1.8} 9. Bg5 {-0.10/8 3} Bf6 {-0.09/220 1.8} 10.
Bf4 {+0.03/8 1.4} Ba6 {-0.10/210 1.8} 11. g3 {+0.16/7 1.3} hxg3
{-0.18/140 1.8} 12. fxg3 {-0.02/8 1.5} Bb7 {-0.09/200 1.7} 13. c5
{-0.01/8 7} b5 {-0.06/180 1.7} 14. a4 {+0.14/8 1.3} a6 {+0.04/180 1.7} 15.
Ne4 {+0.10/8 1.9} Bg7 {+0.00/200 1.6} 16. Ne5 {-0.02/8 1.5} Nh6
{+0.04/180 1.6} 17. Bg2 {-0.09/7 1.9} Rf8 {-0.04/140 1.6} 18. Bf3
{-0.05/7 1.4} Nf5 {+0.02/160 1.6} 19. d4 {-0.19/7 1.6} Nh6 {+0.05/150 1.6}
20. Qe3 {-0.01/7 1.7} Nf5 {+0.00/200 1.5} 21. Qb3 {+0.00/10 1.4} Qa5
{+0.10/170 1.5} 22. Nc4 {-0.01/8 1.9} Qd8 {+0.00/140 1.5} 23. Ne5
{+0.00/8 1.6}
{XBoard adjudication: repetition draw} 1/2-1/2
[/pgn]
Isa download :
-
- Posts: 2488
- Joined: Tue Aug 30, 2016 8:19 pm
- Full name: Rasmus Althoff
Re: horrid positional play in a solid tactical searcher
Yeah right. King Safety also from easy to hard because the easy stuff will already make a big difference.mjlef wrote:On top of this good list I would add King Safety and passed pawn evaluation.
- castling
- having a solid king pawn shield
- if g2 has moved to g3, there should be a bishop on g2, or at least, that bishop should not be missing on the board
- having a hole on g2 or h2 is good, but not on both
- passed pawns heightened in endgames
- supported passed pawns
- mark up passed pawns as good rook files if the rook is behind the pawn