Stockfish 1.5

[meschle]

Impressive results with a 3min blitz match against rybka 2.3.2a (both sp)

Engine Score St Ry S-B
1: Stockfish_15_ja......6.5/10.·········· 1100=11101 22.75
2: Rybka v2.3.2a.w32..3.5/10...........0011=00010 · 22.75

[solis]

Engine Score De St S-B
1: Deep Shredder 12.x64 5.0/10 ····· 3-3-4 25.00
1: Stockfish 15.x64 5.0/10 3-3-4 ····· 25.00

10 games played / Tournament finished
Name of the tournament: Arena games
Site/ Country: Q9550-XP64, United States
Level: Tournament 40/4
Hardware: Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz4x 3500 MHz with 4,096 MB Memory
Operating system: Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)

[solis]

Engine Score De St S-B
1: Deep Shredder 12.x64 5.5/10 ····· 3-2-5 24.75
2: Stockfish 15.0x64 4.5/10 2-3-5 ····· 24.75

10 games played / Tournament finished
Name of the tournament: Engine Testing
Site/ Country: X2-X64, United States
Level: Tournament 40/4
Hardware: AMD Athlon(tm) 64 X2 Dual Core Processor 4800+Dual 2558 MHz with 2,048 MB Memory
Operating system: Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)

[swami]

Stockfish 1.5

Strategic Test Suite Conditions:

Arena GUI
10 seconds per position
600 positions
Core2Quad 32 bits, Q6600, 2 GB RAM, 2.4GHZ
Engine uses 132 Mb Hash.
Single Thread

Subject-wise Scores:

STS (v1.0) - Undermining:
82/100, Grade: S

STS (v2.0) - Open Files and Diagonals:
83/100, Grade: S

STS (v3.0) - Knight Outposts/Centralization/Repositioning:
79/100, Grade: A+

STS (v4.0) - Square Vacancy:
84/100, Grade: S

STS (v5.0) -Bishop vs Knight:
75/100, Grade: A+

STS (v6.0) -Re-Capturing:
78/100, Grade: A+

Overall Performance:

Total Score: 481/600

Overall Average: 80.16 %

Grade: S

[swami]

Thanks, Tord, Marco and Joona. Continue working on it more...

and ofcourse thanks to Jim as well.

[Uri Blass]

I consider using stockfish for analysis of my correspondence games because it has many personalities and I wonder if it is possible to add anti-opposite colour bishop personality parameter.

The idea is simply to make the evaluation closer to draw in case that the material is equal except pawns and the sides have equal number of rooks and equal number of queens single bishop and no knights.

Even if it is going to make stockfish weaker then it still may help in analysis of specific positions.

[mcostalba]

I consider using stockfish for analysis of my correspondence games because it has many personalities and I wonder if it is possible to add anti-opposite colour bishop personality parameter.

The idea is simply to make the evaluation closer to draw in case that the material is equal except pawns and the sides have equal number of rooks and equal number of queens single bishop and no knights.

Even if it is going to make stockfish weaker then it still may help in analysis of specific positions.

If this makes SF weaker I doubt has a sense to commit this change, at least for me, I am already too busy to try to make it stronger and make it weaker is not one of my top priorities right now.

Anyhow please take a look at endgame.h here are all the special endgames that have a specific evaluation function.

I would think you are looking for something like KBPPKB (bishop + pawns vs bishop) in this case you can look at the corresponding function ScalingFunction<KBPPKB>::apply() in endgame.cpp

I hope I have understood your request correctly...

[Uri Blass]

I consider using stockfish for analysis of my correspondence games because it has many personalities and I wonder if it is possible to add anti-opposite colour bishop personality parameter.

The idea is simply to make the evaluation closer to draw in case that the material is equal except pawns and the sides have equal number of rooks and equal number of queens single bishop and no knights.

Even if it is going to make stockfish weaker then it still may help in analysis of specific positions.

If this makes SF weaker I doubt has a sense to commit this change, at least for me, I am already too busy to try to make it stronger and make it weaker is not one of my top priorities right now.

Anyhow please take a look at endgame.h here are all the special endgames that have a specific evaluation function.

I would think you are looking for something like KBPPKB (bishop + pawns vs bishop) in this case you can look at the corresponding function ScalingFunction<KBPPKB>::apply() in endgame.cpp

I hope I have understood your request correctly...

1)The target is not to make stockfish weaker but to have more options
for the user.

My point is that reducing the score when there are opposite bishops may make the engine stronger in specific positions that I want to analyze but weaker in the general case and in this case of course reducing the score is not going to be the default personality.

2)I did not talk about simple endgames but about positions of KQRBPPPPP vs KQRBPPPP when I feel that they are drawn.

The following position happened in one of my games and I believe it is probably a draw so I would like to see computers evaluate it closer to draw inspite of the fact that white is a pawn up in case that I use a special personality.

[D]2r1r1k1/1pq2pp1/p2p2b1/P2P4/2P5/3p1P1P/P2Q1BP1/2R1R1K1 w - - 0 26

The game is already at move 39 of white and I do not ask for advice in a specific game but only for personality that is going to be against opposite color bishops if I try to win.

Uri

[mcostalba]

The following position is from one of my games and I believe it is a draw
so I would like to see computers evaluate it closer to draw inspite of the fact that white is a pawn up.

Ok. In evaluate.cpp there is this code snippet:

Code: Select all

  // If we don't already have an unusual scale factor, check for opposite
  // colored bishop endgames, and use a lower scale for those
  if (   phase < PHASE_MIDGAME
      && pos.opposite_colored_bishops&#40;)
      && (   &#40;factor&#91;WHITE&#93; == SCALE_FACTOR_NORMAL && ei.egValue > Value&#40;0&#41;)
          || &#40;factor&#91;BLACK&#93; == SCALE_FACTOR_NORMAL && ei.egValue < Value&#40;0&#41;)))
  &#123;

It already scales for opposite colored bishop...is this what you meant ?

[Uri Blass]

The following position is from one of my games and I believe it is a draw
so I would like to see computers evaluate it closer to draw inspite of the fact that white is a pawn up.

Ok. In evaluate.cpp there is this code snippet:

Code: Select all

  // If we don't already have an unusual scale factor, check for opposite
  // colored bishop endgames, and use a lower scale for those
  if (   phase < PHASE_MIDGAME
      && pos.opposite_colored_bishops&#40;)
      && (   &#40;factor&#91;WHITE&#93; == SCALE_FACTOR_NORMAL && ei.egValue > Value&#40;0&#41;)
          || &#40;factor&#91;BLACK&#93; == SCALE_FACTOR_NORMAL && ei.egValue < Value&#40;0&#41;)))
  &#123;

It already scales for opposite colored bishop...is this what you meant ?

The question is if this is relevant for the position that I posted.

I do not know what does
phase < PHASE_MIDGAME means exactly.

I understand that scalefactor(50) is used in the case that there are more than 2 bishops
If I understand correctly reducing 50 to smaller value may make the evaluation closer to draw so it may be possible to replace the number 50 by a parameter that the user can change when 50 is the default value.

Uri