Fire 1.31 settings.

[Charles]

I modified Fire1.31. to following settings:

// Search Parameters
Split_At_CN 1
Split_At_EN 0
Verify_Null 1
Null_Reduction 9
Verify_Reduction 8 //default 7
Aspiration_Window 8
AN_Split_Depth 12
CN_Split_Depth 14
EN_Split_Depth 12
PV_Split_Depth 12

// Eval Weights
Material_Weight 100
Positional_Weight 100
Contempt 64
Lazy_Eval_1 150
Lazy_Eval_2 300

// Piece Values
Pawn_Value 100
Knight_Value 300 //default 325
Bishop_Value 300
Rook_Value 500
Queen_Value 900 //default 975
Bishop_Pair 66 //default 50


And my results agains R4 + R3 100G silver suite ponder off tc=0+3s 4 cpu on Arena
Fire_131_w32_NS - Rybka 3_(cont 0) : 63.5/100 36-9-55 64%
Fire_131_w32_NS - Rybka 4 w32 : 49.0/100 22-24-54 49%

[Eduard Nemeth]

My favorite Settings for Fire 1.2 and 1.3.1:

//Piece Values

Pawn=105
Knight=300
Bishop=300
Rook=450
Queen=900
Bishop Pair=45

ED.

[PENSADOR]

I modified Fire1.31. to following settings:

// Search Parameters
Split_At_CN 1
Split_At_EN 0
Verify_Null 1
Null_Reduction 9
Verify_Reduction 8 //default 7
Aspiration_Window 8
AN_Split_Depth 12
CN_Split_Depth 14
EN_Split_Depth 12
PV_Split_Depth 12

// Eval Weights
Material_Weight 100
Positional_Weight 100
Contempt 64
Lazy_Eval_1 150
Lazy_Eval_2 300

// Piece Values
Pawn_Value 100
Knight_Value 300 //default 325
Bishop_Value 300
Rook_Value 500
Queen_Value 900 //default 975
Bishop_Pair 66 //default 50


And my results agains R4 + R3 100G silver suite ponder off tc=0+3s 4 cpu on Arena
Fire_131_w32_NS - Rybka 3_(cont 0) : 63.5/100 36-9-55 64%
Fire_131_w32_NS - Rybka 4 w32 : 49.0/100 22-24-54 49%

It gives very unaccurate evaluation and it says +-1,20 when actually it is +-0,20 and finally draw.

[Taner Altinsoy]

I modified Fire1.31. to following settings:

// Search Parameters
Split_At_CN 1
Split_At_EN 0
Verify_Null 1
Null_Reduction 9
Verify_Reduction 8 //default 7
Aspiration_Window 8
AN_Split_Depth 12
CN_Split_Depth 14
EN_Split_Depth 12
PV_Split_Depth 12

// Eval Weights
Material_Weight 100
Positional_Weight 100
Contempt 64
Lazy_Eval_1 150
Lazy_Eval_2 300

// Piece Values
Pawn_Value 100
Knight_Value 300 //default 325
Bishop_Value 300
Rook_Value 500
Queen_Value 900 //default 975
Bishop_Pair 66 //default 50


And my results agains R4 + R3 100G silver suite ponder off tc=0+3s 4 cpu on Arena
Fire_131_w32_NS - Rybka 3_(cont 0) : 63.5/100 36-9-55 64%
Fire_131_w32_NS - Rybka 4 w32 : 49.0/100 22-24-54 49%

It gives very unaccurate evaluation and it says +-1,20 when actually it is +-0,20 and finally draw.

I want to change fire 1.31 settings as well since I do not like the chess and the success rate against other engines, so I have a question to both of you. How many games did you play with your settings to verify that they are superior to default? Did you play with only one time control or did you check with different time controls as well?

[Charles]

I played the 200 games I posted above and also 40 games more against R4 in a slower tc which R4 won the nunn suite. So I don't know if my settings are better than the default ..

I have no idea what this means: It gives very unaccurate evaluation and it says +-1,20 when actually it is +-0,20 and finally draw. ... What position/game are you referring to ?

[PENSADOR]

In general, with those piece values, evaluation by the GUI is not precise/accurate. It shows a big advantage when it's a slight advantage that ends up in a draw. Instead of showing a realistic value it shows a big advantage.

[Charles]

I dont see what you are saying, the evaluation is not off when i observe the games vs R4. IN many cases it agreed with R4's eval.
What you say can only happen if you set ridiculous values -- and even more telling it would LOSE quite badly if the values give it a wrong evaluation.

Tuning the values is not easy and I am not sure if the default values are best ...of course I am not sure if my values are better either ..

[halestorm]

I'm also toying with Fire settings, and have not tried things other than piece values. Try a 4+2 match of no fewer than 100 games. 250 or 1000 is better. Engine v Engine (mod). Post your results.