my strongest Houdini-setting

[pohl4711]

Hello,

these are the final test-results of my final and strongest pohl4-setting for Houdini 2.0b.
LittleBlitzer GUI, 1 core per engine, no TBs, no books, 2.83GHz Quad PC.

Test 1: 30 seconds + 500 ms Fischer bonus. 250 positions from the Frank Q. database.

Houdini 2.0b xp - Houdini 2.0b x64 263.0 - 237.0 +154/-128/=218 52.60%
Houdini 2.0b xp - Rybka 4.1 329.0 - 171.0 +243/-85/=172 65.80%
Houdini 2.0b xp - Critter 1.2 64-bit 312.5 - 187.5 +200/-75/=225 62.50%
Houdini 2.0b xp - Stockfish 2.1.1 JA 64bit 362.5 - 137.5 +284/-59/=157 72.50%
Houdini 2.0b xp - Komodo64 3 323.5 - 176.5 +247/-100/=153 64.70%

Test 2: 120 seconds + 1000 ms Fischer Bonus. 60 Noomen-positions (Noomen-testsets from year 2006 & 2008)

Houdini 2.0b xp - Houdini 2.0b x64 63.0 - 57.0 +36/-30/=54 52.50%
Houdini 2.0b xp - Rybka 4.1 79.0 - 41.0 +59/-21/=40 65.83%
Houdini 2.0b xp - Critter 1.2 64-bit 74.5 - 45.5 +50/-21/=49 62.08%
Houdini 2.0b xp - Stockfish 2.1.1 JA 64bit 80.0 - 40.0 +59/-19/=42 66.67%
Houdini 2.0b xp - Komodo64 3 79.0 - 41.0 +61/-23/=36 65.83%

I believe that this setting is at least a little stronger, than the default Houdini.


Houdini 2.0b Pohl4 Setting (can be used in Houdini 1.5a, too, but I didnt test that):
64828E7D642D2020202020202020202020202020202020202020202020201A1E2412242628

material values (100,330,342,500,950,45), and
pawn shield outer file = 26
pawn shield inner file = 30
pawn shield middle file = 36
pawn shield diagonal = 18
pawn storm outer file = 36
pawn storm inner file = 38
pawn storm middle file = 40

You can use the HoudiConfig tool. Or a HEX-Editor.


For the replacing with a HEX-Editor search for the ASCII-strings:

Houdini 1.5a default:
5D717974912A20202020202020202020202020202020202020202020202020202020202020

Houdini 2.0b default:
5D717974702A2020202020202020202020202020202020202020202020201E2328161C1E20

and replace them with the Pohl4-Setting ASCII-string from above.

Have fun - Stefan

[jpqy]

Thanks Stefan..will try them out!

JP.

[David Dahlem]

Thanks. I will test your Pohl4 settings.

[Jamal Bubker]

thanks Stefan ! I will try these settings with Houdini 1.5a.

[ivoryknight]

Hello,

these are the final test-results of my final and strongest pohl4-setting for Houdini 2.0b.
LittleBlitzer GUI, 1 core per engine, no TBs, no books, 2.83GHz Quad PC.

Test 1: 30 seconds + 500 ms Fischer bonus. 250 positions from the Frank Q. database.

Houdini 2.0b xp - Houdini 2.0b x64 263.0 - 237.0 +154/-128/=218 52.60%
Houdini 2.0b xp - Rybka 4.1 329.0 - 171.0 +243/-85/=172 65.80%
Houdini 2.0b xp - Critter 1.2 64-bit 312.5 - 187.5 +200/-75/=225 62.50%
Houdini 2.0b xp - Stockfish 2.1.1 JA 64bit 362.5 - 137.5 +284/-59/=157 72.50%
Houdini 2.0b xp - Komodo64 3 323.5 - 176.5 +247/-100/=153 64.70%

Test 2: 120 seconds + 1000 ms Fischer Bonus. 60 Noomen-positions (Noomen-testsets from year 2006 & 2008)

Houdini 2.0b xp - Houdini 2.0b x64 63.0 - 57.0 +36/-30/=54 52.50%
Houdini 2.0b xp - Rybka 4.1 79.0 - 41.0 +59/-21/=40 65.83%
Houdini 2.0b xp - Critter 1.2 64-bit 74.5 - 45.5 +50/-21/=49 62.08%
Houdini 2.0b xp - Stockfish 2.1.1 JA 64bit 80.0 - 40.0 +59/-19/=42 66.67%
Houdini 2.0b xp - Komodo64 3 79.0 - 41.0 +61/-23/=36 65.83%

...

Have fun - Stefan

Much more testing needs to be done. 1,000 games. Also, is it the same at longer time controls? Have any tests been done with H2 in aggressive mode vs H2 in default mode? If so, how do those results compare to yours? Curious to know.

[perejaslav]

Code: Select all

 
1.  Houdini 1.5a x64 Pohl4   	194.0/405
2.  Houdini 1.5a x64         	211.0/405

Testing aborted. No improvements detected over the default
I used fast TC 10"+0.2"

[MikeGL]

Strange.

I can't find your hex strings in my Houdini1.5a both in 64bit and 32bit versions.

5D717974912A20202020202020202020202020202020202020202020202020202020202020


I found a bunch of 20 20 20 20 20 strings but its prefix hex is 04 04 04 or 10 10 10, and not your: 5D 71 79 74 91 2A

[pohl4711]

Strange.

I can't find your hex strings in my Houdini1.5a both in 64bit and 32bit versions.

5D717974912A20202020202020202020202020202020202020202020202020202020202020


I found a bunch of 20 20 20 20 20 strings but its prefix hex is 04 04 04 or 10 10 10, and not your: 5D 71 79 74 91 2A

As I wrote: search for ASCII-Strings, not HEX-Strings...

S.Pohl

[MikeGL]

That was quick.
Thanks, I got it now.