Fire 2.2 xTreme

mare1975 wrote:

32 or 64 bit?

I use 64 bit, but these settings will work with 32 or 64 bit, and with any Fire compile.

Fire 2.2 GH Blaze x64 vs. Houdini 1.5a x64 Match is underway
Time Control: 3 Minutes
The results soon in the Tests Room

@David Dahlem wrote:

mare1975 wrote:

32 or 64 bit?

I use 64 bit, but these settings will work with 32 or 64 bit, and with any Fire compile.

Thanks...

Just amazing these settings of fire 2.2 I think it will become number 1 soon

@anmarahmed wrote:

Just amazing these settings of fire 2.2 I think it will become number 1 soon

results ?

Vladimir wrote:

Fire 2.2 GH Blaze x64 vs. Houdini 1.5a x64 Match is underway
Time Control: 3 Minutes
The results soon in the Tests Room

Fire 2.2 GH Blaze x64 vs. Houdini 1.5a x64 Match is over.
https://www.chess2u.com/t3534-fire-22-blaze-x64-vs-houdini-15a-x64#21082
Houdini 1.5a is still much stronger!

Vladimir wrote:

Vladimir wrote:

Fire 2.2 GH Blaze x64 vs. Houdini 1.5a x64 Match is underway
Time Control: 3 Minutes
The results soon in the Tests Room

Fire 2.2 GH Blaze x64 vs. Houdini 1.5a x64 Match is over.
https://www.chess2u.com/t3534-fire-22-blaze-x64-vs-houdini-15a-x64#21082
Houdini 1.5a is still much stronger!

Thanks Vladimir. I don't know if it makes much difference, but i used the Fire 2.2 xTreme x64 compiled by Peterpan version in my tests, and i tested against a variety of engines.

By the way, did you have any time losses at 3 minutes sudden death time control? I had none at 2+2.

A question to David & Vladimir.

What system you are using..because i see a very high Split Depth in the settings:
AN_Split_Depth 14
CN_Split_Depth 16
PV_Split_Depth 12

Fire 2.2 has as min. value Split depth 12..where Fire 1.6 has default 10 and can put on min. 1 max. 24
I have a core i7 970 @4.155Ghz 6cores (HT OFF) and have Fire 1.6 with SD=10 better then all other Fire engines.So SD=10 do it better then SD=12 on my system.
I'am now testing SD=9 to see if Fire 1.6 doing better then SD=10 on my laptop.I would i can test Fire 2.2 with lower Split Depth!

So,how you come to these high Split Depth values and have you tried with lower values,for sure when you have a slower system!

Ipman.

PS. I will try these Blaze settings..Thanks!

@David Dahlem wrote:

Vladimir wrote:

Vladimir wrote:

Fire 2.2 GH Blaze x64 vs. Houdini 1.5a x64 Match is underway
Time Control: 3 Minutes
The results soon in the Tests Room

Fire 2.2 GH Blaze x64 vs. Houdini 1.5a x64 Match is over.
https://www.chess2u.com/t3534-fire-22-blaze-x64-vs-houdini-15a-x64#21082
Houdini 1.5a is still much stronger!

Thanks Vladimir. I don't know if it makes much difference, but i used the Fire 2.2 xTreme x64 compiled by Peterpan version in my tests, and i tested against a variety of engines.

By the way, did you have any time losses at 3 minutes sudden death time control? I had none at 2+2.

No difference, but Fire 2.2 xTreme GH is a bit faster!
Anyway, I will try another test for you :sm017:

Fire 2.2 xTreme pp Blaze x64 vs. Houdini 1.5a x64 Match is underway
Time Control: 3 Minutes
The results soon in the Tests Room

Vladimir wrote:

Fire 2.2 xTreme pp Blaze x64 vs. Houdini 1.5a x64 Match is underway
Time Control: 3 Minutes
The results soon in the Tests Room

Fire 2.2 xTreme pp Blaze x64 vs. Houdini 1.5a x64 Match is over.
https://www.chess2u.com/t3539-fire-22-xtreme-pp-blaze-x64-vs-houdini-15a-x64
once again, the same previous result. Houdini 1.5a is still much stronger!

ok.. i get hope for fire then it just crushes me.. i love to have an engine to beat houdini but all this false hope with fire 2.0,2.2 cremator,blaze,banzai etc.. is just beating around the bush...

PLEASE there is plenty of options and settings u can make.. how about somebody run a test like mindbreaker did for rybka with the diferent settings.. mindbreaker ran over 85+ settings and finally found a setting that proved to be very very strong ... makking rybka 4.0 exp.42 #2 strongest engine.. with this setting rybka gained well over +50ELO over default settings. I know fire has the potential.. and i believe it can overpower HOudini but if a 100 game test shows fire getting his butt handed to him then no need in cotinuing that setup.. move onto a new setting.. maybe a few minor adjustments..

but after a 100 game battle houdini 1.5a vs. fire 2.2 blaze without triple/total/tablebases(makes little difference since houdini had its GTB's disabled as well..) just engine vs. engine no tb's help. my results after 300 games was this @ 5/0 timer
went thru games and took out time losses, if fire was winning with a big advanatge over houdini and still lost on time i counted it a win for fire 2.2 Blaze.. vice versa goes for houdini if it lost on time but had a very big advantage it got the win..

results

i7-970 @ 4.2 GHz HT off
Ponder OFF
GTB"S off
Robbobases OFF
Houdini 1.5a x64 vs. Fire 2.2 Blaze using peterpans compile.

+141-56=103 in favor of houdini..

Hi No FEAR..it looks like we have same system!

I have now more then 17.000 5min.games..all these settings is nice.but the most i get out is when playing with Split Depth..these other settings don't do much in ELO gain.
Like you say,Mindbreaker did a great work,you need alot off testings to find better settings.
I have done this with Houdini 1.5a ..i don't know in your test what Split Depth you have used.
But i can tell you SD=12 is the best on our system!

Now for Fire..i have already told in my message before the min.Split depth 12 in fire 2.2 is too high,or you need a much faster system then ours..like a SR-2 system.
With Fire 1.6 you can go lower with these Split Depth's..and at the moment SD=9 is doing better then default SD=10 from Fire 1.6 (this is on the laptop)

On my big system is Fire 2.2 Blaze busy..but i would like to see a compile where i can reduce these split depth's under 12..like Fire 1.6 from 1 to 24.

So Fire 2.2 with default SD=12 will give you a worse result on a slower system.

Critters Split Depth has around half value from Houdini..so when default houdini SD=10 then is Critter SD=5 ..default is has SD=6

Testing now busy with Criter SD=5 against Fire 2.2 with blaze settings..already 5 draws.
I will continue with testings against different engines,but when i see it comes below Fire 1.6 with SD=10 i stop it after a few hunderd games.

So i would like to see a Fire 2.2 compile with Split Depth min.1 to max. 24

Ipman

@drmrboss wrote:

The Strength of an engine mainly depends on main cores of lines of codes, search methods and evaluations.

+1

@pogi wrote:

I know which one will beat Houdini 1.5a..... Houdini 2.0

I can't wait!

Carlos wrote:

@drmrboss wrote:

The Strength of an engine mainly depends on main cores of lines of codes, search methods and evaluations.

+1

+2

Changing strength by tweaking the setting is...'artifical'.

fire 2.2 will beat houdini 1.5a when the settings are found.. given the right settings i think fire 2.2 can overtake houdini. houdini = no options for diff settings. as fire has plenty of playing styles and ways of changing its playing strength/speed/depth. i have watched and tried all these settings and so far every newer setting gets closer to beating houdini straight up. other night i use houdini cremator in Playchess for over 14 hours and maintained a 2700 Elo. out of 74 games i lost only 9. drawed alot, won a few, but hey, fire stood up to houdini and beat its pants off and made houdini do a dissapear act and GIVEUP!!!

ipman send me a private message on here bro please.. thanks.

@N0 FEAR wrote:

fire 2.2 will beat houdini 1.5a when the settings are found.. given the right settings i think fire 2.2 can overtake houdini. houdini = no options for diff settings. as fire has plenty of playing styles and ways of changing its playing strength/speed/depth. i have watched and tried all these settings and so far every newer setting gets closer to beating houdini straight up. other night i use houdini cremator in Playchess for over 14 hours and maintained a 2700 Elo. out of 74 games i lost only 9. drawed alot, won a few, but hey, fire stood up to houdini and beat its pants off and made houdini do a dissapear act and GIVEUP!!!

By Qxf6
Stopped the match after 210 games:
Fire2.2GH‡Cremator3 - Houdini1.5a‡ 117.5-92.5 +67 =101 -42 55,9% +41 Elo

Engines: Fire 2.2 xTreme GH x64.exe (with Cremator3 settings and selected Totalbases+Triplebases) vs. Houdini_15a_x64.exe (with Gaviota tablebases)
Time control: 1 minute per game Both engines: 64-bit, 4 CPU Same openings for white & black 256 MB hash
System: i5 760 @ 3 Ghz, Win 7, 64-bit, 4 GB RAM
(Fire settings may not work as well on different systems)

Games (PGN): 2shared.com/file/pPlUZAMS/Cr3-Houdini.html

Cremator3 settings:

http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=22619

fire.cfg

// Fire 2.2 Cremator 3 fire.cfg

// System
Verbose_UCI 0
Write_Log 0
Hash 128
Pawn_Hash 32
Max_Threads 64
MultiPV 1
Random_Range 20

// RobboBases
Use_RobboBases 0
Verbose_RobboBases 0

// TotalBases
AutoLoad_TotalBases 0
TotalBase_Path c:\TotalBases
TotalBase_Cache 1

// TripleBases
AutoLoad_TripleBases 0
TripleBase_Path c:\TripleBases
TripleBase_Hash 1
Dynamic_TripleBase_Cache 1

// Eval Weights
Draw_Weight 100
King_Safety_Weight 110
Material_Weight 100
Mobility_Weight 110
Pawn_Weight 100
Positional_Weight 100
PST_Weight 100

// Lazy Eval
Lazy_Eval_Min 150
Lazy_Eval_Max 300

// Piece Values
Pawn_Value 100
Knight_Value 300
Bishop_Value 310
Rook_Value 505
Queen_Value 930
Bishop_Pair_Value 43

// Prune Thresholds
Prune_Check 10
Prune_Pawn 160
Prune_Minor 500
Prune_Rook 800

// Search Parameters
Aspiration_Window 14
Delta_Cutoff 25000
DepthRed_Min 16
Extend_In_Check 0
Height_Multiplier 64
History_Threshold 50
Low_Depth_Margin 1125
Min_Depth_Multiplier 64
Min_Trans_Move_Depth 18
Null_Reduction 9
QS_Alpha_Threshold 225
Search_Depth_Min 28
Search_Depth_Reduction 8
Top_Min_Depth 19
Undo_Count_Threshold 15
Value_Cut 18000
Verify_Null 1
Verify_Reduction 1

// Split Depths
Split_At_CN 1
AN_Split_Depth 14
CN_Split_Depth 16
PV_Split_Depth 12

// Time Management
Absolute_Factor 45
Battle_Factor 90
Easy_Factor 20
Easy_Factor_Ponder 38
Normal_Factor 45

// UCI Info strings
CPULoad_Info 0
CurrentMove_Info 0
Depth_Info 0
HashFull_Info 0
LowDepth_PVs 0
NPS_Info 0
TBHit_Info 0

Extend_In_Check 1

Also, Robbobases better be used.

I will test this setting to see if it gets better.

Fire 2.2 GH Cremator3 x64 vs. Houdini 1.5a x64 Match is underway
Time Control: 3 Minutes
The results soon in the Tests Room

@Qxf6 wrote:

Extend_In_Check 1

Also, Robbobases better be used.

Thanks. I missed the Extend_In_Check 1 parameter.
Also, personally i never use any endgame bases for any engines.

Thanks for Cremator 3. Based on the performances of your previous settings, i'm expecting great results from these new settings. It's time for the great Houdini to be dethroned!

@David Dahlem wrote:

@Qxf6 wrote:

Extend_In_Check 1

Also, Robbobases better be used.

Thanks. I missed the Extend_In_Check 1 parameter.
Also, personally i never use any endgame bases for any engines.

Thanks for Cremator 3. Based on the performances of your previous settings, i'm expecting great results from these new settings. It's time for the great Houdini to be dethroned!

Thanks. It is still an anti-Houdini personality and the performance is not so great against Critter, Rybka and Stockfish (while Inferno beats them quite easily). Bases: I think that Fire gets a lot of power from using them. But then, my primary interest is correspondence chess where they are an absolute necessity for analyzing endgames but I understand that outside corr chess they may not be so important.