IvanHoeV57 /V55 Tests

[img][You must be registered and logged in to see this link.][/img]
same books , ponder off
Whilst test is an unfair one as DR4 had no Namilov nor large pages we let it run on to test integrity of Ivan 57 and it was exellent, could be faster but , it was stable , no crashes and read bases
Anyoe who can enable Large page and or Namilov please test same engines

Great result! I'm now finishing one large test (Houdini vs Rybka), but I will test Ivanhoe 57 as soon as posible! But before that I will need to know which is the more stable and strong to test vs Houdini the winner in my tests.

Aitor

Why is that version 57? Do you count in reverse?

kicker04 wrote:

Why is that version 57? Do you count in reverse?

yes

"B" wrote:

[img][You must be registered and logged in to see this link.][/img]
same books , ponder off
Whilst test is an unfair one as DR4 had no Namilov nor large pages we let it run on to test integrity of Ivan 57 and it was exellent, could be faster but , it was stable , no crashes and read bases
Anyoe who can enable Large page and or Namilov please test same engines

That is excellant results B! Unforunately 9.57b does not like to run on my system it crashes instantly I knew the decembrists would not releast this engine unless they felt is was better than DR4. I think from whati haveseen and what little i could test agasint dr4. I like what i am seeing. I tried out T57 which crashes a lot but atleast it runs and it looks really strong better than dr4.

Theyre are real issues with ALL these compiles reading Robbobases ,,57b certainly did dureing test but they will no longer load and vanish from paramoters if you use as main engine,,,these compilers simply do not bother to check all works before churning them out

Intel QX 9775 (8x4090 MHz). Blitz 3'+0". Hash 128. Ponder off. Tablebases
5 pieces. book HS-7moves.ctg.
Deep Rybka 4 - IvanHoe T57A : 27.5 -22.5 . 55% -45%. (+13 -8 =29)


Sandi

Good news!
Time control 1' +1, Hash 256, Rybka 4 book 6 moves max, 6 threads vs 6 threads, ponder on,

Ivanhoe T57a X64 +20 wins, 51 draws, -15 losses, +20 elo

Not bad. Not bad. So far i like it!

Is it stronger than Ivanmod 12c?

kicker04 wrote:

Is it stronger than Ivanmod 12c?

not even sure what engines he used ..or played against looking at that post . but weve found T57 rubbish ivan57a is the best of the v57's so far

Sorry i thought it typed what engine it was playing against. T57A was playing DR SSE4.2 X64

This is my finaly results from the same tournament

T57A x64 vs DR4 SSE 4.2 X64 52 Wins, 21 Losses 97 Draws +64 elo rating

Ivanhoe clearly wining in 170 games played. Amazing

[You must be registered and logged in to see this link.] wrote:

Ivanhoe clearly wining in 170 games played. Amazing

In what time? blitz?

[You must be registered and logged in to see this link.] wrote:

Sorry i thought it typed what engine it was playing against. T57A was playing DR SSE4.2 X64

This is my finaly results from the same tournament

T57A x64 vs DR4 SSE 4.2 X64 52 Wins, 21 Losses 97 Draws +64 elo rating

Ivanhoe clearly wining in 170 games played. Amazing

bases, larges pages, crashes ?

The time control was 1' +1" Tb used for rybka 4 but no bases for ivanhoe. The ivanhoe engine did crash at game 171 at which the tourney stopped. But it took a long time to get it to crash. probably about 12 hours. Large page mode not used for rybka 4. With this performance i would say in general that the Ivanhoe's that are version .57 are definately the strongest to date of all ivanhoes and most like the strongest engine in the world above Deep Rybka 4 even if large pages is turned on.

General observations about the T57A engine. This engine seems to play extremely agressive. It's style of play is right on the cutting threshold of reckless! and Quite often at times playing what might appear to be unsound chess. The engine scoring at some postions that are unsual or highly unbalanced seem to go back and forth from winning and back to drawing and/or even sometimes appearing as loosing. When the postion a moment ago appeared to be winning. This appears to be typical of this engine. Not sure you can really rely on the scoring for analysis of this engine while in analysis mode rather then in auto play mode.

Some of the losses that i am sure some of you have seen with this engine. When this engine is loosing it looses badly and sometimes in less than 40 moves. I think one of my games that it lost in the 170 total that it played was like a 24 move game.

But all in all. I feel that the .57 engine is much sharper in tactical and postional and the new code that handles passed pawns seems to really do well in endgames. Completely out playing Rybka 4 in endgames with this heuristic or on the flip side it has handled some tough end games and defended very sharply to get the draw. It definately is an exciting engine to watch.

I hope that the developers of this new breed of ivanhoe expand upon its new abilities and refine it some. and try to make the engine slightly less reckless in its approach to the game. and with this i feel they could tweak a few more elo out of it.


Regards,

[img][You must be registered and logged in to see this link.][/img]


Again a rather unfair test as R4 had NO Large pages NOR Namilov table base wheras Ivan had TripleBases
A simple first to 10 wins test has produced dissapointing result

Perhaps the compilers puts all his faith in Intel compiles now?

Re: IvanHoe
Ivanhoe-B63Mod12c vs. Ivanhoe B57b(Amd64 PP compile) 2'+1", F12GUI

+7=23 -0

I haven't had any problems with the Amd64 compiles, except it keeps losing to 63mod12c!

IvanB57 seems to play better chess, but these losses don't support it. I'll try some single core games, some of my results point to a flaw in multi-p.

posted by sirbac

[img][You must be registered and logged in to see this link.][/img]

we allowed test to run on but sadly things got no better for this compile

Two more tests

Ivanhoe-B63Mod12c(1-cpu) vs. Ivanhoe B57b(2-cpu, Amd64 PeterPan compile) 2'+1", F12GUI

0-1=9

This is a bad result for IvanB57. Could Robo 0.085g3 beat it?

sirbac

================================

Core i7 [You must be registered and logged in to see this link.] 6GB RAM, HT on

Blitz 5', 16 matches, Nauum2

Nalimov EGTB 3-4-5, TripleBases+TotalBases, 1020MB Hash, 8 cores

Result:
Deep Rybka 4 x64 SSE42 - Ivanhoe 57c x64 SSE42 (PeterPan latest compile)
+5=10-1

At match 17th, Ivanhoe crashed ...

Very disappointing results I'd say


Paulo

This was also sent here

another from "sirbac " begs an interesting question and shows all theres been no real improvement of elo

[Event "ippolit7moves, Blitz 1m+1s"]
Ivanhoe-B57bwin32sse2 vs RobboLito 0.085g3 w32

AMD Athlon(tm) 64 X2 Dual Core Processor 4800+ 2511 MHz
W=16.3 plies; 1,525kN/s; IPPOLIT 7moves.ctg
B=15.4 plies; 960kN/s; IPPOLIT 7moves.ctg
+0=9 -1.

Except for node count, how do we know the extra cores are contributing to the analysis? I'll try a test position and see how long to solve it.

Its interesting how there's been very little good results and mostly bad. Looks like they got their work cut out to really try to make this a great engine

That is very interesting! good work!

my test

Level: Blitz 5/1
Hardware: Intel(R) Xeon(R) CPU X5355 @ 2.66GHz with 4,096 MB Memory
Operating system: Windows 7 Ultimate Professional (Build 7600)

-----------------Houdini_x64_4CPU-----------------
Houdini_x64_4CPU - IvanHoe999957a : 35.0/68 14-12-42 (==1=101==0=====1=0=11===01==1=0========0==0==0==1===11=1==1=0=0=01=0) 51% +7
-----------------IvanHoe999957a-----------------
IvanHoe999957a - Houdini_x64_4CPU : 33.0/68 12-14-42 (==0=010==1=====0=1=00===10==0=1========1==1==1==0===00=0==0=1=1=10=1) 49% -7

BTO7 wrote:

My new idea .... POWER TUNING with STS ! Ok folks this is looking promising !! I used STS to power tune PeterPans new 57c Knights and Bishops. The over all gain i got overall was close to 4% overall gain in STS.

Knight back and white..default 300 STS=298 was 3% gain

Bishop pair Black and White default 45 STS=35 WOW!? was 3-5% gain


Then i played 25 games 4+2 ....No ponder ...No books....No TBs....1024 hash ...pawn hash 128 ...2.4Ghz core2duo 4 gig of ram

AFTER TUNING

-----------------Houdini_102w32_2CPU-----------------
Houdini_102w32_2CPU - Ivan57cx32SSE3 : 10.5/25 5-9-11 (===10=010=====010110==000) 42% -56
-----------------Ivan57cx32SSE3-----------------
Ivan57cx32SSE3 - Houdini_102w32_2CPU : 14.5/25 9-5-11 (===01=101=====101001==111) 58% +56


I'm really excited. I have not even tuned the pawns, queen or rooks and this is great results ...the best i have ever had against Houdini no less. Great Job PeterPan !!

Regards
BT

NO... great job you

Now do same with best v57 to date .....Ivan 9.57c ....it is stronger than pps

AMD-X6.x64 Blitz:10', Blitz:10' 0

1 IvanHoe 9.57b x64 3180 +5/-3/=22 53.33% 16.0/30
2 Houdini 1.02 x64 8_CPU 3175 +3/-5/=22 46.67% 14.0/30