IvanHoe V52 Tests

First topic message reminder :

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

KLO.s popcount vrs LeonardodoVincis Non popcount beta . same books + triple bases :cheers:

i fell asleep again :laf gal:

intel i5 cpu 430M @ 2.53 GHZ 4,096 MB Memory
Windows 7 Home Premium Edition 64BITS
Fritz Benchmark:
Speed: 9.58
KN/S: 3860
GUI: Rybka 4
Book: HS-7moves
Hash: 256 MB
RB and TB: OFF
Ponder: OFF
Large pages: ON for both Engines
Blitz: 3+0
CPU usage:both using 4T,Task Manager always open to check 100% cpu usage.
PS:IH is NOT located with RB file there for IMPOSIBLE usage of them + auto load is OFF
PS: 0 " no move " bug in 1330 match.




VANH-TOSH, Blitz 3m 0


1 Ivanhoe 52hUS-x64 +5 +266/=816/-248 50.67% 674.0/1330
2 Houdini 1.03a x64 POPCNT 4_CP -5 +248/=816/-266 49.33% 656.0/1330


@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

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


houdini ivanhoe

wonderfull world,
The road to the perfection is targetting.

peter pans latest is one of his weakest , it fails to beat rykba4

Sandi
Skulltrail Intel QX 9775 Yorkfield (8X4090 MHz)
Fritz Benchmark : 18385 k/n/s
Time 3'+0" (equivalent to 7.5' on a standard Quad and 10' on a fast dual core)
Hash 256. Ponder off. GUI Fritz 12. Tablebases 3-4-5 + Robbo Triple Bases.
Book: HS-7moves plus.ctg.

IvanHoe B52aDx64 - Deep Rybka 4 :
50/100 -50/100 . 50% - 50% (+13 -13 =74)

exactly this is what i have seen with my tournament.
now the target is rybka4

I do not know yet whether it is really stronger than Houdini (one needs more tests to do). But the results of 2 standard 200 games 1+0
night matches with Houdini

1 Ivanhoe 52iUSTM0-x64 +17 +41/=128/-31 52.50% 105.0/200
2 Houdini 1.03a x64 2_CPU -17 +31/=128/-41 47.50% 95.0/200

1 Ivanhoe 52iUSTM0-x64 +7 +44/=116/-40 51.00% 102.0/200
2 Houdini 1.03a x64 2_CPU -7 +40/=116/-44 49.00% 98.0/200
Vlad0

The tournament continue, here the first result


RITCHIE, Blitz 7m 0 1 2 3 4 5 6 7 8
1 Ivanhoe B 52aEx64 +0 -½ +½ -1 +½ -½ +½ -½ 4.0/8
2 Houdini 1.03a x64 2_CPU -1 +½ -½ +0 -½ +½ -½ +½ 4.0/8
3 Ivanhoe 52iUSTM0-x64 +½ -½ +1 -0 +½ 2.5/5
4 Deep Rybka 4 x64 -½ +½ -0 +1 -½ 2.5/5

Core2Duo 2.33
I know, more test is better , so i continue with more games.

one thing : the Knp/s is higher than Houdini all the time : 2200 for 2000
so may be the evaluation function is quickest than houdini, because Houdini use magic bit board and Ivanhoe does not.

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

Peter Pans own test of his latest compile

Blitz 1m+1s 0


1 Houdini 1.03a x64 4_CPU elo +8 +79/=294/-69 51.13% 226.0/442
2 Ivanhoe B 52aFx64 nocrash elo -8 +69/=294/-79 48.87% 216.0/442


:affraid:

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

These 1+1 games take on average over 5 minutes per game and reach deep in middle and deeper in end game play than stupid ultra fast or 1 min bullet games some testers foolishly believe show strenth of engines we KNOW they test the engines far more than ultra fast games , bullet games or games of up to 5 minutes without increment added

we also KNOW not useing book or bases tests engine full abilities

we know another 100 games MIGHT show different reults but we KNOW which engine we would use for real games

these games were Ponder OFF
NO bases
NO books
NO Large Pages

AMD 965 black phenon 3.4ghz , 8GB Ram . windows 64

Congratulations Vlad0 as this is in our AMD tests the strongest of recent compiles

"B" you do teriffic tests

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

Short test to judge speed of ahameds latest compile. which is on a par woth PPs , more games realy needed to state which, is strongest but please test for yourselves


Test was.....
No Books. No Bases No Large Pages Ponder OFF

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

Angels_77 re: Strongest Ivanhoe Today ?
Angels_77 16 minutes ago
and to prove our point re testing WITHOUT RoboBases..
Peter Pan posts...


3 little tests i conducted the last 3 nights... Time control is 1 minute

First one my compile uses default settings and no Robbobases.



1 IvanHoe T0.4 x64 final +3 +34/=168/-32 50.43% 118.0/234
2 Ivanhoe B 52aFx64 -3 +32/=168/-34 49.57% 116.0/234


Second one both compiles uses Robbobases



1 Ivanhoe B 52aFx64 +14 +37/=160/-28 52.00% 117.0/225
2 IvanHoe T0.4 x64 final -14 +28/=160/-37 48.00% 108.0/225



And third one no Robbobases again,but i changed "pawnmatval" to 100 instead of 110 in third test.110 seems to cause a weakness in my compile when playing without robbobases,and perhaps even when playing with it.


Third test WITHOUT robbobases again




1 Ivanhoe B 52aFx64 +18 +31/=137/-21 52.65% 99.5/189
2 IvanHoe T0.4 x64 final -18 +21/=137/-31 47.35% 89.5/189

In all 2 tests Ivanhoe T 0.4 caused 2 "no move" bugs each,and in 1 test 1(5 in total)I did not add that to the score or change the score accordingly


So i have made a new compile just change "pawnmatval" in engine preferences to 100 instead of the dubious 110.


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Angels_77 just now~~

We do think 1 minute tests without increment are is a poor way to test engines viability or how strong it truly is as engines do not reach great depth but do (as stated before ) think testing WITHOUT Bases is best and welcome Peter Pans change of testing

ps

in our tests of several 100 1+1 games (averaging some 5 mins per game ) B 52aF vrs TO.4 Final B52aF (as originaly compiled )is easily the stronger of the two engines

@ BOT7d ...:D

As you can see by our own test above peter pans B 52aF allthough unstable aquits itself well vrs ahmeds follow up to Final

We now run short test with said change of material values and see if it is any different a result. but PLEASE test yourselves AND post results

"B" wrote:

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

Short test to judge speed of ahameds latest compile. which is on a par woth PPs , more games realy needed to state which, is strongest but please test for yourselves


Test was.....
No Books. No Bases No Large Pages Ponder OFF

And here below a short test with material pawn value at 100

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

so NO difference to result at all realy 1 elo

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

this may just prove that if one throws out enough engines eventualy one will be good

But then again...
Sandi posts...

Skulltrail Intel QX 9775 Yorkfield (8X4090 MHz).
Time 3'+0" . Hash 256 . Ponder off. GUI Fritz 12. Tablebases 3-4-5 + Robbo Triple Bases. Book: HS-7moves plus.ctg.

Deep Rybka 4 - IvanHoe T0.4.3 x64 :
52.5 - 47.5 . 52.5% - 47.5% ( +22 -17 = 61 )

and as T0.4.1 beats DR4 in Sandis tests and T0.4.3 beats T0.4.1 in our tests....

Noobster wrote:

Code:

    Program                          Elo    +  -  Games  Score  Av.Op.  Draws

  1 Ivanhoe B 52aFx64p            : 3007    6  6  8915    52.0 %  2993  39.3 %
  2 IvanHoe T0.4.3                : 2994    8  8  4710    48.2 %  3007  39.0 %
  3 IvanHoehUS                    : 2992    8  8  4205    47.8 %  3007  39.5 %

By the amount of games played by ham heads these games must be " ultra fast " engines dont go to any depth at ultra fast speeds so such tests prove absolutely nothing, but "whatever rocks your boat "

no offence intended but whats the point of such tests

"B" wrote:

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

this may just prove that if one throws out enough engines eventualy one will be good

i had them about equal in longer test and his time managment is no better than either , hes a pillock as anyone with half a brain would have released one engine, a combination of both but he thinks "quantity makes quality sooner or later "
all hes done with time management is alter some of the paramiters like "time easy " then hide them so no one can see how simple it is, at least peter pan and vlado are honest and allow anyone to alter paramiters unlike this pillock ham head
self obsessed pillock so he is

Noobster wrote:

Doppelganger wrote:

Noobster wrote:

Doppelganger wrote:

Noobster wrote:

Code:

    Program                          Elo    +  -  Games  Score  Av.Op.  Draws

  1 Ivanhoe B 52aFx64p            : 3007    6  6  8915    52.0 %  2993  39.3 %
  2 IvanHoe T0.4.3                : 2994    8  8  4710    48.2 %  3007  39.0 %
  3 IvanHoehUS                    : 2992    8  8  4205    47.8 %  3007  39.5 %

By the amount of games played by ham heads these games must be " ultra fast " engines dont go to any depth at ultra fast speeds so such tests prove absolutely nothing, but "whatever rocks your boat "

no offence intended but whats the point of such tests

I found the way to save some games from this test and they are.. hm.. they ARE ok. Also, i tested a hamhead's T52E, which is 10 elo weaker tham hUS

ok for what exactly

in ultra fast games or even in short games without increment the engines dont have sufficent thinking time to reach very deep so are never realy tested and games from such tests arent even good for book makeing

I could say that Kevin seems to be using Monte-Carlo method for his book on main exit. And what i do here is Monte-Carlo :D

well keveins a complete asshole who thinks a thousand x 10 minutes is a a day , point is simple
#
at ultra fast time controls engines do NOT reach sufficent depth to give any true indication of there ability nor are such games good for makeing books intended for longer games
end of story

Noobster wrote:

Doppelganger wrote:

at ultra fast time controls engines do NOT reach sufficent depth to give any true indication of there ability nor are such games good for makeing books intended for longer games
end of story

If these are compilations of 1 engine (with some changes, nothing in timemanagement probably), why not?

This topic for tests NOT debates , both of you post in correct places please

Doppelganger wrote:

"B" wrote:

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

this may just prove that if one throws out enough engines eventualy one will be good

i had them about equal in longer test and his time managment is no better than either , hes a pillock as anyone with half a brain would have released one engine, a combination of both but he thinks "quantity makes quality sooner or later "
all hes done with time management is alter some of the paramiters like "time easy " then hide them so no one can see how simple it is, at least peter pan and vlado are honest and allow anyone to alter paramiters unlike this pillock ham head
self obsessed pillock so he is

just post test results here , and debates in other forum please