Similarity Move Test

Not logged inRybka Chess Community Forum

Forum RybkaChess Help Search Register Login

Topic The Rybka Lounge / Computer Chess / Similarity Move Test

1 2

By Sedat Canbaz (*****)

Date 2012-06-11 19:07 Edited 2012-06-11 20:07

Hello Dear Friends,

Nowadays there are a lot of clone/derivative/original work discussions over chess forums
That's why,this time i decided to create a different list,where all engines are played exactly at same position

Similarity Move Test Conditions:
----------------------------------
Processor:QX9650 @3.66GHz
OS:Win XP x64 Prof (service pack 2)
Chess GUIs:Fritz 12 / Arena 3.0
HashTable size:128 MB
Time Control:3 min game
Engine CPUs:All engines are played with 1 core
Before starting each test: engine's hastables are cleaned
Test Opening Position: B12 Caro-Kann,Advance Variation (Side to move:Blacks)

r2qkb1r/pp1nnpp1/2p1p2p/3pPb2/3P4/1N3N2/PPP1BPPP/R1BQ1RK1 b kq - 0 8

Some Notes:
---------------
-The current test is done mainly for recognizing which engines are each other relative
-The Similarity Test can be useful about which engines use same/similar ideas as well
-Similarity Move Test is not done for proving which engines are clone/derivative/original work
-Almost all Top Engines (Houdini;Komodo;Strelka;Stockfish,Critter,Ivanhoe... prefered 'Qc7' move
-Its interesting to note that Rybka 4.1 x64 1c prefered to play 'g5' move (not sure why ?!)
-Fruit 2.1 (open source engine) prefered especially to play 'Ng6'
-Note also 'Ng6' move is the most prefered move by many chess engines
-Zappa 1.0 and Fruit 2.0 are prefered to play 'a5' move (again similarity !)
-Sjeng c't 2010;Spike 1.4;BigLion 1.8f;Junior 6;Sos 5.1 played completely different moves than all

***************************************

59 Engine versions prefered 'Ng6' move

Amateur 2.82 - Ng6
Abrok 5.0 - Ng6
Ant 2006 - Ng6
Arasan 13.4 1c - Ng6
BeoWolf 2.4 - Ng6
Betsy 6.51 - Ng6
Bison 10.1 x64 - Ng6
Brutus 8.05 x64 - Ng6
CapaChess - Ng6
Chezzz 1.03 - Ng6
Comet B68 - Ng6
Crafty 23 CB 1c - Ng6
Cyrano 0.6b17 x64 - Ng6
Danasah 4.45 UCI - Ng6
Danasah 4.45 WB - Ng6
DayDreamer 1.75 x64 - Ng6
Deep Ftritz 13 1c - Ng6
Deep Sjeng 1.6 1c - Ng6
Delphil 26a - Ng6
Dirty 25082011 x64 - Ng6
Dragon 4.43 - Ng6
DrunkenMaster 1.2 - Ng6
EXchess 6.10b x64 - Ng6
Franchesca 0.19 - Ng6
Fruit 2.1 - Ng6
GES 1.36 - Ng6
Gull 1.2 x64 - Ng6
GreKo 9.2 - Ng6
Gromit 3.1 CB - Ng6
GnuChess4TM - Ng6
GnuChess 6.0.2 - Ng6
Glaurung 1.0.1 - Ng6
Jonny 2.7 - Ng6
Junior 7 - Ng6
Junior 8 - Ng6
Hiarcs 7.32 - Ng6
Hiarcs 9 - Ng6
Hiarcs 13.2 - Ng6
Ktulu 5.1 - Ng6
Ktulu 9 - Ng6
LambChop 10.99 - Ng6
LittleThought 052 - Ng6
Loop 2007 x64 - Ng6
Movei 0.08.438 1c - Ng6
Naum 4.2 x64 1c - Ng6
Onno 1.2.70 x64 1c - Ng6
Phalanx XXII - Ng6
Pharaon 3.5.1 1c - Ng6
Popochin 3.2 - Ng6
Shredder 9 - Ng6
Spike 0.9a - Ng6
Supra 101 - Ng6
Toga II 1.2.1 - Ng6
Toga II 1.4 beta5c - Ng6
Tornado 4.88 x64 1c - Ng6
Ufirm 8.02 - Ng6
Umko 1.2 x64 1c - Ng6
Zappa Mexico II x64 1c - Ng6
Zarkov 6.50 - Ng6

***************************************

57 Engine versions prefered 'Qc7' move

Amy 0.87 - Qc7
Anmon 5.75 WB - Qc7
Aristarch 4.50 - Qc7
AliChess 4.20 - Qc7
BobCat 3.25 x64 - Qc7
Buzz 008 x64 - Qc7
Boot 5.1.0 - Qc7
Cerebro 3.0d - Qc7
Chess Tiger 14 - Qc7
Chess Tiger 15 - Qc7
Chess Tiger 2007.1 - Qc7
Critter 0.8 x64 1c - Qc7
Critter 1.4 x64 1c - Qc7
Chronos 1.99 x64 - Qc7
Deep Junior 13 1c - Qc7
Deuterium 110229107 - Qc7
ETchess 130108 - Qc7
El Cinito 3.25 - Qc7
Elf 1.30 - Qc7
EXchess 4.03 - Qc7
Frenzee 3.5.19 - Qc7
FreeChess 2.17b - Qc7
Fritz in Bahrein - Qc7
Gandalf 5.0 - Qc7
Fire 2.2 xTreme x64 1c - Qc7
Fairy-Max v4.8 - Qc7
Fritz 5.32 - Qc7
Fritz 6 - Qc7
Fritz 7 - Qc7
Fruit 1.0 - Qc7
Fruit 090705 x64 1c - Qc7
Gambit Tiger 2.0- Qc7
GreKo 2.817 - Qc7
Ivanhoe B46fE 1c - Qc7
Hannibal 1.1 x64 - Qc7
Houdini 2.0c - Qc7
Igorrit 0086v9 x64 - Qc7
Insomniac 0.69 - Qc7
Komodo 4 x64 - Qc7
MinkoChess 1.3 x64 1c - Qc7
Nejmet 3.06 - Qc7
Nemo 1.01b x64 - Qc7
Philou 3.70 x64 - Qc7
Pseudo 0.7 - Qc7
Quazar 0.4 x64 - Qc7
Rybka 1.0b x64 1c - Qc7
RobboLito 010 - Qc7
Rufian 2.0.2 - Qc7
Rufian 2.1.0 - Qc7
SOS CB - Qc7
Strelka 5.1 x64 - Qc7
Strelka 5.5 x64 - Qc7
Stockfish 2.2.2 x64 1c - Qc7
TSCP 1.81 - Qc7
Thinker 5.4Di x64 1c - Qc7
Xpdnt 061120 - Qc7
Yace 0.99.87 - Qc7
Vitrivius 1.0C - Qc7

***************************************

25 Engine versions prefered 'Nc8' move

Anatoli 035k - Nc8
Amyan 1.59 - Nc8
Amyan 1.72 - Nc8
Arion 1.7 - Nc8
AliBaba 1.18 - Nc8
Bionic 4.01 - Nc8
Cheng3 1.07 - Nc8
Counter 1.2 - Nc8
Colossus 2008b - Nc8
Crafty 19.19 1c - Nc8
Delfi 5.4 1c - Nc8
Faile 1.44 CB - Nc8
Hagrid 0.7.56 - Nc8
Hamsters 0.8.2 1c - Nc8
Icarus 0.18 CB - Nc8
Leila 052f - Nc8
List 512 - Nc8
Mizar 1.0 - Nc8
Nimzo 8 - Nc8
Pepito 1.59 - Nc8
Predator 2.21 x64 - Nc8
Rotor 0.5 - Nc8
Trace 1.37a - Nc8
WildCat 4 WB - Nc8
WildCat 8 UCI - Nc8

***************************************

24 Engine versions prefered 'a5' move

Ayito 02994 - a5
Bugchess2 1.9 x64 - a5
Cheese 1.3 - a5
Chiron 1.0a x64 1c - a5
Gaviota 0.84 x64 - a5
Deep Shredder 12 x64 1c - a5
Equinox 1.10 x64 1c - a5
Fritz 8 - a5
Fruit 2.0 - a5
Goliat Light 1.5 - a5
Kiwi 0.6d x64 - a5
King of Kings 2.57 - a5
Matacz14 HT74 - a5
Quark 2.35 - a5
Pawny 0.2 x64 - a5
Patzer 3.11a - a5
Protector 1.4.0 x64 1c - a5
Prodeo 1.0 - a5
ProDeo 1.74 - a5
SlowChess Blitz 2.1 1c - a5
SmarThink 017a - a5
SmarThink 1.20 x64 - a5
Spark 1.0 x64 1c - a5
Zappa 1.0 - a5

***************************************

21 Engine versions prefered 'g5' move

Anaconda 2.2.2 - g5
Crafty 23.3 x64 1c - g5
Crafty 18.10 CB - g5
ElTurco 0.93 - g5
Eeyore 1.48 - g5
Jonny 4 x64 - g5
Gandalf 6.0 - g5
Natwarlal 0.14 - g5
Naum 1.8 - g5
Patriot 1.3.0 - g5
RedQueen 1.0.0 x64 - g5
Rybka v2.1c - g5
Rybka 4.1 x64 1c - g5
Sjeng WC2008 x64 1c - g5
Scorpio 2.7 x64 1c - g5
Testina 2.2 - g5
Texel 101 x64 - g5
The Baron 2.2.2 - g5
The Baron 2.2.3 - g5
The King 3.33 - g5
The King 3.50 x64 1c - g5
Thinker 4.7a - g5

***************************************
3 Engine versions prefered 'Rc8' move

GarboChess 3 x64 - Rc8
Litle Goliath Revival 3.11 - Rc8
RomiChess P3k x64 - Rc8

***************************************

3 Engine versions prefered 'Bh7' move

Tao 5.7 - Bh7
The Crazy Bishop CB - Bh7
The Crazy Bishop 052 - Bh7

***************************************

2 Engine versions prefered 'Bg4' move

Gaia 3.5 x64 - Bg4
LordKing 2006 Serafinus - Bg4

***************************************

1 Engine version prefered 'b6' move

Sjeng c't 2010 - b6

***************************************
1 Engine version prefered 'Bg6' move

Spike 1.4 1c - Bg6

***************************************

1 Engine version prefered 'Be4' move

Sos 5.1 For Arena - Be4
***************************************

1 Engine version prefered 'f6' move

BigLion 1.8f - f6
***************************************

1 Engine version prefered 'Rg8' move

Junior 6 - Rg8

***************************************

Best Regards,
Sedat Canbaz

By siah

Date 2012-06-12 06:06

By Pia

Date 2012-06-12 07:18

How many positions will be? Similary tester using thousands ^^

Also look at this

New game (128 MB hash)

Analysis by RobboLito 0.085g3 w32:

1.d2-d4 Ng8-f6 2.Nb1-c3 d7-d5 3.Bc1-f4 c7-c5 4.e2-e3 c5xd4 5.e3xd4 a7-a6 6.Ng1-f3 Nb8-c6 7.Bf1-e2 e7-e6 8.0-0 Bf8-d6 9.Qd1-d2 0-0 10.Nf3-e5 Qd8-c7 11.Qd2-e3 Bc8-d7 12.Ne5xd7 Bd6xf4 13.Nd7xf6+ g7xf6
= (0.13) Depth: 22/52 00:06:37 318mN

Analysis by Houdini 1.01 w32 1_CPU:

1.Nb1-c3 Ng8-f6 2.d2-d4 d7-d5 3.Bc1-f4 c7-c5 4.e2-e3 c5xd4 5.e3xd4 a7-a6 6.Ng1-f3 Nb8-c6 7.Bf1-e2 e7-e6 8.0-0 Bf8-d6 9.Qd1-d2 0-0 10.Nf3-e5 Qd8-b6 11.Ne5xc6 Qb6xc6 12.Bf4-e5 Bd6xe5 13.d4xe5
= (0.06) Depth: 23/52 00:05:23 313mN

By Sedat Canbaz (*****)

Date 2012-06-12 08:29 Edited 2012-06-12 08:38

>How many positions will be? Similary tester using thousands ^^

Sure similarity move test by thousands of positions and including more engines will be great...but sorry no time for all

What about you ?
Maybe you can run a such similarity test by thousands of positions and thousands of engine versions ?!
Big thanks in advance, if you will do a such test...

Btw,I have already discussed about this issue on Talkchess forum:
http://www.talkchess.com/forum/viewtopic.php?t=43946&postdays=0&postorder=asc&topic_view=&start=10

One thing more,(in my testings) it seems that more than 80% of the engines are each other relative
Once more i noticed that its very hard to find 100% original work

To be honest,actually i am concentrated more on the engines Elo strenght (not especially on copying code...)
I hope one day (in official tournaments)there will be a new rule :
-Any engine will be allowed,in case of +50 Elo improvement (which is based or have ideas from free open source engine)

Best Regards,
Sedat

By Pia

Date 2012-06-12 10:05

>Maybe you can run a such similarity test by thousands of positions and thousands of engine versions ?!

Yes! How?
Similarity tester version 3 works only with UCI engines.
I wish it also send command to engine to clear hash before feeding new position to test.
And also possibility to set # of CPU cores (to 1).
I've extracted 8238 test positions in case been succeed to write different tester,
but have no clue how program should communicate with engine :cry:

By Sedat Canbaz (*****)

Date 2012-06-12 10:27 Edited 2012-06-12 10:41

>but have no clue how program should communicate with engine

Yes...this similarity tool is too complicated for me as well
And its will be great,if its will be capable for all engines (WB/UCI/CB...) and easier for public use

But however,i think the programmers need to concentrate more how to improve the strength of chess engines,instead of clone issues

Btw,Rybka was unbeatable almost 6 years (a great record),e.g the new king is now Houdini (more than 1 year tops in all ratings)
Really i have no patience to see serious opponents against this monster

Greetings,
Sedat

By Sedat Canbaz (*****)

Date 2012-06-13 23:01

>Yes! How?

Btw,actually there is easier way and its possible to include many positions by using Chess GUIs

For this, we need to create a engine tournament with hundreds of starting positions
And for not wasting many CPU time, we need to adjudicate manually the played moves

Let's see,maybe later i can test some engines (with more positions),but however i don't want to promise...

Best,
Sedat

By Adam Hair (**)

Date 2012-06-14 03:47

Hi Pia,

I forgot to respond to this.

1) I use Wb2Uci to use WB engines with the similarity tester. Though, it should not be much trouble to add the correct WB commands to the similarity tester. I just have not taken the time to figure out what to do.

2) If you have extracted the positions, then you will be able to add the command for clearing hash. Open the lib folder, then the app-sim folder. Open sim.tcl with a text editor such as Notepad ++. Find this line: puts $fh "position startpos moves [index $all $n]". Insert this line just above it: puts $fh "setoption name Clear Hash". Make sure the indentation does not change.

3) To make each engine use 1 cpu, you must use config files. If you double click on a UCI engine, and then type "UCI" in the console window, the list of UCI commands that the engine understands will be listed. You have to find out how each engine specifies how many cpus to use (Threads or Max CPUs or etcetera). Also, see if there are any other options you want to set, such as hash. Here is an example for Houdini:

exe = Houdini_2.0c_64-bit.exe
name=Houdini 2.0c 64-bit
scale = 1.0
Hash = 64
Threads = 1

Then, instead of using sim03w64.exe -t [engine].exe [time], use sim03w64.exe -c [engine].txt [time] .

By Pia

Date 2012-06-14 21:11

Thanks, all worked out. Also using fixed depth instead of interrupting by time results up to 100.00 percents similarity between two copies of same engine.
Here is quick sample
Houdini 1.01 has more in common with RobboLito than Houdini 1.5a itself :grin:

sim version 3

  Key:

  1) Houdini 1.01 w32 1_CPU (time: 100 ms  scale: 1.0)
  2) Houdini_15a_w32 (time: 100 ms  scale: 1.0)
  3) Houdini_15a_w32_copy (time: 100 ms  scale: 1.0)
  4) RobboLito 0.085g3 w32 (time: 100 ms  scale: 1.0)
  5) RobboLito 0.09 w32 (time: 100 ms  scale: 1.0)
  6) RobboLito_0085g3_w32_copy (time: 100 ms  scale: 1.0)

         1     2     3     4     5     6
  1.  ----- 75.75 75.75 88.31 88.30 88.31
  2.  75.75 -----100.00 73.31 73.33 73.31
  3.  75.75100.00 ----- 73.31 73.33 73.31
  4.  88.31 73.31 73.31 ----- 99.76100.00
  5.  88.30 73.33 73.33 99.76 ----- 99.76
  6.  88.31 73.31 73.31100.00 99.76 -----

By Sedat Canbaz (*****)

Date 2012-06-14 21:57 Edited 2012-06-14 22:14

>Houdini 1.01 has more in common with RobboLito than Houdini 1.5a itself

Thanks for your useful measuring

Yes...we see that the percentage of matched moves is too high,but anyway there is one BIG difference:
-Houdini's playing Elo strength (in my opinion,this is the most important :wink:

)

One thing more,it seems the author of Houdini (Robert Houdart) is a honest programer ...

http://www.cruxis.com/chess/houdini.htm
Without many ideas and techniques from the open source chess engines Ippolit and Stockfish, Houdini would not nearly be as strong as it is now.

Btw,its will be great,if you add more engines in your measuring test
For example:Critter;Komodo;Strelka;Rybka;Stockfish;Ivanhoe;Fire;Vitruvius..

Best,
Sedat

By keoki010 (Silver)

Date 2012-06-14 23:09

Yes, despite all the hoopla I think Houdart is honest and trying to improve the engine. :cool:

By RFK

Date 2012-06-18 22:05 Edited 2012-06-18 22:08

>Yes, despite all the hoopla I think Houdart is honest and trying to improve the engine.

+1

And taken a great deal of abuse for it.

By Pia

Date 2012-06-15 03:12

>Btw,its will be great,if you add more engines in your measuring test

That's my intention, of course.
Here's a quick sample.
I will add more engines to the list in a few days.

sim version 3 pia
------ Critter 1.2 32-bit (depth: 2, scale: 1.0) ------
79.19 Critter 1.0 32-bit (depth: 2, scale: 1.0)
77.54 Strelka 5.5 x32 (depth: 2, scale: 1.0)
77.54 Strelka 5.1 (depth: 2, scale: 1.0)
77.54 Strelka 5 (depth: 2, scale: 1.0)
75.52 Houdini_15a_w32_copy (depth: 2, scale: 1.0)
75.52 Houdini_15a_w32 (depth: 2, scale: 1.0)
75.09 Houdini 1.01 w32 1_CPU (depth: 2, scale: 1.0)
73.50 Houdini 2.0c w32 (depth: 2, scale: 1.0)
73.40 RobboLito_0085g3_w32_copy (depth: 2, scale: 1.0)
73.40 RobboLito 0.085g3 w32 (depth: 2, scale: 1.0)
73.39 RobboLito 0.09 w32 (depth: 2, scale: 1.0)
72.97 IPPOLIT 0.080b w32 (depth: 2, scale: 1.0)
71.97 IvanHoe 9.47b w32 (depth: 2, scale: 1.0)
71.79 Fire 1.31 w32 (depth: 2, scale: 1.0)
71.56 Fire 2.2 xTreme w32 (depth: 2, scale: 1.0)
71.40 Fire 1.5 xTreme w32 (depth: 2, scale: 1.0)
60.94 Critter 0.70 32-bit (depth: 2, scale: 1.0)
55.61 Strelka 2.0 B (depth: 2, scale: 1.0)
55.28 Stockfish 1.8 JA (depth: 2, scale: 1.0)
55.12 Critter 0.52b 32-bit (depth: 2, scale: 1.0)
54.81 Stockfish 2.1.1 JA (depth: 2, scale: 1.0)
54.45 Strelka 1.0 Beta (depth: 2, scale: 1.0)
54.44 Strelka 1.8 UCI (depth: 2, scale: 1.0)
54.37 Fruit 2.1 (depth: 2, scale: 1.0)
54.19 Gull 1.0 (depth: 2, scale: 1.0)
51.38 Stockfish 2.2.2 JA (depth: 3, scale: 1.0)
50.10 Deep Rybka 4.1_w32 (depth: -1, scale: 1.0)
50.02 Rybka 3 w32_copy (depth: 2, scale: 1.0)
50.02 Rybka 3 w32 (depth: 2, scale: 1.0)
49.88 Alaric 707 (depth: 2, scale: 1.0)
49.79 Komodo32 3 32bit (depth: 2, scale: 1.0)
49.78 Deep Rybka 4.1_w32 (depth: 1, scale: 1.0)
49.77 Deep Rybka 4.1_w32 (depth: 0, scale: 1.0)
49.65 Komodo32 Version 4 (depth: 2, scale: 1.0)
48.74 Atlas 3.14b (depth: 2, scale: 1.0)
47.60 Rybka WinFinder 1.0 32-bit (depth: 2, scale: 1.0)
47.44 bright-0.2c (depth: 2, scale: 1.0)
47.09 Rybka 1.2 32-bit (depth: 2, scale: 1.0)
47.07 Rybka 1.0 Beta 32-bit (depth: 2, scale: 1.0)
46.98 Rybka 2.3 32-bit (depth: 2, scale: 1.0)
46.92 Rybka 2.1 32-bit (depth: 2, scale: 1.0)
46.89 Deep Rybka 4.1_w32 (depth: 2, scale: 1.0)
46.88 Rybka 2.3.2a 32-bit (depth: 2, scale: 1.0)
46.83 Rybka 2.3.1 32-bit (depth: 2, scale: 1.0)
45.81 Rybka 1.1 32-bit (depth: 2, scale: 1.0)
45.38 Alfil 8.1.1 (depth: 2, scale: 1.0)
43.52 Ares 1.004 (depth: 2, scale: 1.0)
42.28 Amyan 1.72 (depth: 2, scale: 1.0)
41.18 Aristarch 4.50 (depth: 2, scale: 1.0)
39.80 Alex v2.14_win32 (depth: 2, scale: 1.0)
37.30 AliChess 4.25 (depth: 2, scale: 1.0)
34.51 Adam 3.3 (depth: 2, scale: 1.0)
28.77 ALChess v1.5b (depth: 2, scale: 1.0)

Entire data could not fit in here.
For details, attached similarity.data file should be readable by original Similarity tester v. 3 tool from http://komodochess.com/

Attachment: similarity.rar (300k)

By Barnard (Bronze)

Date 2012-06-15 04:01

Hi pia

in your 2 posts,that similarity is related,to what engine each post?

By Barnard (Bronze)

Date 2012-06-15 04:03

i answered myself :red:

is in the header of each test...

By Barnard (Bronze)

Date 2012-06-15 05:21

pia,can you test rybka 1.0?the first public released version

thanks in advance

By Pia

Date 2012-06-15 12:45

sim version 3 pia
------ Rybka 1.0 Beta 32-bit (depth: 2, scale: 1.0) ------
95.57 Rybka WinFinder 1.0 32-bit (depth: 2, scale: 1.0)
66.70 Rybka 1.1 32-bit (depth: 2, scale: 1.0)
64.53 Rybka 1.2 32-bit (depth: 2, scale: 1.0)
64.17 Rybka 2.3 32-bit (depth: 2, scale: 1.0)
64.07 Rybka 2.3.1 32-bit (depth: 2, scale: 1.0)
63.63 Rybka 2.1 32-bit (depth: 2, scale: 1.0)
62.03 Rybka 2.3.2a 32-bit (depth: 2, scale: 1.0)
52.48 Strelka 1.8 UCI (depth: 2, scale: 1.0)
52.38 Strelka 1.0 Beta (depth: 2, scale: 1.0)
52.26 Rybka 3 w32_copy (depth: 2, scale: 1.0)
52.26 Rybka 3 w32 (depth: 2, scale: 1.0)
51.27 Strelka 2.0 B (depth: 2, scale: 1.0)
50.76 Deep Rybka 4.1_w32 (depth: -1, scale: 1.0)
50.56 Deep Rybka 4.1_w32 (depth: 1, scale: 1.0)
50.40 Deep Rybka 4.1_w32 (depth: 0, scale: 1.0)
50.24 Deep Rybka 4.1_w32 (depth: 2, scale: 1.0)
47.50 Stockfish 2.2.2 JA (depth: 3, scale: 1.0)
47.16 Critter 1.0 32-bit (depth: 2, scale: 1.0)
47.07 Critter 1.2 32-bit (depth: 2, scale: 1.0)
46.93 Fruit 2.1 (depth: 2, scale: 1.0)
46.90 RobboLito_0085g3_w32_copy (depth: 2, scale: 1.0)
46.90 RobboLito 0.085g3 w32 (depth: 2, scale: 1.0)
46.88 Vitruvius version 1.0C (depth: 2, scale: 1.0)
46.86 RobboLito 0.09 w32 (depth: 2, scale: 1.0)
46.77 IPPOLIT 0.080b w32 (depth: 2, scale: 1.0)
46.66 Fire 1.5 xTreme w32 (depth: 2, scale: 1.0)
46.48 Fire 2.2 xTreme w32 (depth: 2, scale: 1.0)
46.48 Fire 1.31 w32 (depth: 2, scale: 1.0)
46.46 IvanHoe 9.47b w32 (depth: 2, scale: 1.0)
46.27 Houdini 1.01 w32 1_CPU (depth: 2, scale: 1.0)
45.33 Gull 1.0 (depth: 2, scale: 1.0)
45.16 Houdini 2.0c w32 (depth: 2, scale: 1.0)
44.97 Strelka 5.5 x32 (depth: 2, scale: 1.0)
44.97 Strelka 5.1 (depth: 2, scale: 1.0)
44.97 Strelka 5 (depth: 2, scale: 1.0)
44.96 Alaric 707 (depth: 2, scale: 1.0)
44.65 Stockfish 1.8 JA (depth: 2, scale: 1.0)
44.63 Houdini_15a_w32_copy (depth: 2, scale: 1.0)
44.63 Houdini_15a_w32 (depth: 2, scale: 1.0)
44.34 Stockfish 2.1.1 JA (depth: 2, scale: 1.0)
43.69 Critter 0.70 32-bit (depth: 2, scale: 1.0)
43.52 Komodo32 3 32bit (depth: 2, scale: 1.0)
43.40 Komodo32 Version 4 (depth: 2, scale: 1.0)
41.89 bright-0.2c (depth: 2, scale: 1.0)
41.47 Atlas 3.14b (depth: 2, scale: 1.0)
40.54 Critter 0.52b 32-bit (depth: 2, scale: 1.0)
39.11 Alfil 8.1.1 (depth: 2, scale: 1.0)
38.55 Ares 1.004 (depth: 2, scale: 1.0)
36.82 Amyan 1.72 (depth: 2, scale: 1.0)
36.38 Adam 3.3 (depth: 2, scale: 1.0)
35.85 AliChess 4.25 (depth: 2, scale: 1.0)
34.55 Alex v2.14_win32 (depth: 2, scale: 1.0)
33.76 Aristarch 4.50 (depth: 2, scale: 1.0)
26.01 ALChess v1.5b (depth: 2, scale: 1.0)

By Labyrinth (*****)

Date 2012-06-15 13:44

Weird how the last 8 entries all start with 'A' :-P

By Pia

Date 2012-06-15 14:11

Later there will be also B, C, D, E... ...and Z :wink:

And then some more for A and Z.

By magnumpi (**)

Date 2012-06-15 20:55

Hello Pria, I hope you don't mind if I help you testing other engines :)
BTW thanks Adam, with your instructions everything went smooth

Here are the results of Naum 4.2 and Deep Onno 1.2.70

I've attached the similarity.data file

sim version 3
------ Naum 4.2 (depth: 2, scale: 1.0) ------
78.37 Strelka 2.0 B (depth: 2, scale: 1.0)
74.02 Strelka 1.8 UCI (depth: 2, scale: 1.0)
73.89 Strelka 1.0 Beta (depth: 2, scale: 1.0)
63.52 Fruit 2.1 (depth: 2, scale: 1.0)
61.14 Deep Onno 1-2-70 (depth: 2, scale: 1.0)
60.28 Alaric 707 (depth: 2, scale: 1.0)
58.41 Gull 1.0 (depth: 2, scale: 1.0)
57.10 RobboLito 0.09 w32 (depth: 2, scale: 1.0)
57.09 RobboLito_0085g3_w32_copy (depth: 2, scale: 1.0)
57.09 RobboLito 0.085g3 w32 (depth: 2, scale: 1.0)
56.91 IPPOLIT 0.080b w32 (depth: 2, scale: 1.0)
56.18 Critter 1.0 32-bit (depth: 2, scale: 1.0)
56.14 Critter 1.2 32-bit (depth: 2, scale: 1.0)
55.92 IvanHoe 9.47b w32 (depth: 2, scale: 1.0)
55.60 Houdini 1.01 w32 1_CPU (depth: 2, scale: 1.0)
55.60 Fire 2.2 xTreme w32 (depth: 2, scale: 1.0)
55.47 Houdini_15a_w32_copy (depth: 2, scale: 1.0)
55.47 Houdini_15a_w32 (depth: 2, scale: 1.0)
55.41 Fire 1.5 xTreme w32 (depth: 2, scale: 1.0)
55.27 Fire 1.31 w32 (depth: 2, scale: 1.0)
54.90 Stockfish 2.1.1 JA (depth: 2, scale: 1.0)
54.75 Houdini 2.0c w32 (depth: 2, scale: 1.0)
54.70 Strelka 5.5 x32 (depth: 2, scale: 1.0)
54.70 Strelka 5.1 (depth: 2, scale: 1.0)
54.70 Strelka 5 (depth: 2, scale: 1.0)
53.96 Stockfish 1.8 JA (depth: 2, scale: 1.0)
52.88 bright-0.2c (depth: 2, scale: 1.0)
51.98 Rybka 1.2 32-bit (depth: 2, scale: 1.0)
51.81 Critter 0.70 32-bit (depth: 2, scale: 1.0)
51.70 Rybka 2.1 32-bit (depth: 2, scale: 1.0)
51.30 Rybka 2.3 32-bit (depth: 2, scale: 1.0)
51.15 Rybka WinFinder 1.0 32-bit (depth: 2, scale: 1.0)
51.02 Atlas 3.14b (depth: 2, scale: 1.0)
50.78 Rybka 2.3.1 32-bit (depth: 2, scale: 1.0)
50.52 Stockfish 2.2.2 JA (depth: 3, scale: 1.0)
50.33 Rybka 1.0 Beta 32-bit (depth: 2, scale: 1.0)
50.28 Rybka 1.1 32-bit (depth: 2, scale: 1.0)
49.76 Rybka 2.3.2a 32-bit (depth: 2, scale: 1.0)
49.67 Critter 0.52b 32-bit (depth: 2, scale: 1.0)
49.30 Ares 1.004 (depth: 2, scale: 1.0)
48.83 Alfil 8.1.1 (depth: 2, scale: 1.0)
46.61 Deep Rybka 4.1_w32 (depth: -1, scale: 1.0)
46.58 Deep Rybka 4.1_w32 (depth: 0, scale: 1.0)
46.24 Deep Rybka 4.1_w32 (depth: 1, scale: 1.0)
44.91 Komodo32 3 32bit (depth: 2, scale: 1.0)
44.63 Rybka 3 w32_copy (depth: 2, scale: 1.0)
44.63 Rybka 3 w32 (depth: 2, scale: 1.0)
44.53 Komodo32 Version 4 (depth: 2, scale: 1.0)
43.36 Deep Rybka 4.1_w32 (depth: 2, scale: 1.0)
41.98 Amyan 1.72 (depth: 2, scale: 1.0)
41.96 Alex v2.14_win32 (depth: 2, scale: 1.0)
41.79 AliChess 4.25 (depth: 2, scale: 1.0)
41.70 Aristarch 4.50 (depth: 2, scale: 1.0)
33.24 Adam 3.3 (depth: 2, scale: 1.0)
29.47 ALChess v1.5b (depth: 2, scale: 1.0)

Attachment: naum42_deeponno1270.rar (25k)

By magnumpi (**)

Date 2012-06-15 20:57

sim version 3
------ Deep Onno 1-2-70 (depth: 2, scale: 1.0) ------
73.55 Fruit 2.1 (depth: 2, scale: 1.0)
62.28 Strelka 2.0 B (depth: 2, scale: 1.0)
61.14 Naum 4.2 (depth: 2, scale: 1.0)
60.68 Strelka 1.0 Beta (depth: 2, scale: 1.0)
60.55 Alaric 707 (depth: 2, scale: 1.0)
60.46 Strelka 1.8 UCI (depth: 2, scale: 1.0)
55.74 Gull 1.0 (depth: 2, scale: 1.0)
54.70 Stockfish 2.1.1 JA (depth: 2, scale: 1.0)
54.61 Stockfish 1.8 JA (depth: 2, scale: 1.0)
54.18 bright-0.2c (depth: 2, scale: 1.0)
53.86 Critter 1.0 32-bit (depth: 2, scale: 1.0)
53.42 Critter 1.2 32-bit (depth: 2, scale: 1.0)
53.39 Atlas 3.14b (depth: 2, scale: 1.0)
53.20 IPPOLIT 0.080b w32 (depth: 2, scale: 1.0)
52.83 IvanHoe 9.47b w32 (depth: 2, scale: 1.0)
52.78 Houdini_15a_w32_copy (depth: 2, scale: 1.0)
52.78 Houdini_15a_w32 (depth: 2, scale: 1.0)
52.72 RobboLito 0.09 w32 (depth: 2, scale: 1.0)
52.69 RobboLito_0085g3_w32_copy (depth: 2, scale: 1.0)
52.69 RobboLito 0.085g3 w32 (depth: 2, scale: 1.0)
52.54 Fire 1.5 xTreme w32 (depth: 2, scale: 1.0)
52.46 Fire 2.2 xTreme w32 (depth: 2, scale: 1.0)
52.33 Fire 1.31 w32 (depth: 2, scale: 1.0)
52.23 Strelka 5.5 x32 (depth: 2, scale: 1.0)
52.23 Strelka 5.1 (depth: 2, scale: 1.0)
52.23 Strelka 5 (depth: 2, scale: 1.0)
52.11 Houdini 1.01 w32 1_CPU (depth: 2, scale: 1.0)
51.89 Houdini 2.0c w32 (depth: 2, scale: 1.0)
51.74 Critter 0.70 32-bit (depth: 2, scale: 1.0)
51.61 Critter 0.52b 32-bit (depth: 2, scale: 1.0)
51.51 Stockfish 2.2.2 JA (depth: 3, scale: 1.0)
49.90 Alfil 8.1.1 (depth: 2, scale: 1.0)
47.90 Rybka WinFinder 1.0 32-bit (depth: 2, scale: 1.0)
47.89 Rybka 1.2 32-bit (depth: 2, scale: 1.0)
47.58 Rybka 1.0 Beta 32-bit (depth: 2, scale: 1.0)
47.54 Rybka 2.1 32-bit (depth: 2, scale: 1.0)
47.44 Rybka 2.3 32-bit (depth: 2, scale: 1.0)
47.37 Rybka 2.3.1 32-bit (depth: 2, scale: 1.0)
46.84 Rybka 1.1 32-bit (depth: 2, scale: 1.0)
46.65 Ares 1.004 (depth: 2, scale: 1.0)
46.64 Rybka 2.3.2a 32-bit (depth: 2, scale: 1.0)
45.36 Deep Rybka 4.1_w32 (depth: -1, scale: 1.0)
45.17 Deep Rybka 4.1_w32 (depth: 0, scale: 1.0)
45.13 Deep Rybka 4.1_w32 (depth: 1, scale: 1.0)
44.87 Komodo32 3 32bit (depth: 2, scale: 1.0)
44.10 Komodo32 Version 4 (depth: 2, scale: 1.0)
43.15 Amyan 1.72 (depth: 2, scale: 1.0)
42.80 Rybka 3 w32_copy (depth: 2, scale: 1.0)
42.80 Rybka 3 w32 (depth: 2, scale: 1.0)
42.55 Alex v2.14_win32 (depth: 2, scale: 1.0)
42.24 AliChess 4.25 (depth: 2, scale: 1.0)
42.19 Deep Rybka 4.1_w32 (depth: 2, scale: 1.0)
40.87 Aristarch 4.50 (depth: 2, scale: 1.0)
34.10 Adam 3.3 (depth: 2, scale: 1.0)
29.90 ALChess v1.5b (depth: 2, scale: 1.0)

By Pia

Date 2012-06-15 21:27

All correct. Try to match my speed of batch processing:

@cmd.exe /c for %%a in (*.exe) do call test.bat -t %%a

I'm also using "ucinewgame" command from Richard Vida =^_^=

By Banned for Life (Gold)

Date 2012-06-15 15:22

From this, you can see why the ICGA wouldn't be going down the similarity path:

Rybka 1.0 Beta 32-bit - Fruit 2.1 46.93
Rybka 1.0 Beta 32-bit - Stockfish 2.2.2 47.5

By Barnard (Bronze)

Date 2012-06-15 17:26

That is what i suspected Alan

By Adam Hair (**)

Date 2012-06-15 17:48

Not to start anything (oh no! :smile:

), but what depth does Rybka 1.0 Beta search when given the command 'go depth 2'? In Pia's test, each engine is sent the command to search 2 plies deep. I know that Fruit 2.1 does search only 2 plies when sent that command. It was my understanding that Rybka 1.0 Beta misrepresents the depth searched by 3(?) plies. So the comparison here is Fruit at 2 plies to Rybka at 5(?) plies.

By Banned for Life (Gold)

Date 2012-06-15 18:37

It might have been +2 in depth back then...

By Pia

Date 2012-06-15 20:51

Affirmative, Rybka searches 2 plies deeper. The closer depth, the higher match.

sim version 3 pia

  Key:

  1) Fruit 2.1 (depth: 0, scale: 1.0)
  2) Fruit 2.1 (depth: 1, scale: 1.0)
  3) Fruit 2.1 (depth: 2, scale: 1.0)
  4) Fruit 2.1 (depth: 3, scale: 1.0)
  5) Fruit 2.1 (depth: 4, scale: 1.0)
  6) Fruit 2.1 (depth: 5, scale: 1.0)
  7) Fruit 2.1 (depth: 6, scale: 1.0)

  8) Rybka 1.0 Beta 32-bit (depth: -1, scale: 1.0)
  9) Rybka 1.0 Beta 32-bit (depth: -2, scale: 1.0)
 10) Rybka 1.0 Beta 32-bit (depth: 0, scale: 1.0)
 11) Rybka 1.0 Beta 32-bit (depth: 1, scale: 1.0)
 12) Rybka 1.0 Beta 32-bit (depth: 2, scale: 1.0)
 13) Rybka 1.0 Beta 32-bit (depth: 3, scale: 1.0)
 14) Rybka 1.0 Beta 32-bit (depth: 4, scale: 1.0)
 15) Rybka 1.0 Beta 32-bit (depth: 5, scale: 1.0)

         1     2     3     4     5     6     7     8     9    10    11    12    13    14    15
  1.  -----100.00 61.48 49.21 44.17 41.11 37.45 42.24 42.24 42.24 42.24 39.34 37.00 35.03 32.84
  2. 100.00 ----- 61.48 49.21 44.17 41.11 37.45 42.24 42.24 42.24 42.24 39.34 37.00 35.03 32.84
  3.  61.48 61.48 ----- 61.25 53.76 48.17 43.29 52.00 52.00 52.00 52.00 46.97 43.57 40.24 36.94
  4.  49.21 49.21 61.25 ----- 67.36 59.27 51.89 61.34 61.34 61.34 61.34 54.70 50.56 46.81 43.55 11/-61.34%,
  5.  44.17 44.17 53.76 67.36 ----- 72.14 61.62 53.14 53.14 53.14 53.14 60.40 56.31 52.38 48.64 12/-60.40%,
  6.  41.11 41.11 48.17 59.27 72.14 ----- 73.49 48.79 48.79 48.79 48.79 54.18 58.13 55.84 52.93 13/-58.13%,
  7.  37.45 37.45 43.29 51.89 61.62 73.49 ----- 45.24 45.24 45.24 45.24 50.36 54.05 57.61 56.11 14/-57.61% are best matches

  8.  42.24 42.24 52.00 61.34 53.14 48.79 45.24 -----100.00100.00100.00 69.11 61.80 54.81 50.07
  9.  42.24 42.24 52.00 61.34 53.14 48.79 45.24100.00 -----100.00100.00 69.11 61.80 54.81 50.07
 10.  42.24 42.24 52.00 61.34 53.14 48.79 45.24100.00100.00 -----100.00 69.11 61.80 54.81 50.07
 11.  42.24 42.24 52.00 61.34 53.14 48.79 45.24100.00100.00100.00 ----- 69.11 61.80 54.81 50.07
 12.  39.34 39.34 46.97 54.70 60.40 54.18 50.36 69.11 69.11 69.11 69.11 ----- 74.46 63.77 56.97
 13.  37.00 37.00 43.57 50.56 56.31 58.13 54.05 61.80 61.80 61.80 61.80 74.46 ----- 74.95 65.96
 14.  35.03 35.03 40.24 46.81 52.38 55.84 57.61 54.81 54.81 54.81 54.81 63.77 74.95 ----- 78.19
 15.  32.84 32.84 36.94 43.55 48.64 52.93 56.11 50.07 50.07 50.07 50.07 56.97 65.96 78.19 -----

By Pia

Date 2012-06-17 03:36

>Not to start anything (oh no! )

But why not? This new test can be sometimes looking bad? :twisted:

sim version 3 pia
------ Fruit 2.2.1 (depth: 2, scale: 1.0) ------
95.98 Loop 13.6 (depth: 2, scale: 1.0)
95.97 LoopMP 12.32 (depth: 2, scale: 1.0)
80.76 Fruit 2.1 (depth: 2, scale: 1.0)
76.32 Gambit Fruit 1.0 Beta 4bx (depth: 2, scale: 1.0)
76.30 Gambit Fruit 1.0 Beta 4bx JA (depth: 2, scale: 1.0)

...etc

By Adam Hair (**)

Date 2012-06-17 15:03

Very bad

It looks almost as if the only real differences between Loop and Fruit 2.2.1 is in the search. Virtually none in the evaluation, if I correctly understand how engines work. The similarity decreases some at higher average depths.

Adam

By Sedat Canbaz (*****)

Date 2012-06-17 06:21

>So the comparison here is Fruit at 2 plies to Rybka at 5(?) plies.

Hello Adam,

In other words,its another clear proof that Rybka is not clone of Fruit
If Rybka was a clone,then the both engines's depth would be same (2 plies), right ?

Best,
Sedat

By Adam Hair (**)

Date 2012-06-17 15:11

Sedat,

Anybody (including myself if I did this) who ever claimed that Vas cloned Fruit were mistaken.

The situation with Rybka and Fruit is not clear cut. One thing for certain, the similarity test can not show if Rybka is a derivative of Fruit, nor can it show that it is not a derivative.

Adam

By Barnard (Bronze)

Date 2012-06-15 17:26

Thanks pia,interesting results

By Pia

Date 2012-06-15 03:55

Oh, and just for the mysterious Vitruvius quick update

sim version 3 pia
------ Vitruvius version 1.0C (depth: 2, scale: 1.0) ------
88.52 IvanHoe 9.47b w32 (depth: 2, scale: 1.0)
86.67 Fire 2.2 xTreme w32 (depth: 2, scale: 1.0)
85.36 Fire 1.5 xTreme w32 (depth: 2, scale: 1.0)
79.06 RobboLito_0085g3_w32_copy (depth: 2, scale: 1.0)
79.06 RobboLito 0.085g3 w32 (depth: 2, scale: 1.0)
79.04 RobboLito 0.09 w32 (depth: 2, scale: 1.0)
77.83 IPPOLIT 0.080b w32 (depth: 2, scale: 1.0)
77.35 Houdini 1.01 w32 1_CPU (depth: 2, scale: 1.0)
77.31 Fire 1.31 w32 (depth: 2, scale: 1.0)
71.21 Critter 1.2 32-bit (depth: 2, scale: 1.0)
71.13 Strelka 5.5 x32 (depth: 2, scale: 1.0)
71.13 Strelka 5.1 (depth: 2, scale: 1.0)
71.13 Strelka 5 (depth: 2, scale: 1.0)
69.81 Houdini_15a_w32_copy (depth: 2, scale: 1.0)
69.81 Houdini_15a_w32 (depth: 2, scale: 1.0)
68.45 Houdini 2.0c w32 (depth: 2, scale: 1.0)
68.34 Critter 1.0 32-bit (depth: 2, scale: 1.0)
59.16 Critter 0.70 32-bit (depth: 2, scale: 1.0)
54.48 Strelka 2.0 B (depth: 2, scale: 1.0)
53.96 Gull 1.0 (depth: 2, scale: 1.0)
53.63 Critter 0.52b 32-bit (depth: 2, scale: 1.0)
53.31 Strelka 1.8 UCI (depth: 2, scale: 1.0)
53.11 Strelka 1.0 Beta (depth: 2, scale: 1.0)
52.72 Fruit 2.1 (depth: 2, scale: 1.0)
52.39 Stockfish 2.1.1 JA (depth: 2, scale: 1.0)
52.31 Stockfish 1.8 JA (depth: 2, scale: 1.0)
50.38 Rybka 3 w32_copy (depth: 2, scale: 1.0)
50.38 Rybka 3 w32 (depth: 2, scale: 1.0)
49.50 Deep Rybka 4.1_w32 (depth: 0, scale: 1.0)
49.28 Deep Rybka 4.1_w32 (depth: -1, scale: 1.0)
49.21 Stockfish 2.2.2 JA (depth: 3, scale: 1.0)
49.08 Deep Rybka 4.1_w32 (depth: 1, scale: 1.0)
48.82 Alaric 707 (depth: 2, scale: 1.0)
47.97 Komodo32 3 32bit (depth: 2, scale: 1.0)
47.61 Komodo32 Version 4 (depth: 2, scale: 1.0)
47.55 Atlas 3.14b (depth: 2, scale: 1.0)
47.34 Rybka WinFinder 1.0 32-bit (depth: 2, scale: 1.0)
47.01 Rybka 1.2 32-bit (depth: 2, scale: 1.0)
46.88 Rybka 1.0 Beta 32-bit (depth: 2, scale: 1.0)
46.86 Rybka 2.3 32-bit (depth: 2, scale: 1.0)
46.78 Rybka 2.3.1 32-bit (depth: 2, scale: 1.0)
46.77 Rybka 2.3.2a 32-bit (depth: 2, scale: 1.0)
46.66 Rybka 2.1 32-bit (depth: 2, scale: 1.0)
46.26 bright-0.2c (depth: 2, scale: 1.0)
46.01 Deep Rybka 4.1_w32 (depth: 2, scale: 1.0)
45.85 Rybka 1.1 32-bit (depth: 2, scale: 1.0)
45.11 Alfil 8.1.1 (depth: 2, scale: 1.0)
42.10 Ares 1.004 (depth: 2, scale: 1.0)
40.51 Amyan 1.72 (depth: 2, scale: 1.0)
40.06 Aristarch 4.50 (depth: 2, scale: 1.0)
39.23 Alex v2.14_win32 (depth: 2, scale: 1.0)
38.49 AliChess 4.25 (depth: 2, scale: 1.0)
34.79 Adam 3.3 (depth: 2, scale: 1.0)
28.67 ALChess v1.5b (depth: 2, scale: 1.0)

By Adam Hair (**)

Date 2012-06-15 02:31

Good

I have not gone back to redo the similarity testing after Lukas Cimiotti discovered the reason for the low self-similarity numbers for Rybka (and thus for Robbolito and Houdini). But I am not surprised by your numbers. Without using 'Clear Hash', the numbers I found for those engines were 60% to 70% and approximately 75% self-similarity. I knew that the numbers would be higher when using that command.

By Richard Vida (**)

Date 2012-06-15 08:07

> Without using 'Clear Hash', the numbers I found for those engines were 60% to 70% and approximately 75% self-similarity.

Hi Adam,

At least for Critter - sending 'ucinewgame' before each search reduces noise even more. This clears history counters and some other data structures which 'clear hash' does not.

Richard

By Adam Hair (**)

Date 2012-06-15 17:33

Thanks for the information, Richard.

At the moment, I am using the similarity tool to run 'go deep' experiments (to study the change in move selections for various engines as depth of search increases). Maybe those results will be more useful to the community. At least less divisive.

By the way, I really like your emphasis on making Critter more useful for analysis, even if I am much too weak of a chess player to make use of its new abilities.

Adam

By Nelson Hernandez (Gold)

Date 2012-06-16 23:01

In my database, containing engine and higher-level human games, here's my frequency:

Qc7 36.3%
Bh7 18.3%
g5 17.8%
Ng6 10.4%
Nc8 10.3%
a5    3.1%
Bg6   2.3%
Rc8   0.6%
Be4   0.4%
a6    0.3%
other 0.2%

The top eight are all playable. Most drawish: g5. Most successful: Rc8. Most likely move book would play: g5, narrowly edging Qc7; scores well, far more tested than Rc8. Worst move that has been played: f6 (sorry, Big Lion). Move I have never seen: Rg8 (Junior 6? That goes back to 1998-99 doesn't it? Not enough games in my database from that era, wish I had more).

By Pia

Date 2012-06-17 00:55

>Worst move that has been played: f6 (sorry, Big Lion).

He's probably improves since 1.8f:

Analysis by BigLion 2.23w:

1...Qc7 2.Bd2 0-0-0 3.Nh4 Be4 4.f3 Bh7 5.Ba5 Nb6
+/- (0.80)   Depth: 6/8   00:00:00 75kN
+/= (0.64)   Depth: 9/21   00:00:06 1493kN
1...Nc8 2.Ne1 Be7 3.Bd3 0-0 4.Bxf5 exf5 5.Qd3 Ncb6 6.Qxf5 Nc4 7.Qd3
+/= (0.49)   Depth: 10/22   00:00:30 6945kN
+/= (0.48)   Depth: 11/25   00:01:28 18591kN

>Move I have never seen: Rg8 (Junior 6?

Analysis by Junior 6.0:

1...a5 2.Nh4 Be4
= (0.19)   Depth: 6   00:00:00 0kN
1...Ng6 2.Ne1 h5 3.h3 h4 4.Nf3 c5 5.Bg5
= (0.11)   Depth: 6   00:00:00 2kN
= (0.25)   Depth: 15   00:00:01 3056kN
1...Bh7 2.a4 a6 3.Bd3 Bxd3 4.Qxd3 Nf5 5.c4 Nb6 6.Na5 dxc4 7.Nxc4 Bc5 8.Nxb6 Bxb6 9.Rd1
= (0.19)   Depth: 15   00:00:03 7161kN
= (0.17)   Depth: 20   00:01:05 132mN

By Sedat Canbaz (*****)

Date 2012-06-17 07:19

>Analysis by Junior 6.0

Dear Pia,

My current testings are different than yours,I did not use Analysis mode

In other words,all engines are forced to play the moves
And once more i'd like to mention that Junior 6 engine played exactly 'Rg8' move (in my conditions)

Btw,what about Komodo,Stockfish...?
Do you plan to test those engine too ?

Best,
Sedat

By Pia

Date 2012-06-17 14:29

That would be 7 versions of Komodo and 23 Stockfish~~~~

Test goes slow for some other engines in my testings, I got various errors including one division by zero :smile:

By Sedat Canbaz (*****)

Date 2012-06-17 17:35

>That would be 7 versions of Komodo and 23 Stockfish~~~~

Do you think that 7 engines are enough ?

But its ok... no problem, if you dont want to test them (Komodo,Stockfish...) against 50 engines

Best,
Sedat

By Pia

Date 2012-06-17 18:38 Edited 2012-06-17 18:40

7 versions of Komodo from 1.0 to 4 @ depth 2. When depth 1 completes, there will be 14 versions of Komodo and so on...
Attaching data file now...
P.S. 247 engine samples

Attachment: similarity.7z (822k)

By Sedat Canbaz (*****)

Date 2012-06-17 19:01

Thanks,but i can't open similarity.data

Which view utility is needed to open this kind of ?

By Pia

Date 2012-06-17 19:03

It's on the bottom http://komodochess.com/downloads.php

By Pia

Date 2012-06-18 20:30

Also site for browsing similarity data https://sites.google.com/site/code404html/similarity-tester-version-3

By Sedat Canbaz (*****)

Date 2012-06-18 22:00

Thanks for the link

By Sedat Canbaz (*****)

Date 2012-06-17 07:06 Edited 2012-06-17 07:08

>In my database, containing engine and higher-level human games, here's my frequency:

Thanks Nelson

Rc8 ?
I never tried this more before,but i already enabled Rc8 move in Perfect 2012b book
Lets see what will be overall performance with Rc8 move
And in case of +40% winning percentage i will allow it for the future

For example,my favorite moves for the current B12 position:
1.Qc7
2.g5
3.Bh7
4.Bg6

Greetings,
Sedat

By hsnksc77 (**)

Date 2012-06-12 12:27

hi,

Deep Rybka 4.1 SSE42 x64 01:04:16 g5.

New game

r2qkb1r/pp1nnpp1/2p1p2p/3pPb2/3P4/1N3N2/PPP1BPPP/R1BQ1RK1 b kq - 0 1

Analysis by Deep Rybka 4.1 SSE42 x64:

1...g5 2.Be3 Bg7 3.h3 0-0 4.Qd2 a5 5.Nc5 Nxc5 6.dxc5 Qc7
+/= (0.35)   Depth: 6   00:00:00 3kN
1...g5 2.Be3 Bg7 3.h3 0-0 4.Qd2 a5 5.Nc5 Nxc5 6.dxc5 Qc7
+/= (0.36)   Depth: 7   00:00:00 3kN
1...Ng6 2.Be3 Be7
= (0.23 !)   Depth: 7   00:00:00 8kN
1...Ng6 2.Be3 Be7
= (0.23 !)   Depth: 7   00:00:00 8kN
1...g5 2.Bd2 Bg7 3.h3 a5 4.a4 0-0 5.Bc3 Qc7 6.Qd2 b6
+/= (0.33)   Depth: 8   00:00:00 20kN
1...g5 2.Bd2 Bg7 3.h3 a5 4.a4 0-0 5.Bc3 Qc7 6.Qd2 b6
+/= (0.33)   Depth: 9   00:00:00 28kN
1...g5 2.Be3 Qc7 3.h3 Bg7 4.a4 0-0 5.a5 Be4 6.Nfd2 Nf5
+/= (0.33)   Depth: 10   00:00:00 59kN
1...g5 2.a4 Bg7 3.a5 Qc7 4.Be3 0-0 5.h3 Rad8 6.a6 b6 7.Qd2
+/= (0.33)   Depth: 11   00:00:00 126kN
1...g5 2.a4 Bg7 3.Bd2 0-0 4.a5 Re8 5.h3 Ng6 6.Bb4 Nf4 7.Qd2 Be4 8.Nc5
+/= (0.34)   Depth: 12   00:00:01 203kN
1...g5 2.a4 Bg7 3.Bd2 0-0 4.Bb4 Re8 5.Qd2 Ng6 6.h3 Nf4 7.a5 Be4 8.Bd6 Nxe2+ 9.Qxe2 Rc8
+/= (0.36)   Depth: 13   00:00:02 422kN
1...g5 2.a4 Bg7 3.Bd2 0-0 4.Bb4 a5 5.Bd6 Re8 6.h3 Nc8 7.Ba3 Ncb6 8.Nc5 Nxc5 9.Bxc5 Nc4 10.Bxc4 dxc4 11.Rc1 Bg6
+/= (0.33)   Depth: 14   00:00:04 1017kN
1...g5 2.a4 Qc7 3.a5 Bg7 4.Bd2 c5 5.Nxc5 Nxc5 6.dxc5 a6 7.h3 Nc6 8.Bc3 Nxe5 9.Nxe5 Bxe5 10.Bxe5 Qxe5
+/= (0.31)   Depth: 15   00:00:08 2068kN
1...Bg6 2.Nh4 Be4 3.f3 Bh7 4.f4 Nf5 5.Nxf5 Bxf5 6.Bg4 Bxg4 7.Qxg4 g6 8.Bd2 a5 9.a4 Bg7 10.Qg3 0-0
= (0.25)   Depth: 15   00:00:13 3315kN
1...Bg6 2.Nh4 Be4 3.f3 Bh7 4.f4 Nf5 5.Nxf5 Bxf5 6.Bg4 Bxg4 7.Qxg4 g6 8.Bd2 Be7 9.a4 c5 10.a5 c4 11.Nc1 0-0 12.Ne2
= (0.22)   Depth: 16   00:00:18 4706kN
1...Bg6 2.Nh4 Be4 3.f3 Bh7 4.f4 Nf5 5.Nxf5 Bxf5 6.Bd2 Be7 7.Bg4 g6 8.Bxf5 exf5 9.Qe1 0-0 10.Bb4 Kg7 11.Bxe7 Qxe7 12.a4 Rfe8 13.Qe3 Rac8
= (0.25)   Depth: 17   00:00:23 6090kN
1...Bg6 2.Nh4 Be4 3.f3 Bh7 4.f4 Nf5 5.Nxf5 Bxf5 6.Bd2 Be7 7.Bg4 g6 8.Bxf5 exf5 9.Qe1 0-0 10.Bb4 Kg7 11.Bxe7 Qxe7 12.a4 Rfe8 13.Qe3 Rac8
= (0.25)   Depth: 18   00:00:33 8551kN
1...Bg6 2.Nh4 Be4 3.f3 Bh7 4.f4 Nf5 5.Nxf5 Bxf5 6.Bd2 Be7 7.Bg4 g6 8.Bxf5 exf5 9.Qe1 0-0 10.Bb4 Kg7 11.Bxe7 Qxe7 12.Rd1 Rfe8 13.Nc1 Nf8
= (0.23)   Depth: 19   00:00:49 12803kN
1...Bg6 2.Nh4 Be4 3.f3 Bh7 4.f4 Be4 5.Be3 g6 6.Nd2 Bg7 7.Nxe4 dxe4 8.c4 0-0 9.Qc2 f5 10.exf6 Bxf6 11.Bf2 e3 12.Bg3
= (0.25)   Depth: 20   00:01:37 25237kN
1...Bg6 2.Nh4 Be4 3.f3 Bh7 4.f4 Be4 5.Be3 Nf5 6.Nxf5 Bxf5 7.Bd3 Bxd3 8.Qxd3 g6 9.Bd2 Be7 10.a4 c5 11.c4 dxc4 12.Qxc4
= (0.25)   Depth: 21   00:05:08 81501kN
1...Bg6 2.Bd2 Nf5 3.c4 dxc4 4.Na5 c3 5.Bxc3 Qc7 6.Nc4 b5 7.Ba5 Qb7 8.Ncd2 Ne7 9.Bb4 a5 10.Bd6 a4 11.Nh4 Bh7
+/= (0.31)   Depth: 22   00:29:13 468mN
1...g5 2.a4 Qc7 3.Bd2 0-0-0 4.Ba5 b6 5.Ba6+ Kb8 6.Bc3 f6 7.exf6 Nxf6 8.Bd3 Ne4 9.a5 Bg7 10.axb6 axb6
= (0.23)   Depth: 22   01:04:16 1039mN

By Pia

Date 2012-06-12 13:09

w32 also likes g5

uci
id name Deep Rybka 4.1 w32
id author Vasik Rajlich
...
setoption name max cpus value 1
info string Max CPUs set to 1 (had been 2048).
setoption name hash value 128
info string Hash size changed from 32 Mb to 128 Mb.
setoption name clear hash
info string Hash clear invoked.
position fen r2qkb1r/pp1nnpp1/2p1p2p/3pPb2/3P4/1N3N2/PPP1BPPP/R1BQ1RK1 b kq - 0 8
go btime 180000

info depth 1
...
info depth 12 time 6314 nodes 201612 nps 31930
info depth 12 score cp -37 time 6330 nodes 201612 nps 31850 pv g7g5 a2a4 f8g7
h2h4 g5g4 f3e1 h6h5 e1d3 f7f6 e5f6 g7f6 c1g5 e7g6 b3c5 f6g5 h4g5 d7c5 d3c5 b7b6
bestmove g7g5 ponder a2a4

By Sedat Canbaz (*****)

Date 2012-06-13 09:13 Edited 2012-06-13 09:36

Dear Friends,

Thanks for your testings...

Just i'd like to mention that any engine's move choice is highly depending on played conditions:
-Hardware Speed
-Engine's Cores
-Time Control
-Opening Position
-Hashtable Size
-Endgames
...

Note also that many engines, which are based on free open source (Ippo, Stockfish, Fruit...) play same/similar moves as well

Btw,just my 2 cents more over this issue

Nowadays we are talking about engine move similarities,yes...there are tons of clone discussions,but still we have different views
In my opinion,the problems are appearing due to many engines use similar stuff/ideas or include same copied code as well
On the other hand,i believe the free open source engines are the main reason about why we see a lot clone discussions and also why the engines are so relative

Actually,i think its now too late to find which engines are 100% original, because there is almost no a such original work

One thing more,I have the same view as Alan's statement:
An open source version of any engine can be created by anyone with a good decompiler. Open source does not confer legitimacy.

Let's imagine about opening book moves/lines similarities...
Can we find a X book,which is completely different than all other books (i mean based on 100% own ideas) ?

And last,
Its time for revolution,we need to give a chance for the 'Young Talents',otherwise there will be a big mistake/injustice over them
Note:I mean for those 'Young Talents', who has capacity to create at least 50 Elo improvement over free open source engines

Kind Regards,
Sedat

Topic The Rybka Lounge / Computer Chess / Similarity Move Test

1 2