Home

RK 2004 Tournament
8 engines, 140 rounds, 560 games, Athlon 1.3/64 MB hash
All games played under ChessBase Fritz8-GUI
Time control: 90m+30s, ponder=off, 3-/4-men EGTB
Books: see below, no book learning
 book options=use book, tournament book, optimize
Participants / Book used / Author of engine:
Aristarch 4.41 UCI  (aristarch.boo) by Stefan Zipproth
Fritz 8.0.0.26  (fritz8.ctg) by Frans Morsch
Hiarcs 9  (hiarcs9.ctg) by Mark Uniacke
Junior 8.0.0.2  (junior8.ctg) by Amir Ban
List 513 UCI  (general.ctg) by Fritz Reul
Ruffian 2.1.0 UCI  (Leiden book) by Perola Valfridsson
Shredder 8  (shredder8.ctg) by Stefan Meyer-Kahlen
The King 3.23 X = CM9 T06  (fritz532.ctg) by Johan de Koning
 Tester: Kurt Utzinger / Rolf Bühler, Switzerland

Download Games
 

Remarks to the tournament

First of all we would like to mention that RK 2004 was the longest tournament we have ever played. The conditions are given on the top of this page.  Allow us to mention that we should take care to not make false conclusions about the outcome of the tournament. Looking at the 20 single series of 28 games with 7 games for each engine, we see that Shredder 8 was five times successful to win a serie. But Junior 8 could also win five series. Hiarcs 9 won four series whereas Aristarch 4.41 and Ruffian 2.1.0 only reached this aim once. List 5.13 did not win one single serie but the engine of Fritz Reul had nevertheless 9,5 points more than Aristarch 4.41 at the end of the tournament. This tournament again showed how important it is to play enough games. There are always series in which the best engines get very bad results. Look at Shredder 8 in serie 18 where the program only got 1.5/7. The worst serie (No. 4) was played by Ruffian 2.1.0 with 0.5/7.  And furthermore we should under no circumstances overestimate the results of this tournament since each engine had to play only 20 games vs all other opponents. This is too less and does not allow to state with enough certainty  which is the best program. On the other hand we see the top engines at the first places in almost all tournaments and this can hardly be explained as pure luck.

What would you say about such a [short] tournament: last place for Shredder 8 with only 7.0/21. This refers to the very unlucky series 7, 18 and 20. And again a prove that we can’t take short tournaments too seriously.

In comparison I have looked at the best three series of Aristarch 4.41. And the result is surprising but also confirms how relative all statistic has to be taken:

These two examples may serve as a warning. It would be easy to filter almost every desired data from our tournament to “prove” anything we like. The probability of giving more than a rough estimate regarding the  the strenght of a specific engine raises considerably the more games have been played. Everybody knows that all can happen in a short tournament. And for this reason we should avoid being at war with one another when reading about so called “unbelievable” outcomes in certain matches and tests.

Interesting to note that Junior 8 was involved in five of the six doubles (identical games between the same opponents) in this tournament. These games were cancelled, replayed and replaced. We see not other way to handle such things. By the way, the Israeli program Junior 8 chose with White 16 times the opening ECO B80 with a top score of 13,0-3,0 (11 wins, 2 draws, 3 losses) or a winning score of 81 percent. Somewhat less successful was Junior 8 with ECO E25, the Nimzo-Indian with 4.f3, in 20 games and 13.5/20 or a winning score of 67,5 percent.

All programs had to accept defeats in matches.  Shredder 8 lost 8,5-11,5 vs Fritz 8. But apart from a draw (10,0-10,0) against CM9 T06, Shredder 8 won all other matches. The second placed Junior 8 had to suffer two defeats vs Shredder 8 and Fritz 8. The somehow unappreciated top engine Hiarcs 9 of Mark Uniacke catched our eyes. With some more luck, this program could have won the tournament. However, the bad result of 7,0-13,0 between Hiarcs 9 and Shredder 8 prevented Hiarcs 9 from celebrating this victory. Fritz 8 is fourth which is due to the 8,5-11,5 loss vs Ruffian 2.1.0 and the less convincing win of 10,5-9,5 against Aristarch 4.41. After 140 games for each engine, there is no real gap between 1st and 5th place since the difference is only 7,5 points.  It was a difficult tournament for the amateur programs List 5.13 and Aristarch 4.41. But both these engines were - the same applies to Ruffian 2.1.0 - not at all a disappointment and all of them have played some very good games. We are wondering whether Ruffian 2.1.0 had a handicap with the Leiden book as the almost single move out of this book was 1.d4. As far as The King 3.23 X (CM9 T06) is concerned, please refer to the next chapitre.

We were not fully satisified with The King 3.23 as the X-settings (CM9 T06) lost three matches vs Junior 8 (9,5-10,5) vs Fritz 8 (8,5-11,5) and even against the amateur List 5.13 (9,5-10,5). This and the three drawn matches vs Shredder 8 / Hiarcs 9 / Ruffian 2.1.0 were responsible that The King only got 5th place. In any case, we shall not publish the CM9 T06 setting. We then took the opportunity of the RK 2004 tourney to test some further CM9 settings by deleting, replaying and replacing all games of CM9 T06 with other settings and partly other opening books. In most cases we could stop the experiment half time (after 70 games) due to unsatisfactory results. Below a summary of such trials:

36.5/70 = CM9 T06 with  fritz532.ctg  (original RK 2004 tourney)
36.5/70 = CM9 T06 with  fritz7.ctg
36.0/70 = CM9 R1   with  fritz7.ctg  (setting of Wilhelm Hudetz)
34.0/70 = CM10k    with  fritz532.ctg (Chessmaster 10000 default)
32.6/70 = CM9 SKR with  fritz7.ctg

It’s noteworthy that the latest Chessmaster 10k (The King 3.33) did not convince with the default settings.  We shall however also play the remaining 10 series (70 games) with CM10k. The testing with other settings was even less successful so that we  thought it would be best to give up the ghost. But then things at once changed with the new CM9 T05 setting. This setting had been tested before on the basis of matches against various engines and the results looked very promising:

Resultats of CM9 T05 (40'/40) 5moves.ctg

1) 64 %  20.5/32  vs Aristarch 4.41
2) 50 %  15.0/30  vs Chess Tiger 15
3) 62 %  18.5/30  vs Fritz 8.0.0.26
4) 43 %  13.0/30  vs Hiarcs 9
5) 53 %  17.0/32  vs Junior 8
6) 60 %  18.0/30  vs List 512
7) 72 %  26.5/37  vs Nimzo 7.32
8) 56 %  17.0/30  vs Shredder 8

We were therefore rather optimistic that CM9 T05 would also do well in the RK 2004 tourney with the book fritz7.ctg but what then happened did exceed all our expectations. But please have a look at the following ranking list:

Download Games CM9 T05

We could hardly believe our eyes: CM9 T05 would have won the great tourney with 10,5 points ahead of  Shredder 8. The reason for this surprising outcome can be found when we consider the very clear wins of CM9 T05 vs Aristarch 4.41 (17,0-3,0) and vs Ruffian 2.1.0 (15,5-4,5). Very helpful were of course also the wins of 12,0-8,0 vs Hiarcs 9, Fritz 8 and List 5.13. Despite this sensational result it would be completely wrong to think that CM9 T05 is a dream engine and stronger than all other professional programs.  No doubt, CM9 T05 is a very strong CM9 setting at longer time controls but Johan de Koning’s program was perhaps too fortunate in our experiment.

The parameters for the CM9 T05 Setting  looks as follows:

****************************************************
CM9_T05.CMP
Created by Kurt Utzinger, Rolf Buehler
[Setting under CM9-GUI]
****************************************************

A/D  (Attack/Defender)     [   0]          M/P  (Material/Position)  [   0]
S/P  (Strength of Play)       [ 100]       C/C  (Control of Center)   [ 105]
Rand (Randomness)           [   0]          MOB  (Mobility)                   [ 110]
MSD  (Max Search Depth)  [  99]         K/S  (King Safety)              [ 155]
SS   (Selective Search)       [  12]         P/P  (Passed Pawns)        [ 100]
C/D  (Contempt for Draw) [   0]           P/W  (Pawn Weakness)   [ 100]

OWN                    OPPONENT
Queen  [ 9.1]    Queen   [ 9.0]
Rook    [ 5.0]      Rook    [ 5.0]
Bishop  [ 3.0]    Bishop  [ 3.0]
Knight  [ 3.0]     Knight  [ 3.0]
Pawn    [ 1.0]     Pawn    [ 1.0]


*************************
CM9_T05 textfile
[for use outside CM9-GUI]
*************************
;---- CM9_T05.txt
; A personality created by Kurt Utzinger, Rolf Bühler
cm_parm opp=100 opn=100 opb=100 opr=100 opq=100
cm_parm myp=100 myn=100 myb=100 myr=100 myq=101
cm_parm cc=105 mob=110 ks=155 pp=100 pw=100
cm_parm cfd=0 sop=100 avd=0 rnd=0 sel=12 md=99
;---- File end


***********************************
CM9_T05 InitString=
[for use outside CM9-GUI]
***********************************
[ENGINE]
Name=The King 3.23 SKR
Author=Johan de Koning, NED
Filename=Wb2Uci.exe

[OPTIONS]
Program=c:\programme\ubi soft\chessmaster 9000\TheKing.exe
InitString=cm_parm default\ncm_parm opk=229570\ncm_parm tts=67108864 opp=100 opn=100 opb=100 opr=100 opq=100 myp=100 myn=100 myb=100 myr=100 myq=101 cc=105 mob=110 ks=155 pp=100 pw=100 cfd=0 sop=100 avd=0 rnd=0 sel=12 md=99
HashOnCommandline=false
InitTime=2
Ponder=false
Logfile=false
Edit=setboard
UseSAN=false
Priority=normal
AddHintMove=false
ShowThinkingMove=true
Protocol=1

Course of the tournament

As it can bee seen from the list below of the individual series, all can happen when not playing enough games. When studying the progression of the tourney it looks rather boring:

After serie 1    (28 games)    =  Hiarcs 9       with  5.0/7
After serie 2    (56 games)    =  Hiarcs 9       with  8.0/14
After serie 3    (84 games)    =  Shredder 8  with  12.0/21
After serie 4    (112 games)  =  Shredder 8  with  18.0/28
After serie 5    (140 games)  =  Shredder 8  with  23.0/35
After serie 6    (168 games)  =  Shredder 8  with  27.0/42
After serie 7    (196 games)   = Shredder 8  with  29.5/49
After serie 8    (224 games)  =  Shredder 8  with  33.0/56
After serie 9    (252 games)  =  Shredder 8  with  37.5/63
After serie 10  (280 games)  =  Shredder 8  with  40.5/70
After serie 11  (308 games)  =  Shredder 8  with  44.5/77
After serie 12  (336 games)  =  Shredder 8  with  48.0/84
After serie 13  (364 games)  =  Junior 8        with  52.5/91
After serie 14  (392 games)  =  Shredder 8  with  57.0/98
After serie 15  (420 games)  =  Shredder 8  with  63.0/105
After serie 16  (448 games)  =  Shredder 8  with  67.0/112
After serie 17  (476 games)  =  Shredder 8  with  70.5/119
After serie 18  (504 games)  =  Shredder 8  with  72.0/126
Afterserie 19   (532 games)  =  Shredder 8  with  75.5/133
After serie 20  (560 games)  =  Shredder 8   with   78.5/140

Serie 1 (games 1-28)
Winner = Hiarcs 9

Serie 2 (games 29 - 56)
Winner = Junior 8

Serie 3 (games 57 - 84)
Winner = Shredder 8

Serie 4 (games 85 - 112)
Winner = Shredder 8

Serie 5 (Games 113 - 140)
Winner = Junior 8

Serie 6 (Games 141 - 168)
Winner = Junior 8

Serie 7 (Games 169-196)
Winner = Hiarcs 9

Serie 8 (Games 197-224)
Winner = Aristarch 4.41

Serie 9 (Games 225 - 252)
Winner = Shredder 8

Serie 10 (Games 253 - 280)
Winner = The King 3.23 X / Fritz 8.0.0.26

Serie 11 (Games 281 - 308)
Winner = Shredder 8 / List 5.13 / Fritz 8

Serie 12 (Games 309 - 336)
Winner = Ruffian 2.1.0

Serie 13 (Games 337 - 364)
Winner = Junior 8

Serie 14 (Games 365 - 392)
Winner = Fritz 8

Serie 15 (Games 393 - 420)
Winner = Shredder 8

Serie 16 (Games 421 - 448)
Winner = The King 3.23 X

Serie 17 (Games 449 - 476)
Winner = Junior 8

Serie 18 (Games 477 - 504)
Winner = Fritz 8 / Hiarcs 9

Serie 19 (Games 505 - 532)
Winner = Hiarcs 9

Serie 20 (Games 533 - 560)
Winner = Hiarcs 9

Home