|
1. Go to
107, FQ Ktulu 7.0 by
Rahman Paidar (Iran) played now 236 games in ATL-2.
ATL-2 results: --- ATL-2 rating list after 2.396 games
Program Elo + - Games Score Av.Op. Draws
1 Shredder 9.0 : 2750 26 26 540 71.4 % 2591 29.8 %
2 Ktulu 7.0 : 2653 36 36 236 48.5 % 2663 34.3 %
3 Gandalf 6.01 : 2653 25 24 540 57.3 % 2602 31.3 %
4 TheKing 3.33 Schumacher : 2631 24 24 540 53.8 % 2604 30.9 %
5 ProDeo 1.1 : 2621 26 26 490 52.9 % 2601 31.4 %
6 Ruffian 2.1.0 : 2619 25 25 526 52.1 % 2604 31.6 %
7 SlowChess Blitz WV : 2563 26 26 480 43.8 % 2607 29.2 %
8 SOS 5.1 for Arena : 2561 26 26 480 43.3 % 2607 30.4 %
9 Spike 0.9 : 2527 26 26 480 38.1 % 2611 31.7 %
10 Pharaon 3.2 : 2498 27 27 480 33.8 % 2615 28.3 %
---After 031 ATL-2 games = 2.693
ELO After 056 ATL-2 games = 2.689 ELO - 4 After 105 ATL-2 games = 2.684 ELO - 5 After 178 ATL-2 games = 2.674 ELO -10 After 236 ATL-2 games = 2.653 ELO -21 Ktulu 7.0 REVIEW: Rahman Paidar (programmer of Ktulu) wrote me: Topic are bug messages by users. In some games Ktulu lost on time. I have the same problems in the first match vs. ProDeo and games vs. TheKing. In the first match vs. ProDeo Ktulu lost two games on time with an evaluation over 15.x. Of course the customers will get the first update of Ktulu automacially. Some other results by testers: Michael Jesdinsky (Arena
Test): Harry Schnapp (Arena Mainbook): Hiarcs 9.0 - Ktulu 7.0 I believe the ELO calculation can be better if we add the ChessBase TOP programs in ATL-2 (not possible, I am playing with WB / UCI engines only). At the moment I have positive results against the product family of ChessBase. ChessTiger 15.0 only have a positve results vs. Ktulu 7.0. But the program of Rahman Paidar have problems vs. strong postional engines, like TheKing, ProDeo, ChessTiger. I believe Ktulu will lost the match vs. ProDeo and the ELO will fall again 10-15 ELO. The CEGT results will be better because Ktulu have to play vs. the product familiy of Chess-Base. Harry wrote me after his latest Hiarcs match in German language: Interesting results that Ktulu does well against TOP programs (like the TOP 5) and not so well against probably a little ELO weaker programs.
106, FQ Ktulu 7.0 by Rahman Paidar (Iran) played now 178 games in ATL-2. It seems Ktulu have problems with TheKing 3.33 Schumacher.
ATL-2 results: --- ATL-2 rating list after 2.338
games Program Elo + - Games Score Av.Op. Draws 1 Shredder 9.0 : 2750 26 26 540 71.4 % 2591 29.8 % 2 Ktulu 7.0 : 2674 42 42 178 49.7 % 2676 34.3 % 3 Gandalf 6.01 : 2653 25 24 540 57.3 % 2602 31.3 % 4 TheKing 3.33 Schumacher : 2627 25 25 526 53.4 % 2603 30.8 % 5 ProDeo 1.1 : 2617 26 26 480 52.8 % 2598 31.5 % 6 Ruffian 2.1.0 : 2615 26 25 492 52.2 % 2600 31.3 % 7 SlowChess Blitz WV : 2561 26 26 480 43.8 % 2605 29.2 % 8 SOS 5.1 for Arena : 2558 26 26 480 43.3 % 2605 30.4 % 9 Spike 0.9 : 2525 26 26 480 38.1 % 2609 31.7 % 10 Pharaon 3.2 : 2496 27 27 480 33.8 % 2613 28.3 % ---After 031 ATL-2 games = 2.693 ELO After 056 ATL-2 games = 2.689 ELO - 4 After 105 ATL-2 games = 2.684 ELO - 5 After 178 ATL-2 games = 2.674 ELO -10
The history of the Ktulu 7.0 rating calculation is very interesting. Ktulu loosed 10 ELO after TheKing 3.33 Schumacher games. I believe with the next games the ELO will be better because Ktulu don't have any problems againt Ruffian 2.1.0 by Per-Ola Valfridsson. I am to 80,54367% sure after all my knowledge from 6 years (I have nothing learned in the latest years) with ratings that Ktulu 7.0 have a final result from around 2.670 - 2.680. Some Ktulu buyers sent the first results: Christian Koch (CEGT member): Harry Schnapp (Arena Mainbook): Chess Tiger 15.0 - Ktulu 7.0 Michael Jesdinsky (Arena
Test): Comment in German language by
Harry Schnapp (70 years): After three games more Ktulu 7.0 won with 6.0 : 4.0 vs. Junior 9.0. Harry wrote that Ktulu 7.0 have problems with the playing style from ChessTiger 15.0. Ktulu loosed with 1.5 : 8.5 vs. ChessTiger 15.0 on Harry's machine. In this case we have to wait of the results vs. Fritz 8.0. Against Hiarcs 9.0 Ktulu 7.0 seems to make more as 50%. If Ktulu have problems vs. positional strong programs the ELO in ATL-2 will fall again with around 20 points. Harry Schnapp wrote me for
around one year after the release of Ktulu 5.1: Yes Harry, our team and only a
hand full persons I know have such an opinion about Ktulu 5.1.
105, FQ Please read the complete message ... Ktulu 7.0 by Rahman Paidar (Iran) was released.
You can send your order, please go
to
Prices: Changes from Ktulu 5.1 to 7.0: Rahman Paidar will create an own
registry key for each customer! Note 1: Information to the new Ktulu
7.0 version can be found on this News-Ticker. You can pay with
Note 2:
Thanks for your interest. No, No ...
104, FQ The first official CEGT/ATL rating list was
born! The new selection isn't complete ready but we are working on it. You will know
the ratings from the TOP engines? Yes or No?
103, FQ Ktulu 7.0 by Rahman
Paidar (Iran) was playing 105 games now and can hold the strong results. 40 moves in 20 minutes on three
different systems.
ATL-2 results: For seconds Ktulu won game 18 on
ATL-2 System 1 (Pentium IV 2.67 Mobile Notebook). Same day two hours
later: A forth match in still running on
an older Pentium III 900 MHz vs. Gandalf 6.01 with 40 moves in 20 minutes After 031 ATL-2 games = 2.693
ELO
ATL-2 rating list after 2.265 games! Program Elo + - Games Score Av.Op. Draws 1 Shredder 9.0 : 2750 26 26 537 71.6 % 2589 30.0 % 2 Ktulu 7.0 : 2684 54 54 105 47.6 % 2700 34.3 % 3 Gandalf 6.01 : 2650 25 25 516 57.6 % 2597 31.4 % 4 TheKing 3.33 Schumacher : 2622 26 26 492 53.8 % 2596 30.3 % 5 ProDeo 1.1 : 2614 26 26 480 52.8 % 2595 31.5 % 6 Ruffian 2.1.0 : 2613 26 26 480 52.6 % 2595 31.0 % 7 SlowChess Blitz WV : 2558 26 26 480 43.8 % 2602 29.2 % 8 SOS 5.1 for Arena : 2556 26 26 480 43.3 % 2602 30.4 % 9 Spike 0.9 : 2522 26 26 480 38.1 % 2606 31.7 % 10 Pharaon 3.2 : 2493 27 27 480 33.8 % 2610 28.3 % I will try to give you in the next 2-3 days the same service and will work today on the Gladiatorshop page. Ktulu 7.0 is tomorrow available, later more! After all I saw I believe Ktulu 7.0 have the playing level from Hiarcs 9.0 / Gandalf 6.01 or better! The tactical playing strenght could be the best from all available chess programs! School grade if I compare with other commercial TOP programs: Tatic style = 1 Positional style = 3 Endgame playing = 2 without tablebase support !! I don't know an other tactical engine with such a strong endgame. This can be the advantage or secret from the new Ktulu 7.0 version. Interesting are the tablebase question! In my opinion Ktulu can be better around 20 ELO. I asked Rahman Paidar and here is the answer: Rahman Paidar: "I am not interesting in supporting TB, and I wouldn't do that. I will correct endgame weakness for the next versions." Why not? Ktulu have a lot of endgame knowledge and this is without tablebase support a fantastic result.
102, FQ This was a long night. I am sitting here and follow the Ktulu 7.0 games on the three ATL-2 machines. The computer chess community have a new sensation. Ktulu 7.0 could be the number three in the World. Note:
Look here ... Program Elo + - Games Score Av.Op. Draws 1 Shredder 9.0 : 2750 26 26 524 72.2 % 2584 30.0 % 2 Ktulu 7.0 : 2689 74 75 56 44.6 % 2726 35.7 % 3 Gandalf 6.01 : 2645 26 26 492 57.9 % 2589 30.9 % 4 TheKing 3.33 Schumacher : 2617 26 26 480 53.9 % 2590 30.6 % 5 ProDeo 1.1 : 2611 26 26 480 52.8 % 2591 31.5 % 6 Ruffian 2.1.0 : 2610 26 26 480 52.6 % 2591 31.0 % 7 SlowChess Blitz WV : 2555 26 26 480 43.8 % 2598 29.2 % 8 SOS 5.1 for Arena : 2552 26 26 480 43.3 % 2599 30.4 % 9 Spike 0.9 : 2519 26 26 480 38.1 % 2603 31.7 % 10 Pharaon 3.2 : 2489 27 27 480 33.8 % 2606 28.3 %Download (games -pgn-, log files): 
ATL-2 League
Or this one ... Program Elo + - Games Score Av.Op. Draws 1 Shredder 9.0 : 2750 15 15 1516 69.6 % 2606 29.4 % 2 Fritz 8 : 2697 19 19 992 62.0 % 2612 28.1 % 3 Ktulu 7.0 : 2692 74 75 56 44.6 % 2729 35.7 % 4 Junior 9 : 2682 19 18 992 59.9 % 2613 28.6 % 5 Gandalf 6.0 - 6.01 : 2655 15 15 1484 56.2 % 2612 31.5 % 6 Hiarcs 9 : 2652 18 18 992 55.4 % 2614 33.5 % 7 Chess Tiger 15.0 : 2639 17 17 992 53.3 % 2615 36.9 % 8 TheKing 3.33 Steadfast : 2638 24 24 510 52.5 % 2621 34.7 % 9 TheKing 3.33 Schumacher : 2634 26 26 480 53.9 % 2607 30.6 % 10 Ruffian 2.1.0 : 2629 15 15 1472 52.2 % 2613 32.5 % 11 ProDeo 1.1 : 2619 15 15 1472 50.6 % 2615 30.5 % 12 List 512 : 2618 18 18 992 50.3 % 2617 32.6 % 13 TheKing 3.33 Yoda : 2597 25 26 482 47.9 % 2612 32.8 % 14 Fruit 2.0 : 2589 18 18 992 45.7 % 2619 28.8 % 15 Deep Sjeng 1.6 : 2579 26 26 510 43.4 % 2625 28.4 % 16 SOS 5.0 - 5.1 for Arena : 2578 15 15 1472 44.3 % 2618 32.9 % 17 Aristarch 4.50 : 2574 18 18 992 43.5 % 2619 30.6 % 18 SlowChess Blitz WV : 2566 18 18 962 43.0 % 2615 32.7 % 19 Ktulu 5.1 : 2562 26 26 510 40.8 % 2626 28.6 % 20 Thinker 4.7a : 2559 25 25 482 42.1 % 2614 35.7 % 21 Anaconda 2.0.1 : 2544 26 26 482 39.9 % 2615 32.2 % 22 Spike 0.9 : 2535 26 26 480 38.1 % 2619 31.7 % 23 Delfi 4.5 : 2531 26 26 482 37.9 % 2617 32.2 % 24 Pharaon 3.2 : 2527 19 19 962 37.1 % 2619 31.3 % 25 Patriot 1.3.0 : 2502 27 27 510 32.4 % 2630 25.5 %Ktulu 7.0 is improved with 130 ELO? Rahman, I need a new statement :-) Based on the following results (56 games so far): Ktulu 7.0 (Athlon64 3.8 GHz)
- Shredder 9.0, 10.5 : 9.5 German comment: I never saw a
so strong tactical engine! Ktulu 7.0 (Dual Xeon 2.8 GHz)
- Shredder 9.0, 5.5 : 14.5 After 031 ATL-2 games = 2.693 ELO
Later more ...
101, Chris Tatham Chris Tatham sent me an interesting news for our webpage:
Club Players - Special books In addition to running computer tourneys, I like to use Arena to practice my openings for general over the board games at the local chess club, weekend tournaments etc..
Duplicate games have been removed from source pgn files and in each case maximum book depth is 24 half moves. Thanks Chris!
100,
FQ Rahman Paidar develops in the last year really a tactical monster. I never saw an engine which won 5x back to back against Shredder 9.0. Ktulu 7.0 starts after 31 games with 2.693 ELO. If Ktulu hold this results the playing level will be the same as Junior 9.0 and Fritz 8.0. All is possible! Let us wait for more games! A new sensation?
Not only the commercial program
Gandalf 6.0 is improved to the preview version with ~ 100 ELO! Program Elo + - Games Score Av.Op. Draws 1 Shredder 9.0 : 2750 27 27 511 72.4 % 2582 29.7 % 2 Ktulu 7.0 : 2693 98 100 31 41.9 % 2749 38.7 % 3 Gandalf 6.01 : 2647 26 26 480 58.4 % 2588 31.0 % 4 TheKing 3.33 Schumacher : 2618 26 26 480 53.9 % 2592 30.6 % 5 ProDeo 1.1 : 2612 26 26 480 52.8 % 2592 31.5 % 6 Ruffian 2.1.0 : 2611 26 26 480 52.6 % 2593 31.0 % 7 SlowChess Blitz WV : 2556 26 26 480 43.8 % 2599 29.2 % 8 SOS 5.1 for Arena : 2553 26 26 480 43.3 % 2600 30.4 % 9 Spike 0.9 : 2520 26 26 480 38.1 % 2604 31.7 % 10 Pharaon 3.2 : 2490 27 27 480 33.8 % 2608 28.3 %Yesterday in the evening I started the first games with Ktulu 7.0 on Dual Xeon 2.8 GHz. Ktulu played very good chess but loosed some fights in endgame positions against Shredder 9.0. I am very surprised about the very tactcial playing style of Ktulu. Unfortunately, Ktulu loosed the first games with 0.5 : 3.5. I started in the night the same match on my Athlon64 3.8 GHz and went to bed. In the morning I saw that Ktulu 7.0 won 5 games back to back ... Look here, ATL-2 standings so far: You can download the games vs. Shredder 9.0 on
I can't believe the results and contact the programmer Rahman Paidar (Iran) with a short mail. His answer, text by Rahman Paidar: Read more about Rahman and Jan Kiwitter
... BTW (sorry Rahman): In the maintime I started Gandalf 6.0 vs. Ktulu
7.0 on a slower Pentium III 900MHz machine. I can't wait of the next ATL-2
results and have to see Ktulu against an other opponent. The results were 4.0 :
0.0 for Ktulu 7.0. If you like, you can download these games
--- Ktulu 7.0 is available
end of this week in
Advance order is possible: Update: Ktulu 5.x to Ktulu 7.0 = 14,95
EUR
99, FQ Offenbar gibt es Gerüchte um unsere Arena Chess GUI und dem Schachprogramm Crafty. Per E-Mail wurde mir zugetragen, dass Crafty ein Engine Partner der Arena Chess GUI ist. Ich wurde gefragt, warum diese Information denn nicht auf der Arena Seite verfügbar ist. Dieses Gerücht möchten wir ganz deutlich zurückweisen. Crafty ist kein Partner der Arena Chess GUI und die Engine von Prof. Dr. Robert Hyatt wird es auch niemals werden. Vermutlich kommt das Gerücht dadurch zu Stande, weil ich mal alle Crafty Versionen gesammelt habe und auf Frank's Chess Page zum Download angeboten habe. Diese Sammlung steht heute noch auf der Seite von Leo Dijksman. Wir vermuten, dass die "Engine Partnerschaften" für Besucher unserer Webseiten nicht so ganz klar sind. Daher möchte ich zu der Idee ein paar Worte verlieren. Wir sicherlich bekannt ist, entwickeln wir eine
Benutzeroberfläche (GUI) für Schachprogramme (Engines). Eine eigene Engine wird
nicht entwickelt. Arena unterstützt die beiden Engine Protokolle Winboard
und UCI in Protokoll Version 1 und 2. Seit der ersten verfügbaren Arena
Version war Arena kompatibel zu UCI. In der Zeit, als wir die Webseiten
aufgebaut haben, hatte ich selbst noch Gambit-Soft bei deren Aktivitäten
geholfen und wollte mit der WinBoard Edition CD II und III das seinerzeit
neue UCI Protokoll von
Stefan Meyer-Kahlen und Rudolf Huber so gut es geht supporten.
Martin Blume bevorzugt selbst das UCI Protokoll und steht mit seiner
Entwicklung für Kompatibilitäten. Freie Engine Protokolle sorgen für
Kompatibilitäten und es sollte mittlerweile jedem klar sein, dass wir versuchen
mittels Arena die Protokoll Ideen zu powern. Ein weiteres Beispiel ist
Chess960. Auch Aice ist kompatibel zu Chess960 geworden und
wird in dem Arena Chess960 Setup einfließen. Der Programmierer von Aice
Zurück zu den Partnerschaften: Personen die gerade erst beginnen sich mit der Vielfalt der meist frei zur Verfügung stehenden Engines zu beschäftigen, wird sich ganz sicher nicht mit kryptischen WinBoard Einstellungen auseinandersetzen. Auch wenn Martin einen hervorragenden Support für WinBoard programmiert hat (es ist kinderleicht eine Winboard Engine in Arena einzubinden), heißt das noch lange nicht, dass das meines Erachtens nunmehr eindeutig bessere Protokoll "UCI" nachrangig ist. Stefan Meyer-Kahlen hat durch die kürzlich in das UCI Protokoll eingeflossenen sechs Erweiterungen bewiesen, dass er mit Kritik umgehen kann und bemüht ist, dass UCI Protokoll zu verbessern. Nach meinen Informationen hat zum Beispiel Dieter Bürßner dazu beigetragen UCI zu verbessern. Dieter Bürßner (Programmierer von Yace) hat UCI sehr früh unterstützt. Yace war nach Shredder und SOS das dritte Programm, welches kompatibel zum UCI Protokoll geworden ist. Ein wenig Protokoll-Geschichte kann ja nicht schaden ... Wir stellen uns dem Fortschritt nicht entgegen. Winboard wurde in Zeiten von Frank's Chess Page gepowert (danach von Thomas Mayer und heute von Leo Dijksman), UCI in Zeiten von Arena bzw. der Seite von Alexander Schmidt natürlich auch der Seite von Patrick Buchmann (Patrick hat wie kein anderer Computerschach in Frankreich ... weltweit ... vorangetrieben) und auch Chess960 muss unterstützt werden. Reinhard Scharnagl und Martin Blume stehen auch für Chess960. Stefan Meyer-Kahlen überlegt Chess960 in das UCI Protokoll aufzunehmen. Wir stehen für freie Protokolle, für Kompatibilitäten und möchten Interessen verbinden. Dadurch entstehen Freundschaften und viele Personen werden animiert. Genau das ist das eigentliche Ziel unserer Webseite. Wer diese Worte nun wirklich genau liest wird von selbst darauf kommen, dass Crafty niemals ein Partner unserer Arena GUI werden wird. WinBoard ist nicht out aber wir denken, dass das Konzept von Stefan Meyer-Kahlen für alle Beteiligten einfacher ist. Prof. Dr. Robert Hyatt sieht keine
Veranlassung UCI zu unterstützen. Muss er vielleicht auch nicht, denn Crafty
kommuniziert ausgezeichnet mit dem älteren WinBoard Protokoll. Die Auffassung
von Robert ist vertretbar und ja auch auf unseren Webseiten seit langer Zeit
nachzulesen. Unter
Lasst Euch also nicht von irgend welchen dummen Äußerungen verrückt machen und genießt einfach die wirklich beeindruckenden Engine Entwicklungen. In Foren wird meist ziemlich dumm provoziert, es sind meist die selben Personen die unangenehm auffallen. Besucht unsere Webseiten oder andere Webseiten zum Thema und habt Spaß an den ganzen Entwicklungen. Ob nun WinBoard oder UCI, dass spielt heute dank Benutzeroberflächen wie Arena nun wirklich eine untergeordnete Rolle. Arena bietet zu beiden GUIs eine 1:1 Anbindung die vorbildlich ist. Wir schaffen keine Inkompatibilitäten, sondern sorgen für Kompatibilitäten und das seit den Geburtsstunden von Winboard (Frank's Chess Page, als die ersten Engines kompatibel wurden), UCI und Chess960. Vielleicht wird am Wochenende das nächste interessante Interview zur Verfügung gestellt. Das aktuelle Thema ist Thinker von Lance Perkins (Canada). Freuen wir uns auf dieses Interview, denn eine neue Thinker Version steht in Kürze an. Danach folgt ein Interview mit dem Aice Programmierer Milikas Anastasios. Damit kein falscher Eindruck entsteht? Vielleicht hat jemand Lust den Text in englisch
zu übersetzen. Ich habe keine Lust ... Schreibt uns doch mal eine E-Mail. Was können wir besser machen, was gefällt nicht ... oder möchtest Ihr vielleicht auch mal ein Bericht über Arena zur Verfügung stellen. Wir sind für jede Hilfe dankbar !! Wir wünschen allen Lesern ein schönes Wochenende. Diese Aussage ist vielleicht die spannendste im ganzen Text :-)
98, FQ Sorry, I don't find a better title for this
news.
End, over ... up is the game ??!
You and the others have to visit the following
site, just great! The chess computer community is back with a
great new site. We all know the page of Kurt Kispert and his fantastic
efforts he do for us with information about the older chess computers
Our PHP expert Christopher Conkie will find this news. Thanks Chris for your English corrections of my questions to Fabien (interview with Fabien). I corrected the mistakes for some minutes ago. The Arena admins Michael Diosi and Christopher Conkie help us a lot, each time, each hour, each minute ... Michael, do you recognized that 5 engines are
not longer available yet. Goliath and ESC are two of this :-( Life isn't all beer and skittles Christopher. There's no use crying over spilt milk. Do you understand what I mean? The spirit is willing but the flesh is weak. The chess computers return the earth. This is the message I will give you today ... The chess computers return the earth. WOW
97, FQ Since the Neanderthals we try to produce ELO ratings of available chess programs. Heinz van Kempen, Christian Koch and Charles Smith are played the CEGT-1 and CEGT-2 tournaments. CEGT-3 is still running. With the time control ~ 40 moves in 20 minutes on six fast machines. I am playing my ATL-2 tournament on three fast machines with the same time control (CEGT tournaments with Nunn / Noomen positions under Fritz GUI, ATL-2 tournament with opening books under Arena Chess GUI). What is to do?
CEGT/ATL rating list: Program Elo + - Games Score Av.Op. Draws 1 Shredder 9 2750 2801 +51 16 16 1343 69.8% 2604 29.0% ATL-2 2 Fritz 8 2697 2705 + 8 20 20 863 61.8% 2613 27.6% incompatible 3 Junior 9 2688 2692 + 4 20 20 863 60.5% 2614 29.5% incompatible 4 Gandalf 6.0 - 6.01 2656 2650 - 6 16 16 1318 56.3% 2612 31.3% ATL-2 5 Hiarcs 9 2648 2686 +38 19 19 873 54.7% 2616 33.6% incompatible 6 TheKing 3.33 2634 2676 +42 18 18 1014 52.8% 2615 32.6% ATL-2 7 Ruffian 2.1.0 2633 2667 +34 16 15 1319 52.7% 2614 31.9% ATL-2 8 Chess Tiger 15 2630 2676 +46 18 18 862 51.9% 2617 36.9% incompatible 9 ProDeo 1.1 2618 16 16 1323 50.2% 2617 30.4% ATL-2 10 List 5.12 2613 2648 +35 19 19 863 49.1% 2619 34.1% will start in ATL-2 11 Fruit 2.0 2588 2641 +53 20 20 862 45.4% 2620 29.4% will start in ATL-2 12 Aristarch 4.50 2582 2633 +51 19 19 862 44.4% 2621 30.9% will start in ATL-2 13 Deep Sjeng 1.6 2579 2589 +10 26 26 510 43.4% 2625 28.4% will start in ATL-2 14 SOS 5.0 - 5.1 for Arena 2575 2610 +35 16 16 1282 43.1% 2623 32.7% ATL-2 15 SlowChess Blitz WV 2573 20 20 768 43.3% 2620 31.9% ATL-2 16 Ktulu 5.1 2561 2568 + 7 26 26 510 40.8% 2626 28.6% will start in ATL-2 17 Thinker 4.7a 2555 2572 +17 30 30 352 41.3% 2616 32.1% will start in ATL-2 18 Anaconda 2.0.1 2545 2570 +25 31 31 328 39.8% 2617 34.5% incompatible 19 Delfi 4.5 2540 2564 +24 30 30 353 39.0% 2618 32.6% later 20 Spike 0.9 2534 28 28 430 35.9% 2634 31.4% ATL-2 21 Pharaon 3.2 2524 2566 +42 23 23 628 34.5% 2636 30.1% ATL-2 22 Patriot 1.3.0 2502 2549 +47 27 27 510 32.4% 2630 25.5% laterFritz, Junior, Hiarcs, Tiger and Anaconada are incompatible engines. This means, that this group of engines don't support the free standard engines protocols Winboard or / and UCI. The engines can't participant in ATL-2. If you know a rating list with longer time controls (40 moves in 20 minutes or longer) with so many actual engines and games please give me a short hint! I can't find an other list but I know that people like it to play private tournamnts. The results goes by mail to the programmers. I don't hope that not all users have the same idea because we all like to see all the wonderful webpages with eng-eng results / tournaments. In red you can find the CSS ratings after 480 games only. This rating list are played on one Athlon system only and with 10 minutes + 10 seconds Fischer time control under Fritz GUI. Perhaps the only one I can "a little bit" compare with the CEGT / ATL list. The main problem of the CSS list is, that Klaus Wlotzka used one system only. The CEGT/ATL tournaments are produced on 9 very fast systems (4x Intel and 5x Athlon). In ATL detail page is very nice to see, that engines produced different results on different processors. A very important point today. A good rating list should be played on more systems and should be a good combination between Athlon and Intel processors. ProDeo 1.1, SlowChess Blitz VW and Spike 0.9 aren't tested yet in CSS rating list. Interesting are the very bad CSS results from Gandalf. Gandalf need more time and like Pentium processors. We all know that Shredder is Athlon optimized and Fritz is Pentium optimized. To see that Shredder is around 100 ELO in front of Fritz (CSS ratinglist) isn't very nice. I say it ... a good combination produced good ratings.
The combinated CEGT / ATL rating list will
updated frequently, later more!
96, FQ
v. 2.0
Hint: Fruit 2.0 is number 7 in ATL-1 Rating-List!
The title for this interview could be "The time after Crafty" but I used the title Fabien like "The alternative to Crafty" ... I like both titles because Crafty isn't the only one open source engine. The newcomer Fruit produced in my opinion since a while the most interesting computer chess topics. This gave me a lot of material for our Arena News-Ticker and I am very happy that the programmer of Fruit, Fabien LETOUZEY (France), gave the permission for this interview. My thanks go to Robert W. Allgeuer (Fruit main tester, Luxembourg) which sent additional information to the playing style of Fruit. The interview is created before easter, the comments by Robert W. Allgeuer too (see the end of the interview). Note:
The alternative to Crafty 01. Frank Quisinsky
02. Frank Quisinsky Fabien LETOUZEY Amateur computer chess as I see it is not a global research project, but rather the meeting of different people each with his own goals. In particular, engine authors are only ONE part of the community. What would they be without testers and public tournaments? The community has become relatively large, probably because of Internet. It is only natural that some "side groups" have emerged. CCC is a rendez-vous point, but in my opinion the topic is too general for our needs. There are too many posts for people who are not on-line regularly to follow. I don't think the creation of separate fora is a big problems. It allows more people to participate (e.g. for those who can't or don't want to speak English) and if an interesting exchange takes place somewhere, one can post a reference to it in other places. It may look a bit strange but I think it's working well. It still feels like everybody knows everybody, which is reassuring. I leave it to you whether amateur chess is not as strong as before (whatever "strong" means), as I was not here to see it for myself. It is not at all the impression that I am getting though. I do believe that there are more exchanges nowadays, especially in the form of additional software (anything that is not an engine). I don't know if Fruit was really a special topic before Toga appeared. There is no reason why it would be more important than other engines. I don't think I am especially close to users, for instance I can't remember a feature that I included to please one of them. Instead I have the feeling that the majority was attracted by engines that specialise in attacking the king, not quite my cup of tea ... In other words I have always considered Fruit's playing style as a handicap regarding popularity. I think highly of most programmers because each of them contributes in some way: helping others, giving their point of view, bringing "side" software, etc ... That's the vast majority. The only ones I am not interested in are those who never communicate. They release and that's all; no difference with most professionals. One person that deserves a special mention is Tord Romstad. I think he has been very influential in the last two years, and this has little to do with open source. Among the new generation, Alessandro Scotti is promising! Many people might think that programmers discuss techniques less and less, but it is simply more and more difficult to post about something really new/different. The new generation of programmers can find many answers by digging into fora archive section. I can't hide my ideas, but I did not claim I had any of interest either. I prefer to think that I have "principles"; I follow them, whether good or bad. I expect that anybody who had a look at the source code failed to find anything "special" he was looking for ... Perhaps the real idea is that the components in an engine might not be as important as the glue that is tying them (don't look for too much deep meaning in this sentence though).
03. Frank Quisinsky *** Polyglot,
one of the best tools the amteur chess area have *** Polyglot can be found on WBEC Ridderkerk by Leo Dijksman
Fabien LETOUZEY The choice of UCI is based on software-design principles that are not easy to explain. It's a programmer's thing really, I don't expect engine users to understand. Let me give you a clue though: think about young WinBoard engines that you have tried; how many handled pondering ... without bugs??? Another clue might be that surely, Stefan Meyer-Kahlen knows a lot about good programming, right? So trust him if not me, UCI is good for programmers because it leads to fewer bugs in the code ... At the same time there is no usable UCI interface on Linux, the system I am using. Yes I know of Jose but it is slow and does not allow engine matches. The only interface is XBoard, WinBoard's Unix cousin. This means I had to write a separate program to link Fruit to XBoard and translate both ways: PolyGlot. Note that although PolyGlot was born as separate software due to a programming principle for engine development, other people can benefit from it as well (including on Windows and Mac OS). Soon after its first release UCI engines were allowed in (all, I think) public tournaments using WinBoard-related software, an important event. Nowadays the decision to support either protocol or both is entirely the programmer's and hardly has an impact on users. I agree that adapters are not easy to use (although it's possible to write additional software to easily configure them, any candidate?), but of course there are only smart people in our community :) Now regarding my "choice" of the WinBoard forum. My opinion is that I did not choose that forum, it's the forum that chose me. It sounds megalomaniac but read on ... I had sent beta-version source code to Dann Corbit, who built an executable and posted messages on CCC and the WinBoard forum. CCC was reaction-less (which is OK with me), but there was a "large" amount of action on the other one to the point I was invited there so that I could react. I answered the questions as only seems natural, and never really left the forum afterwards. It became my new "house" (I had regular access to Internet at that time). It was then I understood that CCC was about computer chess alright, but not AMATEUR computer chess. After the release of PolyGlot there was another reason for me to use the WinBoard forum: using PolyGlot makes sense mainly in conjunction with WinBoard, since most other GUIs are UCI-compatible. It was therefore the best way to contact PolyGlot users. Now a precision for the readers of this (Arena) forum: "WinBoard" forum is mostly a misnomer, it really is about amateur computer chess in my opinion. At least that's my use of it. As far as I know, UCI is not considered off-topic in the forum, so there's no need to shy away. I would say the same thing regarding other GUIs (see for example the AEGT sub-forum). Sorry for the long explanation, I hope that you are now convinced that I am only PARTLY inconsistent :)
04. Frank Quisinsky Fabien LETOUZEY My wish to see "Arena for Linux" was genuine but should be seen in general terms: any program that has the same functionality as Arena is fine with me, it does not have to be exactly the Arena software. Let me state clearly that it is very difficult to write portable software that has a graphical interface. There are solutions like Java but they bring their own set of "side effects" like slow down and not looking natural in the host OS (e.g. I guess Java programs on Windows don't look exactly the same as native applications). Porting GUI-less software like chess engines is much easier. Therefore I know that porting Arena itself would not be a piece of cake for Martin. On top of that Arena is written in Delphi, a Windows-specific language. In all fairness I should not ask for an Arena-like program to be written on Linux, but rather do it myself and make it available to others. It is a huge amount of work though (as you know) and my own needs are not large enough to justify such effort. Frank, you could post on CCC to obtain reactions from potential users on this topic. The reason I chose Linux used to be technical. Windows 95 is in my opinion the worst piece of software ever written, and I never considered using it (bless you if you survived it). With Linux I could write my own bugs without wondering who crashed first in case of a problem ... Nowadays the technical matter is less important, thanks to MicroSoft adding standard techniques in his NT series (Windows 2000 and XP are part of the NT family). However there is another reason for me not to use them: the feeling of freedom. When I use Windows to access Internet from cybercafes, I have the feeling that the machine is remote-controlled by big-buck companies. I am sure you all know what I am talking about. In any case there is no question that by not using Windows, I lose access to 95% of chess GUIs, engines etc ... But I place my own freedom higher than this! Short comments by Frank: To CCC: Example:
05. Frank Quisinsky Fabien LETOUZEY You need to know that I have spent the last two years in England (I now live in France again), and travelling for those events was never considered. On top of that I have little interest in participating in tournaments. I much prefer coming as a visitor (yes I have tried both, in other games). Lastly my budget seldom leaves me room for travelling at the moment. *** Fruit,
the sensation of the year 2004 / 2005 ***
06. Frank Quisinsky Fabien LETOUZEY In Fruit I try to mix development time between improving the program (which does not make me a better programmer, my goal) and cleaning up the code. The latter takes considerably more time without any visible difference. Only one addition between versions 1.5 and 2.0 is important: the pawn-shelter bonus in the evaluation function (something most other engines have always had). It probably did not take more than two hours to code. More generally, features that have made a difference since Fruit 1.0 were mostly a matter of one or two week-ends at most. The only other main addition is history pruning, but I think it has a small impact, if at all. Testing is another story though, there I am greatly in-debt to Joachim Rang and Robert Allgeuer. For instance, Robert pointed out that ETC (Enhanced Transposition Cutoff) seemed to hurt Fruit in several versions. I had been using it in versions 1.0 and 1.5, and its removal might be an important decision for Fruit 2.0. The heuristics in itself looks harmless enough and I could not find a bug in my implementation, but complex interactions might be responsible for a drop in strength. In any case, my point is that part of the strength increase between Fruit 1.5 and 2.0 is not even related to me at all!
07. Frank Quisinsky Fabien LETOUZEY The clarity of the code is only due to my programming style and not at all with potential readers in mind. Clearer code leads to fewer bugs. It's an interesting characteristic of programming that one needs years of experience to learn how to AVOID complicated things. More concretely, in chess that leads to slightly-slower programs (say 25% lower NPS) that are easier to modify. For example Fruit spends a lot of time scanning board squares to evaluate mobility. I know several faster ways I could compute the same thing. However none of them would allow me to easily modify the mobility feature later. I should also explain why Bob's code is "less readable" in my opinion. This has nothing to do with poor programming style. Bob has a considerably superior programming experience than I have, and he obviously knows what he is doing. Simply, Crafty's code is full of speed tricks. There is no doubt that it is very fast. Activate both futility pruning and delta pruning in Fruit (Crafty is using both features), and see Fruit's "real" NPS for yourself. Note that these features reduce node count and are slightly beneficial overall, proving once again that NPS is not everything. So this style is better for Bob (stronger engine than naive code) but perhaps not as good for readers. It is probably unavoidable for "old" programs. I am convinced that Fruit would look the same if I worked on it for many years. However I have as principle to rewrite projects from scratch every few years, both to try new ideas and to accommodate with ever-changing programming style. I am sorry to say that whether an engine is someone's "own work" makes little sense to me, although I understand that tournament directors would like a clear yes or no. The reason is that all engines, whether amateur or commercial, share most of the techniques. Alpha-beta (of which PVS, NegaScout and MTD(f) are only derivatives), iterative deepening, check extensions, null move, etc ... are shared by most and have been published, mostly by researchers, some of them more than 30 years ago! Sure there are many different ways to represent the board and pieces but it only affects speed, which rarely amounts to more than a few dozen Elo Points. There is one component that so far is distinct in each engine (although some older ones were probably "inspired" by Crafty): the evaluation function. But there again the evaluation features are hardly ever very original: the principles of sound chess play can be found in hundreds of books. It's hardly a secret that rooks should be placed on open files, something that Fruit does not even know (though rook mobility partly emulates this piece of knowledge). So what is left for improvisation? A lot of course, otherwise all engines would be equal. But say in terms of quantity of code they don't represent so much. Among this "lot" I think there is a large place for things that cannot be extracted: programming style and ways of linking engine components, making them work together. Not something that most would consider as a "chess-engine technique" like null move. OK let's stop here and do a little sum up with Fruit in mind: I can't think of a search feature in it that was not described before. Ditto for evaluation terms (except perhaps a few drawish-endgame rules that activate in one game in a hundred). There are specific principles that I follow in Fruit that gives it a personality somewhat (like never truncating the PV and making sure that mate-depth claims are always correct), but they probably have no impact on strength at all and could even hurt a little. Can I claim that I have written it all on my own? "Yes", I typed all the code myself. Without help??? Certainly not, hence my point: "it makes no sense". Sorry for the dramatic style ... One positive point now: instead of seeing engine authors competing against each others, I see them as cooperating (mostly indirectly) and making progress together, since they have so much in common, whether they want it or not. My opinion anyway ... Something (hopefully shorter) about publishing source code in my case. It has nothing to do with a plan to get help or anything. It is a perfectly natural thing to do for anybody who is NOT using Windows. It allows users of less well-known systems (I can cite many) to use the program. It also serves as a description of the techniques I have chosen, since I don't have the time to intervene in most forum discussions. People can find ideas (with no guarantee) or simply discover a "different" programming style. Last but not least, some programmers can gain the confidence that no special "secret technique" is needed to reach that level and that they should follow their own ideas. It's about freedom.
08. Frank Quisinsky Fabien LETOUZEY I used to think that using heavy forward pruning (not counting null move) was the key. Although all commercial programs seem to use one form (recall I can't use any of the top programs, even amateur ones, since they run only on Windows), I don't believe it's the main difference between amateurs and professionals anymore. In fact the necessary techniques might not be algorithms at all, but planning/development techniques! E.g. in Fruit 1.0 I started with an outrageous "I know nothing" evaluation function. Then instead of adding dozens of common evaluation terms, I watched games and chose a few that "seemed" important in those games. It seems to work very well for me, but it might apply only to evaluation. When the engine's level is already high, the inclusion of a strong player into the development team probably becomes necessary. The importance of serious testing has been stressed in the last few years, for example the popularity of EPD testing has gone down. There is a place for it, but not as a measure of strength. Something that I think is very important for development is consistency! A chess engine is not just a bunch of heuristics put together. I say this because it seems a common misconception. I believe a chess engine should have a design (e.g. development "rules"), a hidden personality if you want. In particular search and evaluation should work together. Now about making Fruit stronger. Let's not talk about a 100-Elo strength increase, that is too much to expect in one go. I believe I can easily get 50 more Elo points, after that I don't know. Maybe Toga is already a proof of that. Note that "easily" does not mean "quickly", but rather "with little effort". This is partly due to the design in Fruit that I have not yet achieved (things I planned from the beginning I still have not tried/added). Also, the "method" of fixing things that look wrong in games (as opposed to making popular additions and see what happens) has done well so far. So I have no reason yet to think that it should suddenly fail to work. Of course it will fail, eventually. I am only say 1800 Elo so my ability to recognise errors is necessary limited. In the case of Fruit 2.0, I would focus on "holes" in the evaluation function. Simply some features I have not put there at all, because they did not look urgent enough. Scoring king attacks is probably a priority, and there are endgame mis-evaluations I am not happy with. Regarding the search I would go the other way around: try to remove as many features as possible. E.g. if extending single replies to check does not gain much, I would prefer to remove it to simplify the code. Note that since Fruit 1.0, history pruning has been the only big change. Unfortunately my interest in working on evaluation is very low because I don't learn anything in the process and I cannot apply the same changes to games other than chess. For this reason I cannot promise important future development on Fruit. To end on a positive note, It's also possible to make progress without any development: by simply tuning existing features. This is why I leave technical UCI options in release version.
09. Frank Quisinsky Fabien LETOUZEY However I don't want the readers to think that Fruit is going to be especially important for the future of computer chess. I think that my contribution will mostly affect the design of new engines through confidence and not help improving already strong ones. Now about amateurs vs. professionals. There should not be a problem at all because amateurs want to enjoy themselves and commercials use marketing techniques to sell their products. The buyer will see ChessMaster in the local shop or Shredder on chessbase.com but not amateur engine XXX hidden in the download section of a site he is never going to visit (because not advertised). On top of that amateur-chess enthusiasts DO buy most top engines, regardless of how strong List is. Because of this, I don't see a problem in the current situation. We can enjoy ourselves without conflicting with the market of others. I can imagine a positive impact of strong amateurs for engine buyers though: I am still surprised to see that apart from Shredder, all top engines use private protocols (if I understand correctly). If amateurs give more power to customers like the right to ask for features, is it such a bad thing? Another way of seeing it is that as long as the big guys don't make a move toward users, it doesn't look like they are in such a difficult position don't you think? In the end, it's freedom of life vs. freedom of commerce. I leave it to philosophers what is right and what is wrong. Now for those who think that releasing open source is has a special impact on the market (which I disagree with), think about the following events: - discussing techniques in fora (exchanging
ideas)
10. Frank Quisinsky Fabien LETOUZEY However I feel it is important that new programmers don't look at Crafty as a Holy Grail (as they used to do a few years ago). I think the main reason why amateurs have been making big progress in the past few years is that they don't try to follow Crafty anymore. Note that Bob is not responsible for this. Perhaps the lack of communication in the mid nineties was. Now sources of information are plentiful. Crafty is one way, not THE way. Fruit is not THE way either, you must find your own way (sorry). That's my message.
11. Frank Quisinsky Perhaps the following constellations? A. B. C. Fabien LETOUZEY
12. Frank Quisinsky Fabien LETOUZEY Now, when the time comes I am not even sure I will want to make exactly those changes, even though they might easily bring me 50 Elo points. I have a clear picture of what I want in Fruit, and what Thomas has done might or might not fit my design (e.g. the order in which I want to add new features). I hope this will help tournament organisers to make their decision. If they want precisions, they should ask ...
13. Frank Quisinsky Fabien LETOUZEY However: I think it's fair enough that other more-recent engines now get attention as well. At the top, there are new promising versions of DanChess, Glaurung, Pro Deo, SlowChess Blitz, SOS, Spike, Zappa and others that need extensive testing. Toga might or might not be added to this list depending on what each tester is trying to achieve. If it's at the expense of Fruit 2.0 then so be it, I will wait for the results with interest anyway. In any case, let us not forget that tournaments belong to their director(s). I will neither try to intervene nor complain. If they need information to make their decision, they should feel free to ask. ***
Note: Fruit 2.0 will start after two other enignes in ATL-2 ***
14. Frank Quisinsky Fabien LETOUZEY 15. Frank Quisinsky Fabien LETOUZEY Regarding Fruit's actual playing style:
Robert W. Allgeuer`s comments to the playing style of Fruit: Text by Robert W. Allgeuer: |
