March 09th, 2005
About Naum, Interview with Aleksandar Naumov
UCI / WB compatible
Home of Naum
Interview with Aleksandar Naumov by Frank Quisinsky
Hint: Naum is 41 in ATL-1 Rating-List, in reality number ~ 14-20 !!!
Aleksandar Naumov about Arena News-Ticker:
By the way Frank, your interview with developers is a great idea. I really like finding out who is the person behind the engine (even when the person is a crazy guy with a knife ;). Engine info you ask about is interesting, but I would also like to know more about the person (age, birth place, profession, education,...).
Thanks Alexander, this comment will give me more energy for the future. And now let us start!
Age, birth place, profession, education ...
Your move :-)
Aleksander second comment to the idea to make little interviews with the programmers:
The idea is so far unorganised by myself. I like your wishes and will organise the idea a little bit. I hope that the programmers will give me the information for Arena page. If Frank write a mail the programmer can be not sure that a little interview will be the result :-)
I am sure the other guys wouldn't mind giving you a small info about themselves. I would like to know more about, for instance, Tord, Fabien and Uri, so guys, don't be shy. Of course, some of them might be a horribly disfigured monsters, so in that case they may not want to send their picture, and I suspect that Uri is some kind of artifical inteligence living in cyberspace, since he seem to reply instantly to all messages posted in chess forums :)
Do you work on Naum at the moment?
Yes, but I am not sure when I am going to release a new version. I am working on improving time management, and didn't improve engine's strength too much. For instance, I fixed a bug Naum had for fixed time for the whole game. Naum would lose on time in a position that is EGTB draw, because it will try to think instead of playing instantly. New version of Naum will almost never lose on time.
Do you look in the ATL-1 games. I can't explain the results. If I look on other ratings I find on WWW Naum have clear better results then on my Dual Xeon 2.8 GHz system. This is a pity because in a test tourney I played "France vs. The World" the Naum results are better. This tournament I played on Athlon 3.2 with ponder = off and with a second machine Pentium IV 2.67 GHz, ponder = off.
Pentium IV 2.67 GHz, 40 moves in 20 minutes
France vs. The World I
|01||Pharaon 3.1||24.0/34||· ·||11||10||=1||00||01||10||11||1=||10||=0||11||11||=1||01||11||11||11||379,25|
|02||AnMon 5.50||21.5/34||00||· ·||11||=1||11||00||11||=0||10||11||11||=1||=1||=0||01||=1||10||=1||355,25|
|03||Tao 5.7 Beta||21.5/34||01||00||· ·||00||11||11||11||0=||=0||=1||1=||11||==||0=||11||11||1=||=1||341,50|
|04||Naum 1.4||21.0/34||=0||=0||11||· ·||11||=0||=1||11||01||=0||0=||11||==||=1||=1||=1||=1||==||346,00|
|05||ETChess 22.10.04 Beta||20.5/34||11||00||00||00||· ·||==||0=||11||==||10||01||1=||01||1=||11||11||11||11||312,75|
|06||The Baron 1.50||19.0/34||10||11||00||=1||==||· ·||1=||10||00||10||11||10||0=||1=||01||00||11||11||313,00|
|07||King of Kings 2.56||19.0/34||01||00||00||=0||1=||0=||· ·||1=||11||10||10||=0||=0||11||1=||1=||11||11||291,75|
|08||Ktulu 6.0||18.5/34||00||=1||1=||00||00||01||0=||· ·||01||1=||10||=1||=0||11||10||1=||11||11||281,50|
|09||Fruit X 07/05 Beta||18.0/34||0=||01||=1||10||==||11||00||10||· ·||00||10||0=||10||11||1=||1=||1=||10||293,50|
|10||Zarkov 4.70||18.0/34||01||00||=0||=1||01||01||01||0=||11||· ·||0=||=1||01||0=||01||11||01||11||285,50|
|11||Patriot 1.30||16.5/34||=1||00||0=||1=||10||00||01||01||01||1=||· ·||0=||10||11||==||==||00||11||275,00|
|12||The Crazy Bishop 0052||15.5/34||00||=0||00||00||0=||01||=1||=0||1=||=0||1=||· ·||11||00||1=||11||11||0=||235,25|
|13||Dragon 4.6||15.0/34||00||=0||==||==||10||1=||=1||=1||01||10||01||00||· ·||00||=0||11||01||0=||253,25|
|14||Movei 00.8.291 Beta||15.0/34||=0||=1||1=||=0||0=||0=||00||00||00||1=||00||11||11||· ·||00||01||11||1=||237,25|
|15||Nejmet 3.07||14.5/34||10||10||00||=0||00||10||0=||01||0=||10||==||0=||=1||11||· ·||01||0=||1=||234,25|
|16||Fafis X 0.01 Beta||10.5/34||00||=0||00||=0||00||11||0=||0=||0=||00||==||00||00||10||10||· ·||==||11||159,00|
|17||Capture R1||10.0/34||00||01||0=||=0||00||00||00||00||0=||10||11||00||10||00||1=||==||· ·||01||158,00|
|18||KnightX 1.86||8.0/34||00||=0||=0||==||00||00||00||00||01||00||00||1=||1=||0=||0=||00||10||· ·||131,00|
306 games played / Tournament finished
Tournament start: 2004.11.14, 23:49:28
Latest update: 2004.12.07, 20:30:59
Site/ Country: Schweich, Germany
Level: Turnier 40/20
Hardware: Dual Mobile Intel(R) Pentium(R) 4 CPU 2.66GHz mit 503 MB Speicher
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
Axon Benchmark: Axon-Benchmark-4 (00:00:38.9 sec.) [2367545 pps.] PowX= 7.376
Tournament with the same participant engines on
Athlon 3.2, 40 moves in 20 minutes!
France vs. The World II
|01||Tao 5.7 Beta||25,5/34||· ·||00||10||11||10||11||01||1=||11||=1||1=||11||11||=1||11||11||=1||01||404,00|
|02||Patriot 1.30||24,0/34||11||· ·||1=||10||=1||1=||11||=0||00||=1||00||01||1=||11||11||11||11||11||388,75|
|03||Fruit X 07/05 Beta||23,5/34||01||0=||· ·||11||01||=0||01||1=||11||11||11||1=||11||=1||=0||==||=1||11||381,50|
|04||Naum 1.4||22,5/34||00||01||00||· ·||11||11||=0||11||=1||=1||=0||11||11||11||10||11||=0||11||348,75|
|05||Pharaon 3.1||19,5/34||01||=0||10||00||· ·||1=||11||01||=0||==||11||00||01||=0||1=||11||11||11||306,25|
|06||Zarkov 4.70||18,0/34||00||0=||=1||00||0=||· ·||10||11||=0||01||=0||=1||11||11||11||01||01||10||277,25|
|07||AnMon 5.50||17,5/34||10||00||10||=1||00||01||· ·||10||01||1=||10||11||0=||11||11||==||=0||0=||284,75|
|08||Ktulu 6.0||17,5/34||0=||=1||0=||00||10||00||01||· ·||=1||==||11||11||00||=0||1=||11||=1||10||277,50|
|09||The Baron 1.50||17,0/34||00||11||00||=0||=1||=1||10||=0||· ·||11||0=||0=||10||1=||01||10||11||=0||276,00|
|10||ETChess 22.10.04 Beta||16,5/34||=0||=0||00||=0||==||10||0=||==||00||· ·||=1||=1||=1||11||11||0=||=1||10||251,00|
|11||King of Kings 2.56||16,0/34||0=||11||00||=1||00||=1||01||00||1=||=0||· ·||10||=1||01||00||01||=1||=1||268,00|
|12||Movei 00.8.291 Beta||14,5/34||00||10||0=||00||11||=0||00||00||1=||=0||01||· ·||01||=1||01||01||1=||1=||225,25|
|13||Dragon 4.6||14,5/34||00||0=||00||00||10||00||1=||11||01||=0||=0||10||· ·||00||11||1=||10||11||217,00|
|14||Capture R1||12,5/34||=0||00||=0||00||=1||00||00||=1||0=||00||10||=0||11||· ·||0=||=1||1=||==||193,50|
|15||Fafis X 0.01 Beta||12,0/34||00||00||=1||01||0=||00||00||0=||10||00||11||10||00||1=||· ·||11||00||01||194,00|
|16||Nejmet 3.07||12,0/34||00||00||==||00||00||10||==||00||01||1=||10||10||0=||=0||00||· ·||11||=1||185,00|
|17||The Crazy Bishop 0052||11,5/34||=0||00||=0||=1||00||10||=1||=0||00||=0||=0||0=||01||0=||11||00||· ·||10||191,00|
|18||KnightX 1.86||11,5/34||10||00||00||00||00||01||1=||01||=1||01||=0||0=||00||==||10||=0||01||· ·||186,50|
306 Partien gespielt / Turnier beendet
Beginn des Turniers: 2004.11.16, 02:05:35
Letzte Aktualisierung: 2004.12.08, 19:05:03
Ort/ Land: Schweich, Germany
Level: Turnier 40/20
Hardware: AMD Athlon(tm) XP 3200+ 2315 MHz mit 1.023 MB Speicher
Betriebssystem: Microsoft Windows XP Home Edition Service Pack 2 (Build 2600)
Axon Benchmark: Axon-Benchmark-4 (00:00:30.6 sec.) [3009722 pps.] PowX= 9.377
I found a big bug that can explain why Naum performed bad in ATL when ponder is on. Naum will actually reset itself every time there is a ponder miss (when it doesn't predict opponent's move correctly). This means that Naum will waste time cleaning all hash tables, and then it will waste more time filling them again, and of course with empty hash tables it will have to think much longer to reach the same depth. I am sure this reduced Naum's strength by around 100 ELO points. This happens only for UCI, so it was better to run Naum as WinBoard engine, but it's too late now :)
Naum is in my opinion a little sensation! The first version of Naum was really strong and today, only one year later, the users like Naum and you can find a lot of tournaments with your engine. How long are you working on Naum? Could you write a little bit about the main improvements to Naum, better ... could you give us the information which programming technics you used is the reason for the high ELO jumpings since Naum 1.0?
The reason for the first version already being strong is simple. I waited until I reached the playing level I was happy with. I hate to lose, and didn't want to release version that will be beaten by everyone. I also wasn't aware until a year ago that there are so many engines and tournaments. I started developing Naum in 2001 for Palm. There is still a Palm version available on my web site that I am proud to say fits it's code, data and transposition tables in less then 256KB of memory. It was hard to develop and test the Palm version, so I decided to create a PC version in 2003. This first version lost nunn match against Gerbil 11:3, but it was very basic and full of bugs. On the PC, development went much faster, and with the info I found on the web, it was easy to implement null moves, IID and other standard stuff. At the end of 2003 I already had a version that was close in strength to Arasan. Naum 1.0 already had all standard search techniques (quiescent search, TTs, extensions, null moves, IID,...) except SEE which I added in version 1.4. Naum wasn't using bitboards (now I use only two bitboards for piece location for both sides). The 1.6 is the first version after 1.2 that I worked hard on (probably 6-8 hours a day while at work :), because I felt that many competing engines are closing in.
From Canada there are three very strong programs available. Naum, King of Kings and Thinker. Do you know the programmer of King of Kings and Thinker, perhaps you have a friendship with the programmers and from time to time you have a meeting with the programmers? This question have a reason Aleksandar. Such a friendship can be the result of a new Canada-ch, perhaps in co-operation with the programmers from USA.
I don't have any contact with other canadian programmers. I think it would be cool to meet them in person at some kind of North American championship just like German Paderborn or Dutch-ch. Any sponsors?...organizers?
Unfortunately, I dosen't try with full power to find out sponsors in the latest years. I have a very interesting contact to a cosmetic concern. This company have intereset on our area and in my opinion is the combination Cosmetic / Chess very interesting. Yes, it's possible to find sponsors for such events, why not! Which such talents I saw in your person and many other engine programmers we can make a lot. For years I organized a "bigger" event -> Deep Shredder vs. The World <-. I wrote mails to Intel and Sparkasse Trier for sponsoring. Both companys had bigger interest and I made a little deal. This is one example only. Such an organization need many time and must be really an event for users. I can give my knowledge and helps persons which try to organize such an event. Unfortunately, I have to give my bad English too :-)
The programmer of Gandalf, Steen Suurballe, like the time control 40 moves in 20 minutes. I am sure you look in the Naum games you can find in WWW. What is the time control you like for testing? I am sure you haven't many interest to search in blitz games Naum errors!
Even blitz games can point to some big problems in the evaluation, but I can easily run them myself. Getting the long time control games is much harder, because I don't have enough free CPU time available for that. That's why I download from WWW and keep only games with 40/20' and up. Long time controls in my oppinion point better to more sophisticated evaluation, king safety and endgame problems. I consider 40/20' to be a perfect level in between blitz (high margin of error) and long (too slow and boring to watch) time controls, but of course the longer time control and faster the hardware, the more value the game has for me as a developer.
Back to Naum. What is your main interest in questions of playing styles? Do you like to create a tactic, strategic or positional engine? After all I can see, your engine are very balanced. The endgame seems to be a problem for Naum too but the most of all available engines have here the same problems. It would be better to have strong chess player in the team? A strong chess player can give a lot of tips for endgame improvements. Are hints from chess players improtant for a programmer to make the endgame better? Do you have a good test team or do you test your engine in self work?
My favourite engine is Hiarcs. I really like positional play of this engine and the way it squeezes the life out of opponent before launching a crushing king attack. I would like to make Naum play like that. I don't think Naum's endgame is worse then in other engines, but, yes, it's bad. Endgame is very hard to evaluate, because I don't think there are many general ideas that apply in all endgame positions. For instance, it's possible to have a position where king on e4 means win and king on d4 loss, because of some zugzwang or specific pawn structure. It's impossible to evaluate something like that. Sometimes you just have to do the deep search, so I think smart extensions are as important in endgame as a good eval. Even something as well known as opposite color bishops works in limited number of positions, and when I tried the simple version of it, I actually got slightly worse blitz results, so I decided not to use it for now. I like when people send me email and point to some problem with the engine, but mostly I do the testing myself. I usually don't have time to go over Naum's games, because I can't have chess board on my screen at work :) (I work on Naum at home only during vacations). Having a strong player to help would be nice, but i don't think it's essential.
I have to see Naum in my new ATL-2 tournament and I believe in around 4-6 weeks I can add your engine.
Now an user typical questions?!
Is the next Naum version with the improvements you gave us in perhaps two months available?
Before releasing a new version, I want to test history pruning that I added recently and improve my king safety. The king safety is currently very fast, but too simple and gets Naum in trouble quite often. I hope I will have time to do some of this and produce at least a beta version by the end of the month.
The latest question for the moment :-) What is your opinion about all the strong free sources? Pepito, Fruit, Phalanx and much others are free available. Do you look in the sources from the stronger available chess programs? If so, could you explain why Phalanx is so strong in King attacks :-) Other programmers find nothing to explain it. I believe its not easy for a programmer to look in other sources and I believe the complete sources understand only the programmer which made the programming, or?
I think that open source programs provide shortcuts to some authors, and I would be happier without them. On the other hand, nothing can replace experience and knowledge you get when you do something yourself rather then just take it from other person's code. You may be able to create a strong engine quickly, but you will never make the top engine that way. Nothing can replace creative thinking. I think a good programmer with chess programming experience shouldn't have any trouble understanding other people's code. I took a quick look only at the search algorithm of the Fruit and Glaurung. Never looked at Phalanx code. Actually I would like to know why is Fruit so damn strong. I didn't see anything special in its search algorithm, so it must be the evaluation. I really have very little interest in other people's code. I have so many ideas of my own that it will take years before I run out of them. I have a great respect for early engine authors (Phalanx, Arasan, Comet and Crafty come to mind). There is so much info about chess programming right now (even without the open source programs) making it much easier to create a strong program.
Thanks Aleksandar for this very interesting interview.
I am sure we alll wish you good luck with Naum for the future!!
Yes, one sympathic programmer more ... loud thinking, sorry!
78, MDFrank has asked me to post this message in the news, what I am doing right now. Unfortunately Frank is ill so he will not be able to update this webpage in the next two weeks. After this period everything will return to normal. In the mean time we wish you a lot of fun with Arena.
From Frank Quisinsky:
Thanks for your helps Michael. I will add News-Ticker message 79 only. An interview I made with the programmer of Naum for different days. I believe end of the next week I can work here with more power. At the moment I am ill and be happy with my rusk. All is OK, but I am not able to sitting to long on the PC.
Home of Matacz
Interview with Maciej Pestka by Frank Quisinsky
Hint: Matacz is not tested by myself yet.
I saw that you have released a new version of your engine. The CCT-7 results are very impressive and we all have to check your engine.
I think I was a little bit lucky the firs day of CCT7. I also noticed that Matacz seems to be better at long time controls.
For some days Grezgorz wrote a little bit about chess in Poland in our News-Ticker. Your program is one of the strongest engines from Poland. For the moment the wishes by users are more engine protocol support in Matacz! WB feature: Edit mode, analyze mode or Nalimov endgame database support. UCI support is just great and for users more easy to handle. Do you have interest to add more protocol support in the near future?
First of all I'm very happy that somebody noticed my program :) I will add more protocol support as soon as I find some time (My 2nd child was born the day of MPPS 3 final - 5 months ago - and since then I have almost no time for my Engine). But my priority number one is now a new opening book which is the weakest point of my engine.
Could you write a little bit to the program structur of Matacz?
The program is written in C++ (or rather in C using C++ syntax). The most detailed description you can find at my webpage. You can find there are all commonly known techniques that Matacz uses. I can only add that Matacz has very small eval (almost the size of Lazy Eval comparing to other engines). It's mainly because of my lack of knowledge about position evaluation at chess (I'm very weak chess player). If you look at the webpage you can notice that many standard techniques are missing (futility pruning, extensions, reductions etc...) some of them are very easy to implement add but needs time to test & tune.
With a logo by Wilhlem Hudetz your program is 50 ELO stronger :-)
I will send Wilhelm a copy of this mail!
Thanks for your nice engine!
Thank you for your interest.
It's good to know that somebody uses my program :)
Stefan Meyer-Kahlen (Germany, programmer of the number 1 'Shredder') added today an very interesting review on HIS OWN News-Ticker. Yes, all what you read is right, Stefan takes time to create a Shredder News-Ticker for his customers. Here you can read the information about his developements.
F A N T A S T I C
I hope other programmers of commercial engines exemplify and will follow!
I hope Stefan will find a lot of translators for his Shredder Classic GUI. In Shredder News-Ticker you can read that a French translation is available. Perhaps one of the hard working Arena "Translators" can give the Shredder Classic GUI the first look and help Stefan too. In my opinion we all have to give the programmer of UCI the max. possible support.
We are very happy that the strongest chess program today 'Shredder' works well under Arena Chess GUI!
And now ...
Sprint to the webpage of Stefan!
v. Blitz WV
WB / UCI compatible, own GUI
Home of SlowChess
Interview with Jonathan Kreuzer by Frank Quisinsky
Hint: SlowChess Blitz 0.4 is number 19 in ATL-1 Rating-List
At the moment three different versions of SlowChess are available.
1. SlowChess 2.96, 2. SlowChess Blitz 0.4, 3. SlowChess Blitz WV
Could you explain a little bit the differents of these available versions?
I believe that the "Gladiators" don't know which of this versions is to add in a tournament.
Slow 2.96, is mostly the same as 2.94. (There were a few tiny changes, hence the new version name.) So it's the oldest of the three. I released it so anyone interested could look at the source code for an older version of Slow Chess. The Blitz versions are newer, and I would consider Slow Blitz WV as a new engine compared to the 2.0-2.96 series, as most of the code relevant to playing style (evaluation & search) has been changed. Slow Blitz 0.4 is an intermediate version. New in Blitz 0.4: New King Safety, many other evaluation changes, like passed pawn, bishops, etc. A few more threats & a few types of checks in the Qsearch. New in Blitz WV: Mobility, some square control (not used much though currently) more new King Safety, endgame knowledge changes, more highly selective search, more threats recognized by the search, etc. Also much slower nps.
Which of your SlowChess versions you like it to see in engine - engine tournaments?
For tournaments it's up to the the director. If they want to use the newest versions, then they should use WV. Otherwise they can run any version they want, it's okay with me even if they use 2.82. Or Blitz and 2.94/2.96 if they want to do more than one engine version by an author.
In ATL-1 tournament SlowChess Blitz 0.4 have an really great performance and is the strongest available chess program from USA. It seems that you develops SlowChess very fast. Do you await for the next versions of SlowChess the same jumpings? Perhaps not all of your main ideas are implement so far?
What are your latest main changes in SlowChess? Could you give us more details about your latest versions of SlowChess? Since SlowChess Blitz 0.4 your program is a lot stronger. I think on around 100 ELO points more ...
I'm guessing there will be less of an ELO change in new versions now that Slow Chess is stronger. I've noticed the difference in rating between versions varies greatly in different people's lists. I think I've covered the main changes in previous posts. By more selective search I mean more aggresive null-move with adaptive R, extended/futility pruning, some Q-search pruning. Threats are tactical things, like Mate-in-1 threats, or forks/skewers, seen 1-ply or more earlier thanks to detection, or in the case of mate threats additionaly used to extend the search. The only special type of move in the Q-search is possible backrow mates, so it's not as tactical as I'd like. There are also some new GUI were features in WV (such as analysis window with multi-variation analysis.) Internally in WV there's personality support with many different settings possible, but there's no personality editor yet, so that's something I hope to make.
Opening expert Harry Schnapp (Germany) sent today the "Little Mainbook". Please read News-Ticker Message 18 and 51 for more information. The "Little Mainbook" is added in Arena Setup 2. Additional information to our mainbook can be found on the detail page!
News-Ticker Overview (it's easy to search a message, try it)
Arena Little Mainbook:
270.524 moves, depth = 25 plays!
Little Mainbook v. 1.0 (1.03 Mb, RAR format)
If two different engines will use this book in engine tournaments the book is very able.
The variants will be no longer as 16 - 25 half moves.
In the latest day many versions of stronger available WB / UCI engines are released.
01. UCI / WB Ufim 6.0
02. WB Matacz CCT 7
03. UCI / WB SlowChess Blitz WV
Links to the engine programmers can be found under Engins, Links
No possible to try to find out from all engines information in detail (the time to do that is the problem). But I wrote some mails to get more information for our News-Ticker system. From time to time you will find information to different engines. See the short interview with Jon Dart (Arasan) for an example.
04. UCI / WB Spike 0.9
The Spike programmers sent today a new beta of Spike 0.9 and an interesting text:
Spike 0.9 is not yet fully finished as we still have a problem with the chessbase-GUI in ponder mode. We will try to fix this problem for the release. We will make a version availiable for download after our hollidays. This version plays exactly like the version we used in paderborn. There are only some improvements in the interface handling that bothered us while playing in tournament. Spike now has some new parameters that can be set in winboard (config-file) and in UCI (config-file and UCI-Parameters). These are search parameters like extensions and prunings.
Spike 0.9 will be available in around two weeks!
WB compatible, own GUI
Home of Arasan
Interview with Jon Dart by Frank Quisinsky
Hint: Arasan 8.1 - 8.3 is number 47 in ATL-1 Rating-List
Changes in Arasan 8.4:
1. Bug fixes in pin detection.
2. Scoring changes.
3. Better lazy eval in scoring module.
4. Improved pawn race scoring, handles more cases.
5. More liberal use of null move in endgames (but less aggressive use of R=3).
6. "make install" target added to Linux Makefile.
7. use feature done=0/done=1 to bracket tablebase initialization.
Arasan is one of the first available amateur engines, one of the first five engines I found in WWW in the time I started with my first internet connection end of 1996. How long are you working on Arasan? In which year your program made the biggest jumping in playing strenght and which new idea by yourself can be the reason?
I wrote an earlier chess program even before Arasan (first in Pascal but then ported to Modula-2), but I never released it. I started Arasan partly to help me learn C++, in the early 90's I think. It was first released in 1994.
I am not sure when the greatest jump in strength occurred. It has been gradual. Anything before version 5.0 was pretty buggy. By version 5.4 (May 2000), I was doing a lot of automated testing and the code was a lot more stable. I think that's the first version that was respectable in terms of quality and playing strength. Version 6.0 added tablebase support, which doesn't often matter, but sometimes helps a lot.
Today over 250 amateur engines are freely available. Much of this engines started directly with a very high ELO performance in a _shorter_ time of developing. What do you think about all the "young stars" and the amazing hight ELO performance by others? After all, do you think that the available information about chess programming for around 8 years are not enough for young programmers to create such monsters? Means, that in the latest years more sources are free available and more pages about chess programming. I am sure this made it today possible to create in a short time a strong program?
Jon Dart, programmer of Arasan (USA)
I am surprised and occasionally discouraged when very new programs show up and are very strong. But I think it is easier now to make a strong program, since there is more information available - when I started out there were some academic articles and books that were mostly short on practical details, plus there was source to Gnuchesss and a very few other programs, mostly not very strong.
I mostly don't know what the new programmers are doing, unless they release source or technical details. I have looked at Pepito, which is a strong program, but don't see anything that is particularly unusual about it - so I am not really sure why it is so strong. Aristarch from its web pages seems to have some unusual forward pruning ideas, as do a few other programs.
The older guard of testers like more the older guard of engines. Many testers have the opinion that a good program can regulary "only" stronger and stronger with time. Many little improvements are more authentic and we can be sure that the programmer try to use really the own ideas. Arasan is perhaps the best example we have today. Many little improvements and today in additional a strong program. With each new version of Arasan we can be sure that you found a little bit to make your program stronger. Do you know that a chess program loosed personality if the first release version is directly with a high ELO performance available?
Not sure I know what you are asking. Arasan has changed a lot over the years and I am not sure it has maintained a consistent playing style, although it has generally been more solid than aggressive - I can tune the eval so it attacks more readily, but I haven't found a way to do that and also make it not play stupid sacs sometimes.
What are your main ideas for Arasan in the future of developing? With years the programmers write big parts of the code again and have long time ideas for make the programs stronger! Any programming technic Arasan must have in your opinion? Example, many programmers I spoken means that MTD is an art of chess programming and not easy to implement. What is in your opinion not easy in chess programming and very interesting to have it in the own program?
I would like to implement SMP support, especially with the new dual-core chips coming out. But from comments I have heard from Bob Hyatt and others, that is not easy to do without also introducing bugs.
I believe you test Arasan yourself with positions only and don't like to play eng-eng matches (not sure). Testers like it to play engine-engine matches. Do you look in such matches by testers? Or do you wish that the users search himself in engine-engine matches and send results / analyses to you. The most programmers have bigger problems to found good and long time beta testers. In my opinion are all the available engines the reason!
I have found that having Arasan play a wide variety of opponents on ICC has been a very good way to test it - I have found and fixed a lot of problems after looking at its games. I also use test suites and I sometimes play engine-engine matches offline. I'd be happy to have external testers send me games, although generally I also need the engine logs to get the most use out of the games. I have sometimes used beta testers, but not recently.
You develops an own GUI. Today much good GUIs are available for testing, like Winboard, Shredder Classic and of course Arena. Do you try other graphical user interfaces? Maybe you test your own engine under your own GUI only? Do you have a bigger interest to improved your own GUI in the future or is the lion's share the engine?
Arasan's own GUI isn't anything I am particularly proud of, but I think it has helped make Arasan popular, because you can install the program and start using it right away, without any setup. So I maintain it. I do however spend most of my time on the engine. I use Winboard also and Arasan supports that well. I don't use Arena myself. I do use commercial programs such as Chess Tiger with their own UIs.
Today we have two "standard protocols", Winboard and UCI are very popular. Arasan have an exessively good support to Winboard. I believe this is enough for you (of course should be enough for the group of testers) but UCI have some interesting features. To reduce ELO in strenght is a nice option for beginners. What do you think about the UCI protocol by Stefan Meyer-Kahlen? Do you give UCI a first look?
I do think the Winboard protocol is not very easy to make work properly and reliably (at least I had trouble with it) - UCI may be better. However, I have Winboard working ok now and Winboard supports the features I need myself. If I get enough requests for UCI support, I may implement it, but right now it is not a priority for me.
In times you have no interest to play chess or to made improvements on Arasan what do you do if you are sitting on your PC? Do you have a favourite other game or other main interests? I like Kyodai Mahjongg, to work a little bit on webpages and find out information to computer chess, like hardware and to test different new tools I can find on Freeware / Shareware pages.
I don't really play other computer games besides chess. I keep some music on the PC, and download some from iTunes and other sites. I try to keep up with technology and general news on the Net.
At last, we need more information about your latest version of Arasan. What do you try to improved in detail in the latest available Arasan versions, perhaps since version 8.0? You know that we all like to hear the details which must come direct from the programmer :-)
Compared to earlier versions, version 8.0 had a re-written and simplier eval, which gave it a pretty large speed boost. I have also started using the Intel C++ compiler for builds, which gives some additional speed advantage. Version 8.0 also introduced a small evaluation cache.
More recently, I have made quite a few endgame scoring bug fixes and improvements over the last few releases, and implemented KPK bitbases in version 8.2. Version 8.4 has some other scoring changes, including a larger penalty for uncovering the king (or bonus if the opponent's king is uncovered) and a larger bonus for rooks on open files. In addition, 8.4 has some changes to lazy evaluation, which improve speed. Overall, compared to version 7.4 (early 2004), version 8.4 is about 33% faster on a typical middlegame position, although the speed improvements have been gradual and there was no one thing except the eval simplification that helped a lot.
The most problems in all the available ratingslist is, that the results based on ONLY one hardware. Shredder 9.0 will be 20-30 ELO stronger on Athlon hardware compare to Intel hardware for only one example. This made testing of engines not easy. After all I saw in the latest time is the CEGT ratinglist the most interesting ratinglist we have today. CEGT based on results on five machines! Different actual systems and a very good time control made this list in my opinion to a reference. Heinz van Kempen and Christian Koch tested at last SOS 5.1 for Arena by Rudolf Huber. SOS 5.1 is after a short time of testing with more as 300 games the strongest available freeware. This is the advantage of many systems.
THE NEW REFERENCE ?!!
Not easy to find additional information about the IPCCC 2005 in Paderborn. After a longer time without news I found on CSS News-Ticker a message!
Perhaps participant programmers will send me information for our webpage?
I believe not many users visiting the tournament in Paderborn. On the official webpage you can find the final results and tables. Unfortunately, the games are not to 100% complete!
Hydra by Dr. Christian Donninger (Austria) won again the IPCCC with 8 of 9 the tournament. On ranking two Deep Shredder 9.0 by Stefan Meyer-Kahlen (Germany) with 7.5 of 9 and on ranking three Gandalf 6.0 by Steen Suurballe (Denmark) with 5.5 / 9. Very strong are the forth place from Spike by Volker Böhm & Ralf Schäfer (Germany) with 5.5 of 9. For me a little sensation that Spike have in the first both tournament (CCT-7 and IPCCC 2005) such good results. On ATL-1 ratingist Spike have a good place and to the lastest release version Spike is improved more as 100 ELO!
Rudolf Huber, SOS 5.1 for Arena
IPCCC 2005 (official webpage):
IPCCC 2004 (detail page from ChessBase):
Not the actual IPCCC tournament but you can find on this very nice site from ChessBase some information and pics about the participant programs from the last year. Twelf of the participant programmers are the same this year.
2004: With Yace (3/7), Comet (2/7), Fritz (5.5/7) and BlackBishop (1/7) = 11.5 of 28 = 41,07%
2005: With Spike (5.5/9), Jonny (4.5/9), Nexus (4.5/9), Neurologic (3/9) = 17.5 of 36 = 46,61%
This year the IPCCC 2005 was stronger as the year before!
Spike: http://spike.lazypics.de/, Shredder http://www.shredderchess.com
Left to right: Ralf Schäfer (Spike), middle Volker Böhm (Spike) and right site
ACHILLES he called ... Number -1- Stefan Meyer-Kahlen (Shredder)
Of course, the Spike team played under Arena Chess GUI.
The next Spike release version will be available after programmers holiday in around one month!
March 01st, 2005
More pics can be found under:
http://www.team-oh.de/ (Timo Klaustermeyer)
Congratulations to the team of Hydra and Spike!
1. DGT released a new setup of the collection of main programs. Queen 5.0 (not the chess engine Queen by Leen Ammeraal), Toma 1.12a, DGT eBoard driver 1.32, WoodIn 3.0, and the DGT board tester. The download file of this DGT Board Setup 1.2 is only 7.63Mb.
The "new" DGT webpage have a nice look, you can find all information about the DGT chess boards, different pieces for the DGT boards and about DGT clock XP.
2. Very interesting is the message "Kramnik opens DGT Chess lab" on DGT and ChessBase webpage.
Read it: http://www.chessbase.com/newsdetail.asp?newsid=2173
Hermen Reitsma and Ben Bulsink (DGT projects)
3. Arena have really a fantastic native support to DGT Board / DGT Clock XP. In Arena 1.1 Martin Blume added the fully support for the Clock XP. Since the first available version of Arena, for around 2 1/2 years, Martin added many interesting features. I will write soon a review about the DGT Board / DGT Clock XP in combination with Arena Chess GUI in German. This need a little bit time.
The Chess960 cube by Reinhard Scharnagl found the right place :-)
Maxdata Pentium IV 2.67 Mobile Notebook,
DGT Chess Board Serial, DGT Clock XP
SiteCom USB adapter (my Notebook don't have a serial connection)
My blue Allnet Wireless Router have nothing to do with DGT :-)
Two pics of my first DGT Clock XP support test. A little bit cabel chaos and the cigarette on the pics isn't very nice. But you have to look on the DGT clock XP. You can see the evaluation from Gandalf 6.0. This is just great if you play a game on the DGT Board in additional with the DGT Clock XP. Just chess computer feeling, must thinking on Tasc R30. Only Arena GUI supported this feature to show the evaluation from a chess engines. We have much more ideas to show information on the DGT Clock XP.
68, FQ, GS
I asked the programmer of Armageddon Grzegorz Sidorowicz (Poland) for more information about computer chess in Poland. Grzegorz sent me much very interesting information. I add parts of the mails with permission by Grzegorz in my News-Ticker ...
Currently I build new engine called Drwal. It will be open source program. His move generator is 5 times faster than Armageddon engine. But it doesn't mean end of Armageddon developing. Drwal engine I will put in Armageddon insted Armageddon's engine. It means new Armageddon version will be called Armageddon Chess III. I don't write about my plans because after my big lose in Polish Computer Chess Championship I want to take revenge. It can be very difficult because Polish Chess Programs making big progress. I'm ambitious kind of person but in case of another lost I will have to stop farther developing of my programs. I think about it and in case of my lost I want to leave very nice and solid open source program which can help to other people. But I'm not in a hurry it can take me even anothe one year. I want to make program which can use many processors and many machines. I work realy hard (I do it faster than Arena translation :-))
And now some information about Polish coputer chess. Polish computer chess environment it is realy lively collection of people who work very hard. Polish chess engine have been very easy to notice at CCT7 and final results are very near to realistic. In my humble opinion Gosu is at the moment the strongest Polish chess engine but I know personaly author of Tytan and he is very slow but also he is very solid people who can make suprise in every moment. On the last Polish Computer Chess Championship author of Butcher said he want to rewrite his program and I think in consequence Butcher can occur much stronger in near future. But all this program about I talk are well known. We have also some unknown programs who can also make big suprise. I think mainly about Atak. This program have got fastest move generator in Poland and who now maybe in world also :-) His results in WBEC aren't good but I know his author and I know some other results. This program without any knowledge killers, transposition table, and any prunning or advenced ordering techinques make not bad results. We must remember final program will have probably everything. Every year in Polish computer chess is something unusual and I think we can expect something this year. I believe Polish computer chess have future. I will guard it but as long as we have people like Maciej Szmit who is organizer of Polish Computer Chess Championship and person who produce new computer chess programmers I am calm about our future.
Perhaps you can give me the information which programs from Poland are at the moment the strongest.
Gosu. I'm absolutly sure. In every my tests and my colleagues always win Gosu. But diffrence aren't big and every thing can happend this year.
I asked for the date of the next Polish-ch ...
June, But exact date is unknown. We have some battles about rules of our champioship but I believe Maciej will still want doing good work.
Thanks for the translation, you make me happy !!
You make me too. I think about text on your webpage ;-)
It is not quite readable but quite funy :-)
Ralf Schäfer (Germany) released Jose 1.35. Jose works fine under Mac OS X, Linux and of course Windows. Jose have strong database options, nice 3D look, a good engine support, like UCI II and is in different languages available. Jose is Freeware, Peter produce really a piece of art in the latest two years!
Have a look on the screens, the other information by Ralf!
The IPCCC 2005 in Paderborn is still running. The favourits Shredder, Brutus and Gandalf are on the first positions. Unfortunately, SOS losed today both games and with 3.5 / 7 SOS is on ranking 5 / 16. A lot of engines with 3.5 points also the newcomer Spike. Tomorrow SOS have to play vs. Gandalf. I made today some experiments with SOS and it seems that SOS is a tactical monster, in endgame very strong but with problems in positional positions.
Visit the official site in Paderborn!
You can find all results, you can replay and download the games!
For around two weeks I started ATL-2. I try to used the tournament system by Hans Walz. This system need a lot of time. I switches the rules and I will play now with 10 engines in each group. Each of the ten participant engines will produce 540 games with 40 moves in 20 minutes in around 3-4 months. In three- four months it's more interesting to start the next group of engines.
The ATL-2 pages are daily updated.
You can find for each match the log files and games in *.pgn format.
Here the first ratinglist after 440 games:
Program Elo + - Games Score Av.Op. Draws 1 Shredder 9.0 : 2700 39 39 200 62.0 % 2615 37.0 % 2 Gandalf 6.01 : 2630 40 40 200 49.2 % 2635 30.5 % 3 Ruffian 2.1.0 : 2612 48 48 140 46.1 % 2640 30.7 % 4 TheKing 3.33 Schumacher : 2607 39 39 200 45.0 % 2642 34.0 % 5 ProDeo 1.1 : 2606 50 51 140 45.0 % 2641 24.3 %
An other very interesting tournament (CEGT) by Heinz van Kempen and Christian Koch can be found under
This tournament on five computers is playing under Fritz GUI. Please have a look on this very nice page! In my opinion the best available rating list so far. For me its nice to see from time to time tourneys under other Chess GUI. The ELO ratinglist is very realistic. In my opinion is the hardware the biggest problem for all the different results we can find in WWW. On AMD different engines are playing stronger, others on Intel. The good combination of hardware and the time control 40 moves in 40 minutes will make this tournament to an event for us.
After my quick test SOS 5.1 for Arena can be 40-50 ELO better as the preview version SOS 4.0 for Arena. In this case SOS is with Aristarch, List, Ruffian 1.05 and ProDeo 1.1 the strongest free available engine so far.
At the moment the IPCCC 2005 is still running and SOS beat Matador in round 1, remis vs. Jonny in round 2 and again won vs. the Dutch amateur Diep in round 3. Unfortunately, Shredder loosed in round 3 vs. Brutus and now its not easy for Shredder to win the IPCCC again. In round 4 SOS will play vs. Brutus and all is possible because SOS like the NPS killer. It's just a great tourney, perhaps the most interesting IPCCC ever played!
I like this group of engines a lot!
I don't find the Shredder logo but the Achilles LOGO is good too :-)
You can follow the tourney on the official IPCCC 2005 webpage LIVE. Here you can found the games, tables and much more interesting information. Today three rounds are to play and we all wait of the results of round 4 and 5.
Again, we wish our main program and all other participant programmer good luck. For me very interesting the results of Spike, this engine is clearly improved and have with 2/3 also a good start.
AGAIN: Special hint for Achilles :-) / Shredder programmer Stefan Meyer-Kahlen:
Movie Troy (Summary, Capital three ... in German):
"Achilles, Achilles ... Schau in die Gesichter dieser Männer, Du kannst hunderte (15) von Ihnen retten, mit einem einzigen Hieb Deines Schwertes (Endspieles) kannst Du diesen Krieg beenden, lasse Sie zu Ihren Frauen (Rechnern) heimkehren"
GO Spike GO, means GO SOS GO ...
The programmer of Armageddon Grzegorz Sidorowicz (Poland) and one of the organizers of Polish computer chess events sent me the Polish Arena 1.1 translation. This is just great because each translation is really one of the biggest events for us. We have many Polish visitors / Gladiators here. I have a Polish / Russian name, speak very very bad Polish. My mother comes in wartime from Silesia (Schlesien) to Germany. I am nine years young as my Polish neighbor learns me chess. With my Polish friends in school time I played the first computer chess games (my first chess computer Chess Challenger 7 vs. the Chess Challenger 10) and my computer won the most games :-)
Do zobaczenia przyjaciel Grzegorz!
Napisc do mnie dookola Polska komputer szachy informac(z)ja ... not sure believe with "c" ... very very bad, but I hope Grzegorz can read it!
Back to the main topic?
Got to Armageddon webpage and visit the great table with results of Polisch chess engines :-)
Today Matacz, Butcher and Tytan are clear improved ...
Back to the topic, second try ...
The translation can be found in User Files selection!
FOR EXPERTS ONLY
We are all experts :-)
You are a proud user of a lot of UCI engines. After all you know UCI engines have settings which are easy to config. Each UCI settings have an own window for the configuration.
Example: UCI Deep Shredder 9.0
Possible settings of the number 1, Deep Shredder 9.0 by Stefan Meyer-Kahlen
So far so good, a little bit publicity for Stefan :-)
Oh, back to the topic, means the secrets ...
Each UCI engine have a standard setting.
Much engines 1MB for Hash or ponder=on in standard setting, other engines ponder=off ... you know what I mean!
This is a glorious mess in UCI area!
You can give all UCI engines the same settings under Arena Chess GUI:
Engines / Managed / UCI (not new)
If you don't used this option you have to config for each engine the hash table size, tablebase cache, tablebase directory.
This can be a lot of work!
Perhaps you will make experiments in this area and you don't have interest to activate for each UCI engine the same settings?
The problem are the mentioned above standard settings!
And here is the secret:
You can config for all UCI engines YOUR settings without to load each UCI engine under Arena Chess GUI in one file!
Yes, this is possible ... with an editor like Notepad, TextPad or UltraEdit and the new BackUp option from Arena 1.1!
Create a new subdirectory:
c:\Chess\Arena ... in this example Arena files are on this position!
c:\Chess\Arena\BackUp ... create the subdirectory "BackUp"
Options / BackUp Arena settings
Save the engine settings.
Arena will save the engines settings from Windows registry.
You can find a file with the name: ArenaEngines.reg in your new BackUp directory!
Example, extract from my ArenaEngines.reg:
"History Move Number"="3"
"Quiescence Check Plies"="1"
Now you can edit by hand the settings from all UCI engines you have added in Arena.
If you are ready you have to saved the file ArenaEngines.reg
Options / Restore Arena settings
And you can set your changes in Windows registry. Reboot your system and ready!
Very easy ... THATS ARENA
Of course, you can save for each UCI engine your setting (personality, save as new engine).
Arena Chess GUI have in my opinion a very big healthy margin to all other available computer chess GUI (included are the commercials) in questions of engines compatibilty, engine configurations and a lot of others main feature!
The first beta of Arena 1.2 isn't in the next weeks available. Martin Blume made / will make some bigger improvements and tested so far other interesting bigger features. It's the wrong time to write more about the work by Martin. It's so far not possible to make changes in the available version 1.1. Martin made to many changes after the release version of Arena 1.1. To fix a small bug isn't possible. We hope of understandings!
After two months support by mail I don't got many bug messages. Version 1.1 works very stable. In the time Arena 1.2 Beta 1 will be available we have a hard work because we have to check the complete program and each option. The result will be a fantastic new Arena releae version end of the year 2005.
Furthermore, Martin works with full power in the latest six months each day on Arena.
He need a little break!
I will try to make in this time our News-Ticker system very interesting for you and hope of a little bit support by engines programmers. Information by engine programmers are very popular. The Gladiators like, for one example, the pic from Amyan programmer Antonio Dieguez :-)
I will write different reviews and Alex and myself are thinking on new interviews. On days I haven't many to do I will make a little bit Support here.
Some Gladiators reported that Arena crashed sometimes!?
Of course, this can have a lot of reasons. I need for such bug reports more details.
Try to create a debug file and please send me this file.
For me it's not possible to give you a good answer if you write: "Arena crashed, please help me"
Engines / debug - windows
as_debug (pic of the debug option, UCI Pharaon 3.2 is loaded)
I will try to give you different helps:
01. Each eng-eng match should be config as tournament. Please give each tournament or each new engine match a new name. If not, you overwrite the very very important *.at tournament configuration file and this should be a reason that Arena crashed. This is the mistake the most Gladiators made!
Engines / Tournament / Settings / New
02. A tournament which is over can be not start again. If you config:
Engines / Tournament / Start, Quit / Quit Arena after tournament
Arena isn't crashed, that is your setting! Please don't start a ready tournament again if you activated this Arena option :-)
03. During the tournament you stopped a running game. No problem for you, for me, for Arena and the others. But sometimes you cann't start the tournament again because Arena crashed. Please restart Arena and start an other engines from your list of engines. Then start the tournament again and the tournament will works fine. The reason can be the latest settings of the two engines which played the latest tournament game.
04. Restart engines after each game should be ON
Engines / Tournament / Start, Quit / Restart engines after each game
05. Restart Arena after so many games ...
I used: 40 moves in 5 minutes (20 games), 40 moves in 10 minutes (10 games), 40 moves in 20 minutes (5 games).
Engines / Tournament / Start, Quit / Restart Arena after so many games
The same as the -reuse WinBoard option!
Much engines have problems with unloading.
Some other good settings for tournament players ...
Options / Apperance / General
Continuing finished games is allowed can be a reason for crashes from WinBoard compatible engines!
Should be off if you will play a tournament with a combination of UCI / WB engines!
Engines / Managed / Options
With UCI Filter option different engines (example: SOS) works faster for ultra fast time controls.
The engines don't loosed important time for the moves.
Game in 3 minutes works fine in my test.
With game in two minutes some engines produced "loosed on time games".
Engines / Tournament / Options
Autoflag is important:
In order to continue a tournament if one of the engines doesn't play anymore, you should use "Autoflag". Then the engine runs out of time and at least the next game will begin normally.
The configuration of Engines is very importand:
RAM for hashtables, RAM for tablebase cash, RAM for tablebase using!
If you have 200 MB free you cann't give 128 for hashtables if you used 5-pieces tablebases! You can see with Crafty in textmode that the 5-pieces tablebases need a lot of RAM. Arena need around 15Mb. You have to test the settings before you will start an tournament or its possible that Arena crased if you gave the engine more RAM as you have.
Please check if the engines you are using are configured correctly. E.g. no engine should use too much hashtable space, otherwise Windows can not use fast RAM and has to swap to very slow hard disk space.
In a lot of the support cases engines are the reason for crashes under Arena Chess GUI!
Please try to reproduce a crash with engines like UCI Yace, UCI Shredder, WB Crafty!
Bevor you contact a programmer, you have to try to find out the problem by yourself. Please use the debug-windows from Arena Chess GUI, have a look in the readme files by the programmers and please have a look in the readme file from Arena!
60, FQ, VA
Text by Volker Annuss
When an engine sends too much data to arena in a short time, something is blocking and the CPU usage goes down to nearly zero. This does not happen on all machines, and I don't know the exact reason why it happens. To work around this problem I wrote the UCI-Filter program, that discards some of the data.
The UCI Filter option in Arena does something similar, so my program is obsolete now.
The old filter program with source code and a description is still available at
Arena menu system:
Engines / Managed / options
I used: Allways used UCI filter!