An idea for a new WCCC format - what do you think?

[Vasik Rajlich]

IMHO the two main things are having many participants and providing live coverage of every single game. Everything else is secondary.

Vas

This sounds like the CCT. What could be done to elevate the CCT to the stature of the WCCC?

The main thing would be to change the name. In this game, boldness pays.

You could make a decent argument that CCT has already been the top computer chess event for the past seven or eight years, but without a bigger title it won't get that level of respect.

In my opinion, open hardware is also a must for a world title. Otherwise, the competition devolves into the old microcomputer championships. It also ignores chess as the drosophila of computer research, in this case massive cluster and multi-core computing.

I tend to agree, but that's an implementation detail. The main thing is to decide these sorts of things by some sort of transparent, democratic process which reflects the wishes of the majority of programmers.

Vas

[Vasik Rajlich]

A simple solution to catching cheaters? Why not have everyone who enters the tournement be required to send a copy of the engine that they inend to use in the tournement to the director. In that way if someone is accused of cheating then the game could be replayed by the director using the engine supplied by the contestors. If the play did not match the game played then that player could be dismissed from the tournement.

You will flag every parallel engine as a cheater then.

Also, what are you going to do without the same hardware?

This does not work at all.

The non-determinism isn't a show stopper, but IMO requiring executables is too draconian. Of course, executables will need to be required from first-time participants, as is already the case for existing online events.

I do like the idea of requiring participants to submit their log files.

Vas

[Gian-Carlo Pascutto]

A simple solution to catching cheaters?

Another anti-cheat could be to require upload of log-files of thinking output for all (or a selected bunch) to see. Cheating then would still not be impossible but made a bit harder.

Stan

This is a good idea. To prevent or stop most accusations of cheating at a live event, I believe 2 things may be good enough:

1) The log files being available LIVE (not after-the-fact). Sufficient detail in the log files themselves.

2) A provision that the author provides the source code of the program to a mutually trusted third party if there are suspicions about the origin of the program.

[Gian-Carlo Pascutto]

You will flag every parallel engine as a cheater then.

Also, what are you going to do without the same hardware?

This does not work at all.

The non-determinism isn't a show stopper, but IMO requiring executables is too draconian. Of course, executables will need to be required from first-time participants, as is already the case for existing online events.

I repeat my question: what are you going to do without the same hardware? (And I don't mean in terms of speed...)

Non-determinism will be a complete show-stopper in something like a Monte-Carlo engine, or a massively parallel system.

[diep]

A simple solution to catching cheaters?

Another anti-cheat could be to require upload of log-files of thinking output for all (or a selected bunch) to see. Cheating then would still not be impossible but made a bit harder.

Stan

This is a good idea. To prevent or stop most accusations of cheating at a live event, I believe 2 things may be good enough:

1) The log files being available LIVE (not after-the-fact). Sufficient detail in the log files themselves.

2) A provision that the author provides the source code of the program to a mutually trusted third party if there are suspicions about the origin of the program.

Like most programs, diep logs every mainline, finishes of plies etc.

I remember i asked once for a logfile from junior.

"Which logfile?" was the answer.

Very helpful the answer was:
"we can provide a pgn though of the game".

Vincent

[bob]

Hello

1. It be handled over a internet chess server such as ICC, Playchess
or something else, This could lead to some level of sponsorship
via a fee to the server to show it.

I personaly dont like the idea of playing a world championship on a remote server is a good idea. As cheating is easy you will have a suspicious situation as soon as an "outsider" wins some games vs the big players. Fakeing an engine output or nodes is much easier than to write a top class chess engine. I think an on site event is preventing cheating much better than an remote server thing (nothing is 100% of course and cheating was, is, and will be always possible and happen, but you should not make it that easy!

Bye
Ingo

PS: Just think about an "inbeween" program which is randomly multipling nodes within a certain frame and cutting or mixing PVs - some say it is already excisting!

You do realize this already happens? "people" attend the event, but they use "remote" computers. So it is already easy, and playing on a server does not make it any easier than it already is.

I think that the main difference is the risk for the cheater.

If you discover that somebody cheats in a remote server then the person may easily hide and you do not know who cheated because people can use false name and later participate in another remote server event with a different name.

If you discover that outsider cheats in on site event then the outsider cannot hide and everybody saw the outsider in the event(assuming programmers need to be there) so the risk for the cheater is bigger(maybe the outsider can try to do plastic surgery in order to paticipate later in another on site event but it is clearly harder and I do not think that people are going to do it for cheating).

Uri

That's easy enough to get around. Not known? Can't enter. I would not let anonymous people enter anyway...

[diep]

There seem to be several constraints that keep the WCCC as is which
many complain is too long and too expensive.

...snip

Charles you're a nice guy, yet there is a lot of issues that needs to get adressed. So don't interpret my words wrong.

a) you live at a different planet than the majority which attends the world champs computerchess.

At the world champs computerchess there comes a few types of participants

a1) regular competitors or those who have a real strong engine built themselves and make a chance to win and/or do real well, based upon the current strong performance of the engine. All of them are parallel engines.

These guys show up to win and/or do well. A world title is important to win. For those who make a chance to win it, they put in a lot of effort, and some even pay a lot of money for hotel, tickets and machines or time invested to get good hardware. Nowadays this has become less than the past. Some years ago there was like 30 of these type of programs. The word 'commercial entry' is not appropriate to them, yet it is guys who have a lot of experience there.

You can write down the names of these guys already on a list and probably most of them will show up, when having the opportunity, in Pamplona, in random order the programmer names: Zwanzger, Huber, Isenberg, Pijl, Bushinsky, Rajlich/Cozzie, Meyer-Kahlen, Kolls brothers, Pascutto, Uniacke, Diepeveen.

a2) a few amateurs from far away show up. I'd call them just like group a1 "computerchess addicts". They are the real amateurs. usually this is just 1 or 2 a year at most. I remember 1 event where Charles showed up with Noonianchess in Maastricht. That was very much appreciated Charles!

a3) sometimes a participant who was real real strong a few years ago tries again to feel and smell how computerchess is nowadays, such as Ferret in 2001, with the expectation to revive his engine (though that doesn't usually happen). Of course right now it means they first want to parallellize their engine before showing up (that stopped Uniacke a few years also from joining). I'm hoping of course for a re-entry of Tiger. Obviously great programmers from some years ago have proven they can improve each time. If you do it once you can do it again.

a4) Government sponsored entries. Most semi-government guys only show up when getting paid. Good example is Le Fous Fous (the crazy bishop) and numerous clone entries. Some years ago there was engines like Zugzwang and Cilkchess and P.Conners, all 3 of them supercomputer programs. They all disappeared and only show up when their ticket and hotel gets paid. Also in this list falls open-source programs with local operators. Crafty is the only regular participant from this list. Though i had him on the phone some time ago i never met Bob in person! Also not 'coincidental' they usually only join when such program is very strong.

With respect to a4, I really have to quote the best email that ever reached my email box in 21th century. Short before world champs 2005:

Code: Select all

From: "Fabien Letouzey" 
To: "Vincent Diepeveen"
Subject: WCCC ambiance?
Date: Tue, 02 Aug 2005 09:21:23 +0100

Hi Vincent,

I have just been invited for the WCCC.  The reason is the most insulting 
possible: great VIP Remi Coulom could not come, so he proposed me instead.

I don't mind getting crushed by books and SMP search and fight hard to avoid 
the last place, but the single question for me is whether such events are 
nice?!

BTW I would really enjoy meeting you (not just see you like the two previous times I came to your country).

Thanks for your advice,

Fabien.

a5) local guys. This groups size is dependant upon nation. Big nations have a lot of engines that show up for sure. Knowing the huge number of dutch engines that exist, you can of course smell my disappointment so little of them showed up in 2007 world champs. My hope in Pamplona is of course that a lot of French entries show up besides Spanish entries. A world championship in your own nation is a big attractor to participants. See for example Italy 2006. None of the italian engines joined in 2000-2005 nor 2007 nor 2008 and very few i expect to show up in 2009, despite Italy to Spain being a very short distance. There was a similar effect in 1997 in France and 1999 Germany. Local guys sometimes also operate engines. In big nations there is more local guys who are willing to operate. The reason why i call this a special group is because this group ONLY joins when the event is closeby. Persons from this group who travel a long end to join a world champs somewhere i would put in group a2 (the computer chess addicts).

The local group always shows up of course and t
Past years, the groups a1 and a4 have become a lot smaller.

It is especially group a1 that has become smaller.

Now my hope is of course that Omid David Tabibi joins a1 as well as in the future the engine programmed by Costalba/Kannan together with the tomato truck manager from Spain. Also i hope of course Now again joins now that its author has retired.

Just write down a list of regular participants from the 80s and 90s that no longer show up past few years, just from head (and i'm missing probably a lot) :

- Tiger
- Rebel
- CS Tal
- Kallisto
- The King
- Fritz
- MchessPro
- Ferret
- Nightmare
- Nimzo/Hydra
- Lambchop
- Goliath/Little Goliath
- Deep Thought/Deep Blue
- Genius
- Now
- Centaur
- P.Conners
- Zugzwang
- Virtual Chess
- Darkthought
- Cilkchess

From the above non-university programs, with exception of Ferret, they all got sold under one or another name.

The big driving force behind computerchess was the prospect of a little money by means of sales. That still doesn't make you commercial, as it is peanuts money for a lot of effort. The ICGA definition of 'commercial' only applies to Shredder and Fritz IMHO and not to any other engine, as that's hardly sales what these guys have.

Now the discussion here on the internet is only about a2. Yet historically these hardly ever showed up anyway. A total wrong discussion. That Charles himself belongs to this group is basically telling us how fanatic he is in his hobby.

It is a group that is very exceptional though.

The proof is however very solid IMHO that it is group a1 which has become smaller and group a4.

Of course i hope to move Omid David Tabibi and Harm-Geert Muller into group a1 soon

p.s. if i look in ICCA history really disappointing is most dutch participants.

There were so many a few years ago.... ...just see even Paris 1997, which is 2 nations away from here. 3 hours by TGV (at 300KM/h)!

Vincent

[Graham Banks]

You can write down the names of these guys already on a list and probably most of them will show up, when having the opportunity, in Pamplona, in random order the programmer names: Zwanzger, Huber, Isenberg, Pijl, Bushinsky, Rajlich/Cozzie, Meyer-Kahlen, Kolls brothers, Pascutto, Uniacke, Diepeveen.

Why is Rajlich bracketed with Cozzie?

[Vasik Rajlich]

You will flag every parallel engine as a cheater then.

Also, what are you going to do without the same hardware?

This does not work at all.

The non-determinism isn't a show stopper, but IMO requiring executables is too draconian. Of course, executables will need to be required from first-time participants, as is already the case for existing online events.

I repeat my question: what are you going to do without the same hardware? (And I don't mean in terms of speed...)

Non-determinism will be a complete show-stopper in something like a Monte-Carlo engine, or a massively parallel system.

If you have an executable which was supposed to play the games, while the games were actually played with a much stronger (and therefore much different) program, it will be pretty obvious even with the non-determinism. You can try the game moves and see how different the scores are, how the scores change with deeper search, etc.

I have two problems here:

1) It's going to be really tough for a commercial author to give a tournament director an executable which is much stronger than his last public version. Those are usually kept under lock and key.

2) The rule won't handle a bunch of obscure cases, like programs which run on custom hardware or OS, etc.

Requiring log files seems more promising.

Vas

[Gian-Carlo Pascutto]

If you have an executable which was supposed to play the games, while the games were actually played with a much stronger (and therefore much different) program, it will be pretty obvious even with the non-determinism. You can try the game moves and see how different the scores are, how the scores change with deeper search, etc.

I agree, but it might be difficult in practice to show anything conclusively.

I have two problems here:

1) It's going to be really tough for a commercial author to give a tournament director an executable which is much stronger than his last public version. Those are usually kept under lock and key.

This one really surprises me! You played and still play in ICGA tournaments, which require handing over the sources to the tournament organization, when contested. Surely, giving out an executable is much less sensitive.

2) The rule won't handle a bunch of obscure cases, like programs which run on custom hardware or OS, etc.

That's exactly what I had in mind. It doesn't even need to be custom, just uncommon and hard to get.

Requiring log files seems more promising.

Yes. A live logfile would make cheating a lot harder. Provided it contains more than just the moves, of course