World Computer Chess Championship

[mjlef]

One programmer told me he felt the ICGA was not transparent enough. Towards making it more transparent, I am creating this public thread so anyone interesting in the World Computer Chess Championship, World Software Chess Championship and other events can comment and give feedback. Programmers interested in attending and entering their engines can let us know here. Suggestions for rules changes, people who have ideas for ICGA Journal articles, ways to improve events.... This can be an open forum. So, lets get started.

The International Computer Games Association (ICGA) website is here: icga.org

The dates for the World Computer Chess Championship and events are July 13th-19th. It will be held in Stockholm, Sweden. Details and more information will be placed on the website.

Now, the floor is open. I would be happy to answer questions that I can, and pass on the ones I cannot answer to the appropriate people.

Mark Lefler
ICGA Programmers Representative

[mcostalba]

Thank you Mark for this initiative, I am sure it will be appreciated by the community.

[jdart]

Since you asked:

I think already more than one person (although not me, so far) have contrasted what TCEC is doing with the ICGA competitions.

There are >1000 chess programs now, unlike some decades ago, and there are maybe 30-50 notably strong ones. TCEC was able to get a field that included certainly the top three, including particularly Houdini, and a good many of the others in the top level, including Rybka 4.1. So it is much more representative of the state of the art in software.

Furthermore it runs more rounds than WCCC, and so it can give some sounder indication of which engine is on top.

They are able to do that because it's online, there is no travel involved, and they were able to get voluntary participation from a large segment of the top field. (Not to state the obvious but also: they didn't ban anyone).

I understand there are differences and some are deliberate: WCCC is open hardware, which lets some participants join who have unusual hardware (Johnny for example). And it is not restricting the book.

Personally I like "own book" tournaments because I think the book is a part of the overall program strength, you get more variety in games, and you don't usually get the early opening blunders that otherwise occur. But quite a few engine authors don't bother with book construction, so they would be disadvantaged in that kind of competition.

I don't know what to do about open hardware but having a separate competition (preferably online) based on that could be interesting and if WCCC wanted to take on that role but with the goal of similar participation to what TCEC had, I'd think that a good thing. Or sponsor or endorse such a competition together with the TCEC team, perhaps. Would be very interesting if AlphaZero could participate.

--Jon

[CMCanavessi]

I was reading about the history of WCCC and checking their results and something caught my attention: from 1983 to 1999 there were always more than 20 engines competing, peaking in 1999 with 30.
Last year, there were 4.

The only thing that one can ask is what the hell happened?

[mjlef]

Thanks for the detailed response which I forwarded to David Levy. TCEC type tournaments are great. ICGA has begun transmitting the WCCC in the last couple of years. And a few years ago I proposed to David using TCEC or something similar as a platform to bring in more interest in the events.

You brought up three of the big issues: how do we handle unlimited hardware in a TCEC like system, opening books, and expenses. I agree opening preparation is a big part of an event like this. Book makers hold their own secrets, so somehow doing this on an open system add complexity.

Any suggestions on how to best do these?

ICGA can sometimes help with travel expenses, as they did this past year. it think some of the reasons to have in person tournaments has changed. One big, but rarely talked about issue is cheating. When the programs were not so strong, have the Tournament Director able to see all the operators to prevent "phoning a friend" was needed. But it is unclear who they could call that could play better than the programs (well, another program I suppose).

[jdart]

Playchess and InfinityChess already run unlimited hardware online tournaments and there is also precedent in the form of the CCT tournaments, etc. It requires participants be able to stay connected however.

It would be nice if there was some standard hardware available for those entries who do not have a reasonably powerful machine at their disposal. TCEC of course has that already, they just use it for all entrants.

I guess you cannot police a remote entrant but if you suspect for example it is using an engine other than what it advertises, it is probably possible to test that to some extent if full engine output is available.

--Jon

[mcostalba]

Thanks for the detailed response which I forwarded to David Levy.

This guy, David Levy, has some problem with the keyboard? Or perhaps he has some problem interacting with mere mortals?

Which of the two?

If I am the chief of an organization that want to restart a new channel with the community, I would put myself on the front line. Letting know the people who I am and what I think, without the support of (kind) middlemen.

If he is not able to communicate with people then do yourself a favor, change him.

P.S: In my day job I am in a position that allows me to write what I write, knowing what I am writing.

[Ovyron]

If he is not able to communicate with people then do yourself a favor, change him.

Fully agreed!

I think the WCCC could have the potential to become again the World Computer Chess Championship, not just in name, and this could be the very first step to achieve it.

[davidlevylondon]

Here I am, on the "front line."

I am responding to some of the posts which have appeared here during the past few days. In general I shall try to respond to such posts from time to time, but I cannot do this on anything like a daily basis so please be patient.

To Jon Dart: I would like to see the winners and other top programs from the TCEC and similar events participating in the ICGA events. Partly with that in mind we are providing travel and accommodation subsidies this year for the ICGA event in Stockholm, which will run from July 13th-19th inclusive. The detailed schedule will be published shortly on the ICGA web site and elsewhere. Ditto the rules. We are catering for up to 12 participating programs this year. We have the uniform platform tournament (WCSC) as well as the open tournament.

I agree with you that "own book" is the best approach. As you say, the book is part of the program, and so why not benefit programs that have had a lot of effort spent on their openings books?

To Carlos Canavessi - "What the hell happend?" A perfectly reasonable question. IMO it is the cost of attending which made the tournaments less appealing. In the early days chess programers could learn a lot from each other by attending, but later that aspect of the events did not seem to be so important, so fewer programmers were prepared to pay the travel and hotel costs. There are doubtless other reasons as well, but I believe cost to be the main one.

[Ovyron]

Thanks for participating David.

Now I want to inquire about specifially ICGA's ruling regarding the potential participation of the Houdini Chess engine in the WCCC. Houdini is clearly the top 2 chess engine as its latest version is ahead of Komodo, and both of them a buried behind dozens of Stockfish-derived engines that I'll count as a single entity.

Houdini's origin is shrouded in controversy as its author claims it's an original engine while other programmers claim it used Ippolit/Robbolito as its base. I indeed recall using them together and despite its elo Houdini was not a good engine for analysis just like Ippolit/Robbolito were only good in the elo chase (this has changed with version 5 or 6 dramatically improving on this aspect).

What would the ICGA do about it?

Houdini has no problems participating in TCEC, and with the second best engine absent from the WCCC we wouldn't really know if it would have managed to win the championship if it participated, so a clear advanage of TCEC is not only number of participants, but the quality of them, and if ICGA's rules are stopping top engines to participate (note the best engine Stockfish is also absent - 4 engines per Championship is fine as long as they're the best 4 in the world) I think it's time to revise them.