Suggestions for new CCT rules

[CRoberson]

I understand that viewpoints can differ, and I believe there is certainly a place for all types of competitions - but to me, the ideal of computer chess is that the logic of the program controls and guides the game as much as possible, and a major part of the art of programming a great engine is giving it the positional and strategic understanding of the game needed to play well in the opening without a book.

I believe the analogy between humans and computers as to the role of the opening book is also false; a human always retains their free will as to which book line to follow and when and where to leave their book preparation. In the case of standard UCI + GUI, the engine algorithms are not even running during the "in book" section. The continuing evolution of computer chess going forward seems to need a split between the development of analytical theory and deep analysis from the opening through the midgame, and the development of programs with the ability to synthesize "deep knowledge" of the game on the fly with positional evaluation and no book.

This is my first post to this forum and my opinion is approximately worthless, just sharing the perspective of a long-time chess and computer hobbyist.

The fact that the GUI doesn't consult the engine during book moves doesn't guarantee that the engine isn't part of the process. For those of us that use a UCI GUI that supports automated learning, the engine and the GUI work together over time to adjust the book. Those adjustments happen from move 1 throughout the book for me. So, there is cooperation and coordination going on, but over time as opposed to this instant.

Also, some use position learning which is not exactly the same thing.

The ICGA, ACCA and CCT events are not chess events. They are primarly programming/research competitions. Nobody had a problem with that 30 years ago, because the chess wasn't GM level. These limitless types of competitions allow all the researchers/scientists to push the boundaries in all directions. That is part of the reason clones aren't allowed - they don't invent anything new. Spectators only started complaining and treating it as a chess competition when we approached GM level play, but it isn't really a chess competition. It just looks like one.

Wow, that is the ultimate pass of the Turing test! We do it so well that people have forgotten that it is not a chess tournament.

[CRoberson]

Also Nelson Hernandez makes a good point on the Rybka forum about book limiting, some openings need more moves in books than others. Do you know how the CB "Import to relative ECO length" works? I think something like that is what is needed.

Nelson made another smart observation on the Rybka forum - that the Hiarcs team outfoxed the Sjeng team, not allowing them to use the h/w advantage in that game.

That is actually a superior point; not just a good one. Nelson may not even know how good a point that is.

It comes down to letting us come up with creative ways as to "outfoxing" other strategies and tactics. You have bigger hardware so I have a good book. Hah, I trump your good book with my big hardware and a book that will take us out of opening quickly. Hah, I trump that with .......

[Christopher Conkie]

For years, there has been a WCCC of all machines must have unique books. Same rule exists for ACCA tournaments.

You should limit all books to 20 ply. Having games that get to move 50 and engines are still in book, is just dense. This would avoid all the rubbish about who is using what book.

If one were an outsider looking in, you could be mistaken for thinking that these people had lost the plot. I have no idea about you Charles but when I watch computer chess I want to see how the engine plays, not something we have seen before churned out by some book.

That is my suggestion.

20 ply, no more.

[CRoberson]

how about switching from kibitz to whispers?

a) all the same info is there
b) the tourney is now held on a server where accounts are free so there
is no problem to view the output
c) almost all observers including authors AND operators arent viewing from the remote machine they are operating anyway
d) it would completely eliminate fears that the opponent has found a method to take advantage of the kibitzed info

Whispering doesn't work. The point of kibitzing is so that your opponent can and does see your thinking process. Opponent being the human. This is a way of ensuring that people aren't cheating.

[CRoberson]

For years, there has been a WCCC of all machines must have unique books. Same rule exists for ACCA tournaments.

You should limit all books to 20 ply. Having games that get to move 50 and engines are still in book, is just dense. This would avoid all the rubbish about who is using what book.

If one were an outsider looking in, you could be mistaken for thinking that these people had lost the plot. I have no idea about you Charles but when I watch computer chess I want to see how the engine plays, not something we have seen before churned out by some book.

That is my suggestion.

20 ply, no more.

Generally speaking, the vast majority of Telepath's games are out of book by move 20. Many are out of book by move 10. Rarely, I have seen a few go to around 30. I don't like those myself, because they get to drawn endgames out of book. When those happen, I cut them out of the book.

[garybelton]

After reading this in the hiarcs public forum perhaps there should be a rule 6

6. The TD will enforce his own rules at all times

This would certainly take away any possibility of unsportsmanlike conduct from the teams.

You miss the point...

It didn't appear anyone was using a stopwatch from the time of the disconnect until the game initiated again.

The reconnection was quick enough that it was not apparent if it had been over 5 minutes or not. Komodo operator said nothing, TD said nothing and soon enough, the game was resumed.

It wasn't until some 40+ minutes later after Komodo's position was CLEARLY lost that the operator then brought up the complaint that Junior had not reconnected within the 5-minute timeframe.

The "sportsmanship" that Harvey refers to is the fact that the complaint surfaced once Komodo's luck had changed (read: "find a way out of the loss...)

If an operator or author needs to lodge a complaint or make a challenge, it needs to happen when the event or situation occurs, not at a much later time. Don't agree to play on and then backpedal when you don't get the luck you had hoped for...

But then again, with these latest genre of engines popping up in the wake of the published Rybka code... Honor amongst thieves is an oxymoron...

-elc.

[Zatarra]

how about switching from kibitz to whispers?

a) all the same info is there
b) the tourney is now held on a server where accounts are free so there
is no problem to view the output
c) almost all observers including authors AND operators arent viewing from the remote machine they are operating anyway
d) it would completely eliminate fears that the opponent has found a method to take advantage of the kibitzed info

Whispering doesn't work. The point of kibitzing is so that your opponent can and does see your thinking process. Opponent being the human. This is a way of ensuring that people aren't cheating.

um... read item B. opponents readiing are human, and thus can make an observing account. there is actually more potential for cheating via kibs than whispers