Some tweaks to the CCT rules

[bob]

I agree with changing the disconnect time to 10 minutes. We have gotten much more complicated in recent years than the old days.

I disagree with the kibitzing issue penalty. Yes, they must kibitz and not whisper. It is too easy to accidentally start up with kibitzing off. For me, ChessPartner handles it and I just have to click on a button and set the proper ICC/FICS variables. However, the cluster version of Telepath doesn't work with ChessPartner, it works with Xboard. In Xboard, I have to start it up with an extra command line option to send to Telepath. Very easy to forget that.

Fixing the forgotten kibitz in ChessPartner may not require a disconnect but it does in Xboard.

Doesn't for me, I just use the zippy2 password and tell crafty to kibitz and it is done. I then add it to the crafty.rc file and it is fixed for the rest of the games...

I have a single complaint, and it can easily be addressed. I do not buy all of this bullshit about "I can't kibitz on a ponder hit" or other such nonsense. Fix the program(s). Crafty kibitzes after _every_ move.

The reminds me to tell you should look at the logs between Gaviota and Crafty this weekend. There was one move that crafty kibitzed "n/a". A bug may have been exposed in that particular move.

That happens on _very_ rare occasions. It is a known race condition I am looking into. I had one other case where it kibitzed a move but played something else. I reran it several times and it always played the same move as it played in the game, but the log file never showed that output. I believe it happened when a time-out occurred at _exactly_ the correct node so that it didn't get displayed...

Miguel

Everyone else can do this too. Or else play somewhere else. This is not that hard to fix, but some simply refuse to do so. My philosophy is to either (a) follow _all_ the rules; (b) don't play. No middle ground because none is necessary. yet each online event produces more exceptions.

So, I agree with Miguel. When a person notices opponent not kibitzing, they inform the opponent and then the TD. The opponent must fix it immediately even if it means using one of the disconnects. If they have to use a second disconnect because they botched the job the first time then that is fine too.

The biggest problem we have with kibitzing is with the people that continually try to use Arena. It has never worked with kibitzing in any online tournament I've seen. The Arena guys claim it works, but I've never seen it work.

So, some early warning message may be in order to warn people that Arena is unlikely to work with kibitzing on FICS/ICC.

The second biggest problem is people complaining that it is not kibitzing when it is. The problem is when kib and allowkib variables are incorrectly set. This problem comes up multiple times per tournament. The only reason I call it the second biggest issue is because it is so easy to fix, however it appears more often than the first one.

The third sort of nonproblem with kibitzing is deep books. Some can not kibitz that it is a book move and their opponents think it is not kibitzing on move 27. This one is sort of easy to deal with; just check the clocks. There is the special case of falling out of and back into book.

Or why not just do the obvious and kibitz book moves? I do it. _anyone_ can do it. Or else just not play...

That issue came up 2 years ago at the ACCA Pan AM event in Thinker vs Crafty. People were complaining that Thinker wasn't kibitzing. Lance said it was still in book while Crafty had been out of book for almost 10 moves. Bob checked his log files and Thinker was indeed making immediate moves. When it finally came out of book, it started kibitzing.

[Dann Corbit]

As to the book makers:

Two teams will not be able to use a book created from the same book maker.

i.e. Ted Langreck makes Crafty's books. He would be prevented from making a book for another participant. Just as a participant would be prevented from using a book created by him.

If we only allow one engine per author, then we only allow one book per book maker. That is fair right?

This sounds utterly absurd to me.

Why? Should it be OK for me to work on several programs? We have five members working on Crafty. We could split it into 5 different programs that are almost identical, and each of us enter a different version and greatly increase our chances of winning. This is a team vs team competition using programs developed by each _team_. The WCCC adopted this rule several years ago and its a good one. One person per program, whether he codes or writes the book does not matter. Can't be on two different teams at the same time. Huge conflict of interest.

I doubt all the competitors are interested in developing a custom book. Should Mark's performance.bin be banned just because he's no longer around to give someone explicit permission? I think an exception should be made for free publicly available books. At the least I see no conflict of interest there.

Several engines use Ed's book.

There are other books that are used by multiple engines with permission.

I have written a dozen or so books (none of them great ones, of course) but something for beginning engine authors to use. I guess we will have to ban all of those engines from contests (or force them to run bare metal).
I don't recall ever putting in a back door or booby trap to hinder any particular engine {at least not intentionally } (though I admit anything is possible).

[bob]

As to the book makers:

Two teams will not be able to use a book created from the same book maker.

i.e. Ted Langreck makes Crafty's books. He would be prevented from making a book for another participant. Just as a participant would be prevented from using a book created by him.

If we only allow one engine per author, then we only allow one book per book maker. That is fair right?

This sounds utterly absurd to me.

Why? Should it be OK for me to work on several programs? We have five members working on Crafty. We could split it into 5 different programs that are almost identical, and each of us enter a different version and greatly increase our chances of winning. This is a team vs team competition using programs developed by each _team_. The WCCC adopted this rule several years ago and its a good one. One person per program, whether he codes or writes the book does not matter. Can't be on two different teams at the same time. Huge conflict of interest.

I doubt all the competitors are interested in developing a custom book. Should Mark's performance.bin be banned just because he's no longer around to give someone explicit permission? I think an exception should be made for free publicly available books. At the least I see no conflict of interest there.

If they can be used by _anybody_ then I would have no problem at all. I do have a problem with a book that can be used by more than 1, but not _everybody_ that wants to use it...

[JVMerlino]

Even with the engines, I think it requires case by case analysis. For instance, John Merlino worked on ChessMaster and has now written Myrddin. If ChessMaster enters a contest can Myrddin?

I don't think John was ever a programmer on the chessmaster engine. So I don't see a problem. I do see a problem with a programmer jumping around, because again, this poses a conflict of interest with potential misbehaviour.

True. Although I worked extensively on the Chessmaster GUI and features, and definitely had discussions with Johan about what the engine needed to implement to support the new features, I never even saw a line of code for The King, let alone helped write any.

jm

[hgm]

For me, its a command line start up option that I didn't encode into the xboard/uci protocol subsystem. I could do that.

For the uci engines, polylgot could be modified to kib "book move", but not all the uci based programs use polyglot. If they use Aquarium, ChessAssistant or ChessPartner they don't have that level of control, so its not an option for them.

What is there to kibitz on a book move or single legal move? Do we just want to print the message "book move" or "only move"? Especially the latter seems to add little value.

As to the technical aspects: I guess it depends a lot on which program is consulting the book. For UCI engines running under Polyglot, this could be the engine, the GUI, or Polyglot. For WinBoard engines it could be the GUI or the engine.

When XBoard / WinBoard are responsible for handling the book, the assciated kibitz can only be sent by XBoard itself, as the other entities do not have the information. But -autoKibitz already does that, and the opponent will see somehing like this (copied from the Engine-Output window, so read bottom to top!):

Code: Select all

!!! +0.24/8 (23.14 sec, 3593817 nodes, 155 knps) PV=g1f3 a7a5 e1g1 
\   d7d6 d2d3 c8f5 c1e3 h7h5 e3c5 d6c5
!!! +0.00/0 (0.00 sec, 0 nodes, 0 knps) PV=a2a3 (xbook)
!!! +0.00/0 (0.00 sec, 0 nodes, 0 knps) PV=f1g2 (xbook)
!!! +0.00/0 (0.00 sec, 0 nodes, 0 knps) PV=g2g3 (xbook)
!!! +0.00/0 (0.00 sec, 0 nodes, 0 knps) PV=b1c3 (xbook)
!!! +0.00/0 (0.00 sec, 0 nodes, 0 knps) PV=c2c4 (xbook)

Then there is the possibility to have Polyglot handle the book. I don't know how it handles kibitzing of book moves. But this is not a recommended mode in ICS play: any book that could be handled by Polyglot can also be handled by WinBoard / XBoard. Better to do it there, and let WinBoard handle both book and kibitzing. For the foolhardy: There are Polyglot versions around where kibitzing can be switched on and off, and the kibitz command (kibitz or whisper) can be set through the GUI interface.

Finally the engine can do its own kibitzing. WinBoard engines can send whatever they want through "tellics". I know little of UCI, but I think UCI engines can send something called PV info lines, which will be passed to the GUI by Polyglot as Thinking Output, with dummy (zero) values for the score, time and node count, with the same depth as the last PV that was reported by the engine. The text of the info line is then in the PV field of the thinking output. If this is the last line of Thinking Output WinBoard receives before the move, this line will be kibitzed to the ICS when -autoKibitz is on. And you could print anything you want there, e.g. "MOVE (engine book)".

So I don't think there is any technical problem. If we think just sending move + a message teling it came out of a book is not enough, just specify what more you would want to see on a book move, and I could put it into the Thinking Output that XBoard kibitzes on a GUI book hit.