On-line blitz tourney July

[Michel]

Oops,

I see that GnuCheese is not in the crosstable. Was it forfeited for some reason?

[hgm]

I guess you did not join it.

[Michel]

I was absent. It was my first (attempted) participipation. I interpreted the announcement as saying that it is sufficient to start up xboard with the right parameters to participate. I assumed that all engines present at the indicated time would be force joined by the td.

[Michel]

I am trying to find find the tourney announcement in my log files but
I can't find it.

How are tournaments announced on this ICM? It doesn't seem to be mamer
as I searched for that.

[marcelk]

I am trying to find find the tourney announcement in my log files but
I can't find it.

How are tournaments announced on this ICM? It doesn't seem to be mamer
as I searched for that.

Mind that this is a very old ICS code base. It is all not as automated as on other servers. Joining has to be done manually, there are no standardized announcements and I doubt if the admin can force any commands onto players. (Also tasks such as drawing by 50 moves rule have to be done manually.)

Basically if you can't be present and observe all games, you can't play. I noticed GnuCheese logged in but I didn't realize it wasn't playing until I saw this post. Next time better...

[hgm]

I was absent. It was my first (attempted) participipation. I interpreted the announcement as saying that it is sufficient to start up xboard with the right parameters to participate. I assumed that all engines present at the indicated time would be force joined by the td.

OK, sorry about the misunderstanding. I could have added GNUCheese to the tourney by remote, had I known that you intended me to do so. It is not standard procedure that I add any computer that is logged in. I tried that on several occasions in the past, and it did not work out well. (Operators were not there and the engine got stuck, or it was not complying with kibitz rules, etc. I am sure that GNUCheese would have been technically flawless, though.)

As I recall it, there was some sort announcement by mamer when I open the tourney. Could it be that you were not open to the tournament channel (49)?

I just tried with a dummy tourney, and it prints this:

Code: Select all

fics% 
mamer(49): *****Tourney Announcement*****                                                                    Trny #2  4 0 r  SWISS 0-9999 0 plr(s).  tell mamer join 2. Avg:   0.0
fics%

That 50-move and rep-draws claims are not recognized by the ICS is indeed a pain. I definitely am going to fix that, but I did not want to install a new version of the ICS just a few hours before the tourney.

[Michel]

As I recall it, there was some sort announcement by mamer when I open the tourney. Could it be that you were not open to the tournament channel (49)?

Yes that seems to be it. I should have thought of that. On FICS tournaments are announced by the td outside of any channels (if you set tourneyinfo to true that is). I had forgotten about the existence of a dedicated tournament channel.

That 50-move and rep-draws claims are not recognized by the ICS is indeed a pain.

Ok I did not realize this was a known bug.

[hgm]

Ok I did not realize this was a known bug.

It became a known bug because you reported it! (Thanks, btw.)

But in one of the games yesterday (Spartacus-Rookie) it surfaced again. This time neither of the engines was claiming, so the game just went on forever. (There still seems a bug, however, in the way the new WinBoard version displayed the 50-move counter, as it did not seem to be reset, although I saw Pawns and later Bishops disappear from the board.)

[hgm]

A quick comment. GnuCheese just lost the following game against Trace on time.

Code: Select all

[Event "ICS Rated blitz match"]
[Site "localhost"]
[Date "2011.07.08"]
[Round "-"]
[White "GnuCheese"]
[Black "Trace"]
[Result "0-1"]
[WhiteElo "1782"]
[BlackElo "1643"]
[TimeControl "120+12"]

1. e4 e6 2. d4 d5 3. Nd2 dxe4 4. Nxe4 Nd7 5. Nf3 Ngf6 6. Nxf6+ Nxf6 7. Bd3
b6 8. Qe2 Bb7 9. O-O Be7 10. Bf4 O-O 11. Be5 Nd7 12. Rad1 Nxe5 13. dxe5 Qc8
14. Be4 Bxe4 15. Qxe4 c5 16. h4 Qc7 17. Ng5 Bxg5 18. hxg5 Rad8 19. f4 c4
20. Kh2 c3 21. Rxd8 Rxd8 22. b3 Qd7 23. Kg3 b5 24. a3 a5 25. Rf3 Qd2 26.
Qc6 Qe1+ 27. Kg4 b4 28. axb4 axb4 29. Qb6 Qd1 30. Kg3 h5 31. gxh6 Rd2 32.
Qb8+ Kh7 33. Qf8 Qe1+ 34. Kh2 Qh4+ 35. Rh3 Qxf4+ 36. Rg3 Qh4+ 37. Rh3 Qf4+
38. Rg3 Qh4+ 39. Rh3 Qf4+
{GnuCheese forfeits on time} 0-1

However when checking the log file I noticed that this is a draw by threefold
repetition. GnuCheese correctly claimed the draw but the ICS
interpreted it as a draw offer and simply forwarded it to the opponent who
ignored it.
GnuCheese refused to make a further move and hence lost on time.


PS. I tried to compile the ICS but it did not work out of the box. I have
not tried again.

It turns out there is rep-draw code in this ICS, but that it is buggy: the problem here is that the first time this position was reached, it was reached by an irreversible move. The rep-check just starts checking positions one move too late.

[hgm]

This whole rep-draw code is a bit weird. It compared the position after the current move and the current move + 1 each to all positions since after the last irreversible move + 1 to the current move - 1, and if either of them matched two or more times, a draw offer was considered a claim.

Now the current move + 1 of course does not exist yet (the FEN is an empty string). One suspects that all move numbers are one off, and it should really have compared lastReversible to current-2 against current-1 and current. Current can of course never match current-1, because the other side is on move, so comparing upto current-2 makes sense. (In fact it cannot even match current-2 in games without null moves, so running upto current-4 would have been good enough, but a bit extra doesn't hurt.)

The question of course is why it also compares current-1 against all these positions. This mean you can still claim after a move that breaks the repetition, if you could claim before that move. I guess this is a work-around for the race condition, where you claim after your move, but the opponent could quickly interject a reply between your move and the claim. freechess.org now uses the command "draw MOVE" for this situation; perhaps I should implement that too, and forget about comparing the current-1 position.

I need to make some changes in the rep-draw code anyway, because in Xiangqi not every repetition is a draw.