TalkChess.com

Sven Schüle wrote:

bob wrote:

sje wrote:Last minute modifications during a tournament have led to some very humorous events over the years; I encourage other authors to do this as it's a sure source of entertainment.

OK, you have a point. between round modifications are good for lots of fun. Pawns on the 8th rank. Disappearing pieces. infinite loops. Subscript violations and resulting crashes that happen every time the game is restarted.

More than 20 years ago one of my very first chess programs played a small private tournament against programs of some friends. In one game my program displayed a "√" character on the square f5 ... Stefan Edlich might remember it. The "square root bug" was caused by not restoring the board correctly when taking back an ep capture, leaving the square in an incorrect state.

I had to do a "last minute fix" there, of course

And then there's a report I read about an event some three decades ago where a last minute change got the alpha-beta working perfectly -- perfectly badly, picking the worst possible move once out of book. Consider it the first working Losing Chess engine, invented by accident.

I am happy to provide any tourney organiser access to Sjeng's cluster at short notice and random during the event for any validation purposes.

Please read "Sjeng's cluster" in the above post as "my hardware".

sje wrote:

Sven Schüle wrote:

bob wrote:

sje wrote:Last minute modifications during a tournament have led to some very humorous events over the years; I encourage other authors to do this as it's a sure source of entertainment.

OK, you have a point. between round modifications are good for lots of fun. Pawns on the 8th rank. Disappearing pieces. infinite loops. Subscript violations and resulting crashes that happen every time the game is restarted.

More than 20 years ago one of my very first chess programs played a small private tournament against programs of some friends. In one game my program displayed a "√" character on the square f5 ... Stefan Edlich might remember it. The "square root bug" was caused by not restoring the board correctly when taking back an ep capture, leaving the square in an incorrect state.

I had to do a "last minute fix" there, of course

And then there's a report I read about an event some three decades ago where a last minute change got the alpha-beta working perfectly -- perfectly badly, picking the worst possible move once out of book. Consider it the first working Losing Chess engine, invented by accident.

I did this in 1977. We played in a human tournament, and fixed a couple of bugs the night before. In the first game, we played an incredible queen sac, and our opponent declined to accept it. When I saw the move, I simple told the system to reload the previous version and keep playing, which was perfectly legal back then by USCF rules. After the game, my opponent asked "If I had re-captured the queen, would I have been mated?" I told him "No, you would have easily won, it was a bug..."

For that one move, it literally was picking the worst possible move, not the best.

Suj wrote:I am happy to provide any tourney organiser access to Sjeng's cluster at short notice and random during the event for any validation purposes.

That is generally the way things are done. But in the case of Rybka, where we have to go back quite a ways to see which (if any) versions were legal, that becomes very problematic. It is hard enough to figure out about which version played in which tournament. It is likely that the hardware is no longer available.

We are trying to address the "which version played" by saving copies, but even that is problematic if someone wants to be dishonest...

sje wrote:And then there's a report I read about an event some three decades ago where a last minute change got the alpha-beta working perfectly -- perfectly badly, picking the worst possible move once out of book. Consider it the first working Losing Chess engine, invented by accident.

In the last ICGA Olympiad I drastically changed the piece values and piece-square tables of my engine after the first two rounds (4 games) of the 5x5 Shogi tournament. Because I had never seen that game played before, (I did not even know it existed before I came to the Olympiad; the engine I played with was designed for regular (9x9) Shogi, and I had just shrunk the board size the previous night in my hotel room), and those first four games had taught me there were some very important strategic principles my engine was totally unaware of.

After that, I continued the tourney, winning all remaining games, grabbing the silver medal.

So it can be done!

(Funny thing was that later I discovered there was a horrendous bug in my new assignment of piece values: I had given the Bishop the value that was intended for the promoted Bishop! )

hgm wrote:

sje wrote:And then there's a report I read about an event some three decades ago where a last minute change got the alpha-beta working perfectly -- perfectly badly, picking the worst possible move once out of book. Consider it the first working Losing Chess engine, invented by accident.

In the last ICGA Olympiad I drastically changed the piece values and piece-square tables of my engine after the first two rounds (4 games) of the 5x5 Shogi tournament. Because I had never seen that game played before, (I did not even know it existed before I came to the Olympiad; the engine I played with was designed for regular (9x9) Shogi, and I had just shrunk the board size the previous night in my hotel room), and those first four games had taught me there were some very important strategic principles my engine was totally unaware of.

After that, I continued the tourney, winning all remaining games, grabbing the silver medal.

So it can be done!

(Funny thing was that later I discovered there was a horrendous bug in my new assignment of piece values: I had given the Bishop the value that was intended for the promoted Bishop! )

A little luck never hurts. I won the 1983 WCCC in New York, using a parallel version (the first parallel version of Cray Blitz) that had never even played a complete game. We were helping debug the Cray parallel library (similar to pthread library today) the day before the first round in fact. And we played 5 flawless games, no hangs, no odd things, completely amazing...

I had forgotten that on the way to NYC (we drove from Hattiesburg, MS, a 24 hour non-stop drive) I stopped at several points along the way to call Harry Nelson, who was up at Cray Research. We debugged as I drove, using pay phones every hour or two.

CRoberson wrote:The 2011 Fifth Annual ACCA World Computer Rapid Chess Championships is tentatively scheduled for the weekend of July 23/24 2011. It is tentative only to give others a chance to tell me of a serious timing conflict. After looking around the usual tournaments, I couldn't find a conflict.

The time control and rules are mostly as they have been. The only change may be that all participants provide a binary of the version entering. This is for tracking purposes and ease of catching clones. I haven't completely decided on that rule yet. It is open for discussion.

I will have the web pages up this week. You can respond here or PM me if you have any questions, comments or requests.

Hi Charles,

Rule: "8.Participants are allowed 2 disconnects due to accidents"

(The underlining is mine)
May we assume that disconnecting on purpose for changing books/engine version etc is not allowed during a game?

BTW: Good luck!

This is a concept at least 50-60 years old and is called change control. Never implement anything that hasn't been thoroughly tested and aged unless it's an absolute emergency. All it takes is one misstep to realize the validity of the rule.

Dave

Andres Valverde wrote:

CRoberson wrote:The 2011 Fifth Annual ACCA World Computer Rapid Chess Championships is tentatively scheduled for the weekend of July 23/24 2011. It is tentative only to give others a chance to tell me of a serious timing conflict. After looking around the usual tournaments, I couldn't find a conflict.

The time control and rules are mostly as they have been. The only change may be that all participants provide a binary of the version entering. This is for tracking purposes and ease of catching clones. I haven't completely decided on that rule yet. It is open for discussion.

I will have the web pages up this week. You can respond here or PM me if you have any questions, comments or requests.

Hi Charles,

Rule: "8.Participants are allowed 2 disconnects due to accidents"

(The underlining is mine)
May we assume that disconnecting on purpose for changing books/engine version etc is not allowed during a game?

BTW: Good luck!

I think normally the human should not interfere while the engine is running and playing. Accidents - Internet connection problems, power interruptions, harware failures, and maybe problems related to software interface used in connecting to ICS.

Personally I don't really care if my opponent disconnects for whatever reason, as long as I don't have to do the rematching of engines and don't have to count how many times my opponent disconnects - it should be the responsibility of my opponent to report to tournament administrator (TA) that he disconnected 3 times already . It is also better and ethical that everytime the engine disconnects, the operator when online, should inform the TA that he had disconnected by a number of times .

Andres Valverde wrote:

CRoberson wrote:The 2011 Fifth Annual ACCA World Computer Rapid Chess Championships is tentatively scheduled for the weekend of July 23/24 2011. It is tentative only to give others a chance to tell me of a serious timing conflict. After looking around the usual tournaments, I couldn't find a conflict.

The time control and rules are mostly as they have been. The only change may be that all participants provide a binary of the version entering. This is for tracking purposes and ease of catching clones. I haven't completely decided on that rule yet. It is open for discussion.

I will have the web pages up this week. You can respond here or PM me if you have any questions, comments or requests.

Hi Charles,

Rule: "8.Participants are allowed 2 disconnects due to accidents"

(The underlining is mine)
May we assume that disconnecting on purpose for changing books/engine version etc is not allowed during a game?

BTW: Good luck!

I guess it is if the operator accidently forgot to set the book....