Computer draughts tournament announcement

TAILLE · Post by **TAILLE** » Wed Jul 15, 2009 23:07

BertTuyt wrote:To be honest, i really don't like to do adjustments during the game.
I know (rumours) from the chess world, that some programmers also do on-line analysis of the game , and in this way add interesting positions to the hash tables.

Also i very much dislike, moving the mouse over the board during analysis, and "doing some imaginary move sequences".
Im sure no-one within the Draughts world has cheating modes built in their programs.

But to avoid any discussion, lets totally refrain from any major, non-standard, operator involvement.

Bert

I am also sure that within the Draughts world, no cheating modes are built in a program.

My general view is the following : as programmers we all know that it is very very easy to give a program extra information for example by the way of entering the last opponent move (it exists almost an infinite way to click on a given square). That means that you can either suggest a move or you can suggest taking more time on the next move or whatever you want.

Fortunetly I do not believe that suggesting a move has any interest because I guess all programs are far better than any programmer. Maybe it remains only the possibility to give the program extra information concerning the time to use on the next move.

A solution exist : anybody explain to the opponent programmer the operations needed by his program and each programmer will operate the opponent program !

Be serious, as far as I am concerned I decide to trust my opponent!

BTW be not surprise to see Damy answering very quickly in some occasions : if Damy guess correctly the opponent answer during the opponent time then it avoids spending to much time when it is Damy turn.

TAILLE · Post by **TAILLE** » Wed Jul 15, 2009 23:28

Ed Gilbert wrote:This is a difficult situation. As you noted, it is not always possible to rewind the match back a few moves and then continue, because your program may not be designed to resume in this way. As you noted, you may not know how much time you had used for those moves that were undone. In my case, the only way I can change the clock is by resetting the count of moves played to 0. I could of course reset the clock to the remaining time left, and set the game moves to the number of remaining moves, but my search time algorithm varies during the game and takes into consideration that after a certain number of moves I am primarily getting results from an endgame database and thus do not need nearly as much time, so it would wrongly think it is at the start of a new game and not manage time correctly for the situation. While I agree with you that we should try not to penalize operator error, we also shouldn't allow an operator error to make the situation worse for his opponent. I think in these situations that if your opponent made errors causing his program's board position to differ from the table board, then you should have the option to request that he play from the position on the table, and only if you are willing to allow a rewind then you can offer this as an option. So in this case if you think you can rewind and resume the match without too much disruption to your game, you can offer this, but if not you can insist he play from the table board position. Does this sound reasonable?
-- Ed

Yes Ed, of course I agree with you.
Curiously the Leo Nagel problem (Cerberus position different from board position) was really detected by Damy ! Cerberus made surely a good move according to its position but it was a blunder according to the board position. When I saw Damy evaluation following this blunder, I told Leo Nagel that Cerberus made a losing move and this allowed Leo Nagel to detect the problem.
Then I had two choices :
1) I do not accept to rewind the game and I win this game immediatly (it would have been very easy for me to explain to Jaap Bus that I cannot rewind the game without penalizing Damy)
2) I accept to rewind the game and I trust Leo Nagel in rewinding the correct number of moves.
My first goal when I went to Culemborg was to play games against other programs. To win the tournament was only a secondary goal. Without any hesitation I did my best on the second solution.

TAILLE · Post by **TAILLE** » Wed Jul 15, 2009 23:48

Ed Gilbert wrote:my search time algorithm varies during the game and takes into consideration that after a certain number of moves I am primarily getting results from an endgame database and thus do not need nearly as much time, so it would wrongly think it is at the start of a new game and not manage time correctly for the situation.

BTW Ed I faced exactly this problem when Damy program crashed against Tornado in the last french open tournament in Arleux.
After having lost a lot of time due to two successive crashes (in fact due to the same problem) I managed to continue the game from the board position but here I had a third problem : Damy considered that it was the start of a new game and it plays the following 15 moves very quickly (less than 1 second per move) considering that it played from the initial position of the game, in order to keep the maximum of time for the following 30 moves!
Now I solved this last problem.

Ed Gilbert · Post by **Ed Gilbert** » Thu Jul 16, 2009 00:03

Gerard, thank you for explaining your time management more clearly. I understand now how you can modulate the t3 time to keep your damy clock in agreement with the table clock. I guess the main disadvantage to your scheme is the extra mouse click, which is a bit unintuitive. I could see myself forgetting to do the final click and then be in big trouble!

-- Ed

Ed Gilbert · Post by **Ed Gilbert** » Thu Jul 16, 2009 00:20

Fortunetly I do not believe that suggesting a move has any interest because I guess all programs are far better than any programmer.

Indeed, I would have to be crazy to offer move advice to kingsrow!

I recently created a "hello world" draughts program as a demo to exercise a conversion of Martin Fierz's CheckerBoard GUI to 10x10 draughts. This hello world program has no hashtable, no move ordering of any kind, and its eval function is simply material count and center control, nothing more. Against kingsrow it loses every game. Against me it wins every game! :-(

-- Ed

TAILLE · Post by **TAILLE** » Thu Jul 16, 2009 00:22

Hi Ed,

Ed Gilbert wrote:Gerard, thank you for explaining your time management more clearly. I understand now how you can modulate the t3 time to keep your damy clock in agreement with the table clock. I guess the main disadvantage to your scheme is the extra mouse click, which is a bit unintuitive. I could see myself forgetting to do the final click and then be in big trouble!

-- Ed

Yes I see your point. In that case you can also decide in your program that t3 can not exceed e.g. 3s.

wellnesswrotter · Post by **wellnesswrotter** » Thu Jul 16, 2009 09:31

Almost Jonah 2008
With the only difference, that i am a stronger player than my program.

To resolve cheating in GameMoveInput you should
A. operate you're opponents computer ( )
B. use standaard input GUI ( )

Ed Gilbert wrote:
Fortunetly I do not believe that suggesting a move has any interest because I guess all programs are far better than any programmer.
Indeed, I would have to be crazy to offer move advice to kingsrow!

I recently created a "hello world" draughts program as a demo to exercise a conversion of Martin Fierz's CheckerBoard GUI to 10x10 draughts. This hello world program has no hashtable, no move ordering of any kind, and its eval function is simply material count and center control, nothing more. Against kingsrow it loses every game. Against me it wins every game!

-- Ed

BertTuyt · Post by **BertTuyt** » Thu Jul 16, 2009 20:41

Just a question to all programmers.
Damage like other programs has the DamExchange implemented.
Both in server as client mode.
So far we never "played" this way during tournament.
However what is the general opinion.
Should we allow this mode in the future ??

Bert

wellnesswrotter · Post by **wellnesswrotter** » Thu Jul 16, 2009 21:14

In the future, this certainly is going (must) happen.

BUT
- In the beginning, have a normal board ready if dameXchange fails
- It takes away some charm
- You need 'network' infrastructure
- You are not bound to the same location

TAILLE · Post by **TAILLE** » Thu Jul 16, 2009 22:18

For the time being I did not introduce DamExchange in Damy program. BTW, by reading quickly the protocol description made by Frank Mesander I understood that a move is described by the source square, the destination square and the list of squares concerned by a capture. That looks fine but my move generator do not give me the source and destination square when they are identical. My question is the following : in this case (source = destination) is it allowed to code these fields 00 00 or any other value ?

Ed Gilbert · Post by **Ed Gilbert** » Fri Jul 17, 2009 13:16

That looks fine but my move generator do not give me the source and destination square when they are identical. My question is the following : in this case (source = destination) is it allowed to code these fields 00 00 or any other value ?

No, that will not work. You have to put the correct source and destination squares in these fields. A simple solution is to add one more field in your movelist structure which only needs to be assigned when source and destination squares are the same. Since this can only happen when at least 4 pieces are captured, it should add negligible overhead to the move generator.

-- Ed

Ed Gilbert · Post by **Ed Gilbert** » Fri Jul 17, 2009 13:43

Just a question to all programmers.
Damage like other programs has the DamExchange implemented.
Both in server as client mode.
So far we never "played" this way during tournament.
However what is the general opinion.
Should we allow this mode in the future ??

I guess we can allow it, but I doubt we will be able to *require* it any time soon, which would be necessary to get any real benefits from it, like more games played, and perhaps at faster time controls because no operator time. If you look at the chess world, they have hundreds of programs, and most of them have the ability to play automated games through a standard internet chess server interface because there are free GUIs available that have this capability built-in. But AFAIK when they have the kinds of tournaments where they all gather at some physical place (as opposed to virtually on the internet) they use real boards and chess clocks between the computers.

-- Ed

Ed Gilbert · Post by **Ed Gilbert** » Fri Jul 17, 2009 14:05

Bert,

Slightly off topic, but did you fix the problem with your parallel search? At least I assumed it was a problem because you used a single search thread at Arleux.

When I ran those series of engine matches against Truus, I remember that Feike pointed out how odd it was that there was such a small difference between the matches using 3 threads vs. only 1. At the time I attributed it to the fact that all the matches were very lopsided so maybe not much room for further improvement against truus, but I wanted to do something else to try to check that the parallel search is making a real improvement. I had to postpone those experiments while the 8pc db was building, but now that the computers are idle I can do those things. I ran a match this week putting kingsrow/3-threads vs. kingsrow/1-thread. The results were +9 -1 =148. So I think it is working ok.

-- Ed

Ed Gilbert · Post by **Ed Gilbert** » Fri Jul 17, 2009 14:11

Almost Jonah 2008
With the only difference, that i am a stronger player than my program.

Johan, you are an anomaly!

wellnesswrotter · Post by **wellnesswrotter** » Fri Jul 17, 2009 14:42

Johan, you are an anomaly!

Not really, Jonah is (still) weak and i am (again) in the top 100 of Dutch players[/quote]