Discussion about development of draughts in the time of computer and Internet.
-
Ed Gilbert
- Posts: 859
- Joined: Sat Apr 28, 2007 14:53
- Real name: Ed Gilbert
- Location: Morristown, NJ USA
-
Contact:
Post
by Ed Gilbert » Wed Jul 23, 2014 13:29
I think that the experience of chess programs clearly shows that ultra fast games are very good predictors for longer time controls.
A fundamental difference is that the chess matches are controlled by a tournament manager that measures and keeps track of the time used by all the engines. With DXP, each engine keeps track of its own time used. This is unacceptable for blitz games. For example, if an engine did not include some of its non-search activities in its accounting of time used, that would give it an unfair advantage, and it would be difficult to detect.
-- Ed
-
Rein Halbersma
- Posts: 1722
- Joined: Wed Apr 14, 2004 16:04
-
Contact:
Post
by Rein Halbersma » Wed Jul 23, 2014 13:49
Ed Gilbert wrote: I think that the experience of chess programs clearly shows that ultra fast games are very good predictors for longer time controls.
A fundamental difference is that the chess matches are controlled by a tournament manager that measures and keeps track of the time used by all the engines. With DXP, each engine keeps track of its own time used. This is unacceptable for blitz games. For example, if an engine did not include some of its non-search activities in its accounting of time used, that would give it an unfair advantage, and it would be difficult to detect.
-- Ed
I am not really concerned with fairness. For automated draughts tournaments or public rating lists, yes, one would need a tournament manager (like cutechess-cli or Winboard) to ensure a fair outcome. But for private testing, all I care for is the ability to run ultra fast games. If one engine uses a slightly different definition of time keeping, I can work around that using asymmetric time controls (e.g. give myself a little more time, or the other program a little less).
-
Ed Gilbert
- Posts: 859
- Joined: Sat Apr 28, 2007 14:53
- Real name: Ed Gilbert
- Location: Morristown, NJ USA
-
Contact:
Post
by Ed Gilbert » Wed Jul 23, 2014 16:43
If we are going to change the GAMEREQ format
I don't see sending for example "0.5" as being a change in the GAMEREQ format. Frank's description of this field says, "the time in minutes for each player for the whole game; must be specified in 3 positions;". Nothing in his description prohibits sending a decimal point along with ASCII digits. This allows you to send game times a short as .01 minutes (0.6 seconds). This seems like plenty of resolution.
I am not really concerned with fairness. For automated draughts tournaments or public rating lists, yes, one would need a tournament manager (like cutechess-cli or Winboard) to ensure a fair outcome. But for private testing, all I care for is the ability to run ultra fast games.
That's fine, but if this capability is added to engines, then it will be used to run tournaments (after all, it was requested by Krzysztof), and the results will be published on public forums.
-- Ed
-
Rein Halbersma
- Posts: 1722
- Joined: Wed Apr 14, 2004 16:04
-
Contact:
Post
by Rein Halbersma » Wed Jul 23, 2014 19:39
Ed Gilbert wrote:If we are going to change the GAMEREQ format
I don't see sending for example "0.5" as being a change in the GAMEREQ format. Frank's description of this field says, "the time in minutes for each player for the whole game; must be specified in 3 positions;". Nothing in his description prohibits sending a decimal point along with ASCII digits. This allows you to send game times a short as .01 minutes (0.6 seconds). This seems like plenty of resolution.
Ok, .01through 9999 minutes is plenty, tnx for pointing that out !
-
Krzysztof Grzelak
- Posts: 1368
- Joined: Thu Jun 20, 2013 17:16
- Real name: Krzysztof Grzelak
Post
by Krzysztof Grzelak » Thu Jul 24, 2014 10:50
Ed Gilbert wrote:
That's fine, but if this capability is added to engines, then it will be used to run tournaments (after all, it was requested by Krzysztof), and the results will be published on public forums.
Ed it is necessary publicly to publish not everything on the forum. I asked whether there was such possibility so that so quickly he plays programs. I am not a programmer and I don't know a lot about it but willingly at home I would conduct such tests. As for the publication to this forum honestly I would ask owners of programs whether I can publicly present conclusions of these tests. Rein is also testing chess programs in particular Stockfish, Komodo, Houdini. I would like to add that I am dreaming of such a program like in chess (not necessarily for very short times) but up to draughts programs. It the program in which it was can was add a few checkered programmes, adjust the hash as well as the base of endings of given programme.
Krzysztof
-
Krzysztof Grzelak
- Posts: 1368
- Joined: Thu Jun 20, 2013 17:16
- Real name: Krzysztof Grzelak
Post
by Krzysztof Grzelak » Thu Jul 24, 2014 18:03
I apologise that I am asking Ed, Bert, Rein and Michel. Whether there is a possibility that such a program will arise.
Krzysztof
-
MichelG
- Posts: 244
- Joined: Sun Dec 28, 2003 20:24
-
Contact:
Post
by MichelG » Tue Aug 05, 2014 13:18
In conclusion:
- the current damexchange format already allows for faster games by setting time to non-integer value. Only Dragon and Kingsrow appear to support this, but they limit the shortest time to 0.9 and 1.0 minutes a game while playing against external engines.
- Rein has suggested a damexchange format that allows a greater range of game-times and fisher clocks. If people have a desire to use fisher clocks in damexchange, we should further discuss if there is enough enthusiasm for that. (i don't really have a great desire for it)
For me, i do want to have some minimum time per game in dragon. In the next version i will allow games at 0.5 minutes per game. I invite others to support that as well.
However, I don't really want other programs to 'mine' or finetune against dragon by playing very large numbers of games. There will be a measure to prevent that.
Michel
-
Krzysztof Grzelak
- Posts: 1368
- Joined: Thu Jun 20, 2013 17:16
- Real name: Krzysztof Grzelak
Post
by Krzysztof Grzelak » Tue Aug 05, 2014 14:59
He thanks warmly for the statement Michel. That's true that I wanted to play a bit faster between programs - 30 seconds to the party. It is necessary to make not everything available publicly on the forum, not which things willingly I am keeping to myself what is regarding draughts programs.I would like so that somebody at one time creates such a program in which a few programs can participate for different times without the interference of man. I will give an example. I would like to conduct championships of 3 rounds from five with programs for the time on exemplary 10 minutes without the interference of man. Whether somebody I will write such program similar to of chess programs, such a manager of games. What you think about this idea Ed, Bert, Rein and You Michel.
-
Rein Halbersma
- Posts: 1722
- Joined: Wed Apr 14, 2004 16:04
-
Contact:
Post
by Rein Halbersma » Wed Aug 06, 2014 13:25
MichelG wrote:
However, I don't really want other programs to 'mine' or finetune against dragon by playing very large numbers of games. There will be a measure to prevent that.
If Dragon has some legitimate trouble with super-fast time controls, it would be no use "mining" or "exploiting" that just to claim really good match results (for advertisement purposes). The whole point of super-fast time controls is to get faster high-quality statistical feedback on engine improvements. For that to be useful, the opponent Engine has to play well at the time control, otherwise you cannot extrapolate the match results to longer time control.
In chess, nobody really does that because it is accepted there that there are independent people running rating-list tournaments at regular blitz time intervals that all engines support. The super-fast time controls are only used in private testing. If Dragon or others wouldn't support it, then that would limit the super-fast engine to self-testing. This is no problem of course.
-
Krzysztof Grzelak
- Posts: 1368
- Joined: Thu Jun 20, 2013 17:16
- Real name: Krzysztof Grzelak
Post
by Krzysztof Grzelak » Wed Jan 28, 2015 17:44
Rein Halbersma wrote:I am not really concerned with fairness. For automated draughts tournaments or public rating lists, yes, one would need a tournament manager (like cutechess-cli or Winboard) to ensure a fair outcome. But for private testing, all I care for is the ability to run ultra fast games. If one engine uses a slightly different definition of time keeping, I can work around that using asymmetric time controls (e.g. give myself a little more time, or the other program a little less).
Rein I have a question for you. Please write or using program Winboard, you can play you can play tournament between programs checkers. Using ' tournament manager '.
-
Rein Halbersma
- Posts: 1722
- Joined: Wed Apr 14, 2004 16:04
-
Contact:
Post
by Rein Halbersma » Wed Jan 28, 2015 19:30
Krzysztof Grzelak wrote:Rein Halbersma wrote:I am not really concerned with fairness. For automated draughts tournaments or public rating lists, yes, one would need a tournament manager (like cutechess-cli or Winboard) to ensure a fair outcome. But for private testing, all I care for is the ability to run ultra fast games. If one engine uses a slightly different definition of time keeping, I can work around that using asymmetric time controls (e.g. give myself a little more time, or the other program a little less).
Rein I have a question for you. Please write or using program Winboard, you can play you can play tournament between programs checkers. Using ' tournament manager '.
You should write H.G. Muller an email. I haven't had the time to adapt my engine to the Winboard protocol.