A couple of surprising results, but they're very close so that can happen. BTW, what books are Amy and Yace using, or what kind of pgn-file did you use?The first round in my winboard tournament has come to an end. The results somewhat differ with my expectation:
1)Amy 0.7 - Yace 0.20 (0-1)
2)AnMon 5.09 - SOS 991103 (0-1)
3)Comet B23 - The Crazy Bishop 0045 (0-1)
4)Crafty 17.11 - Gromit 3.0 (0-1)
[Event "KUT_O2 P3/650 32 HT PB=off 60/60'"]
Reply to Mogens:A couple of surprising results, but they're very close so that can happen. BTW, what books are Amy and Yace using, or what kind of pgn-file did you use?The first round in my winboard tournament has come to an end. The results somewhat differ with my expectation:
1)Amy 0.7 - Yace 0.20 (0-1)
2)AnMon 5.09 - SOS 991103 (0-1)
3)Comet B23 - The Crazy Bishop 0045 (0-1)
4)Crafty 17.11 - Gromit 3.0 (0-1)
[Event "KUT_O2 P3/650 32 HT PB=off 60/60'"]
Best wishes...
Mogens
PS: Thank you for signing my Guestbook.
Reply to Mogens:A couple of surprising results, but they're very close so that can happen.The first round in my winboard tournament has come to an end. The results somewhat differ with my expectation:
1)Amy 0.7 - Yace 0.20 (0-1)
2)AnMon 5.09 - SOS 991103 (0-1)
3)Comet B23 - The Crazy Bishop 0045 (0-1)
4)Crafty 17.11 - Gromit 3.0 (0-1)
[Event "KUT_O2 P3/650 32 HT PB=off 60/60'"]
BTW, what books are Amy and Yace using, or what kind of pgn-file did you use?
Mogens
I'm using the "normal" books for Amy and Yace. You have created special books >I think but I like letting play programs with their original "outfits".
There's another similar solution, which Dieter is working on at the moment. I don't think I should be the one to reveal some of the details, so I'll leave the explaining to Dieter when he believe it works.Also since Yace and Amy don't have any book learning, I suppose, someone who had enough games could help make the "normal" opening book better..By cutting out lines that Yace or Amy don't play well..
Besides a auto book learning feature, I'm not sure what other solution there is..There's another similar solution, which Dieter is working on at the moment.Also since Yace and Amy don't have any book learning, I suppose, someone who >>had enough games could help make the "normal" opening book better..By cutting >>out lines that Yace or Amy don't play well..
A book made using 2600.pgn isn't good, it's too narrow as you say. The book used with my experimental tournament is based on that and I foresee problems for Amy and Yace. Amy lost 4-0 to Zchess and Yace is about to lose the second game in a row against Bringer and then the score would be 0-2. Incidentally, Bringer uses the same book with the same depth.I'm also beginning to wonder if the book made out of p2600.pgn is a good book. Since it is lacking in quite a few opening lines..Probably because GMs just don't play certain lines..
Perhaps I should make a book with the same file that you can use to make crafty's book..
Well you didn't state that it was due to opening or anything..A book made using 2600.pgn isn't good, it's too narrow as you say. The book >used with my experimental tournament is based on that and I foresee problems >for Amy and Yace.
Amy lost 4-0 to Zchess
and Yace is about to lose the second game in a row against Bringer and then >the score would be 0-2. Incidentally, Bringer uses the same book with the same >depth.Please do, and post your experiences.Perhaps I should make a book with the same file that you can use to make >crafty's book..
I'm currently trying to expand my pgnbase carefully. I'm approaching 60000 >games, but there's a lot of doubles I think.
It was the same openings, but not complete doubles (Amy changed its mind at move 15 and 22 I think). Zchess har learning and a slightly deeper book (you can't determine depth) than Amy (used flatten 1).Well you didn't state that it was due to opening or anything..Amy lost 4-0 to Zchess
Or are the games doubles?
But if I understand you correctly, all engines use the same book..So, this probably means that the book suits Zchess better? Anyway the result is only a little surprising..
Again the result is a minor surprise..Just to clarify, Are you blaming Yace and Amy's poor result on the narrow book?
Finally, what is the point of your tournament, by making all the engines play using the same book?
If we get "upsets" can we say that Bringer and Zchess are better suited to use the 2600.pgn books?
Or if the results are almost as expected, you can see that the book is "netural"
?
If the engine is out of book even after very common opening moves (and goes on to play "weak" moves), it means that you need to collect more games for that certain variation..
The problem as you know with collecting only 2600 player games is that some lines are not represented..
I forgot about that..I have zero experience with making books up to now..But I think i will start trying them out now..It was the same openings, but not complete doubles (Amy changed its mind at >move 15 and 22 I think). Zchess har learning and a slightly deeper book (you >can't determine depth) than Amy (used flatten 1).Well you didn't state that it was due to opening or anything..Amy lost 4-0 to Zchess
Or are the games doubles?
But if I understand you correctly, all engines use the same book..So, this >>probably means that the book suits Zchess better? Anyway the result is only a >>little surprising..
Again the result is a minor surprise..Just to clarify, Are you blaming Yace >>and Amy's poor result on the narrow book?
Finally, what is the point of your tournament, by making all the engines play >>using the same book?
Not all bookmakers used are similar. With some you can determine the number of >ply, with others you can't. Try creating a book for Green Light Chess with >2600.pgn using the bookcreator.
Not in the Yace-Bringer match, because the conditions are similar. But the >book differences I mentioned and the fact that some programs utilize book >and/or position learning constitutes an advantage. If I turn learning off I'll >just get a lot of duplicate games, which is uninteresting. So the experiment >has two factors; learning and bookdepth. One too many you might say, but >that's the way it is.
I'm hoping to get an impression of the strength of the new Zchess and Yace off >course. Furthermore, I'm hoping to get some information about the learning >implemented in various programs (Zchess seems to repeat winning moves).
I'm collecting games by ECO, but some openings are weakly represented when it >comes to the number of games.
Book learning will be delayed, until the Tour de France is over ...Besides a auto book learning feature, I'm not sure what other solution there >is..There's another similar solution, which Dieter is working on at the moment.
I'm also beginning to wonder if the book made out of p2600.pgn is a good book.
Since it is lacking in quite a few opening lines..Probably because GMs just
don't play certain lines..
Not in every opening. Sometimes it's good to get out of book early, because too much depth could result in a bad position for a computer program. Maybe depth is good in closed positions and bad in open positions, I don't know.I'm not familar with the differences in book creation..But are the differences that great?
Good..I will download the bigger Yace book, you have put up..See if this book allows Yace to play the Evan's Gambit properly..And whether there are other holes..
Perhaps..If you assume that the engines can play open positions perfectly well by itself..Not in every opening. Sometimes it's good to get out of book early, because >too much depth could result in a bad position for a computer program. Maybe >depth is good in closed positions and bad in open positions, I don't know.I'm not familar with the differences in book creation..But are the >>differences that great?
If we use the Evans gambit as an example: Would it be best if Yace started >thinking at depth 10 or depth 15? I have no idea, but I'm leaning towards >depth 15. What do you think?
You are looking for simple ideas..This is good..Book learning will be delayed, until the Tour de France is over ...
Currently, I am storing some flags in the book, that can be set by the user.
Something like
1. e4 f4?
(I parse ??, ?, ?!, !?, !, !!)
The idea is, to append some lines to the book.pgn, that are marked like this.
Then the book move selection code would use these hints. (Everything
implemented since a long time besides the actual usage of the hints).
Please suggest better ideas.
I think, you are correct. But it should be easy, to create your own book. You
can download any of the PGN files available on the net and try it out. If
you are creating a book frequently, or intend to add lines frequently, just
write a file bookc.inp like
And then at the command prompt:
yace bookc.inp
I suspect, that a larger maxply than 20 might be good for Yace.
To add the lines, you were missing, just add them at the beginning
or at the end of the book *three* times. (If the move count is one,
Yace will not use it, if it is two, Yace will only use it, when it is the
only book move available).
I'd be very interested in your experience with different books.
Do you mean 15 ply or 15 moves? The depth 30 ply will apply for all the games reaching 15 moves or higher. But in a lot of variations the depth would be less, because not all variations are represented, e.g. your Evans gambit example. So if you got a thin pgnfile, increasing depth would only help in common lines.One question though , when you set the opening book to draw from depth 15, wouldn't that apply across all opening?. Unless you intend to cut short games which you don't want too much depth..
Let's use ply.. I think 15 ply might be too little.20 is respectable..But of course 30 would be best...But I'm afraid the book will be too big..Do you mean 15 ply or 15 moves? The depth 30 ply will apply for all the games >reaching 15 moves or higher.One question though , when you set the opening book to draw from depth 15, >>wouldn't that apply across all opening?. Unless you intend to cut short games >>which you don't want too much depth..
But in a lot of variations the depth would be less, because not all variations >are represented, e.g. your Evans gambit example. So if you got a thin pgnfile, >increasing depth would only help in common lines.
Perhaps I was unclear. Just open the PGN file with your favoriteI'm sorry, I don't quite understand how you add lines to a book that >already exist..
Question: What is book.pgn?
Is that the file with the lines that you want to add?Question: What will this do?And then at the command prompt:
yace bookc.inp
I'm totally lost here..How do you add the new lines 3 times?
You can't with the book creation possibilities I'm aware of. E.g. Crafty, Bringer and Yace use a set ply depth. I can't remember what the Little Goliath bookmaker does, but you can choose between strong and wide. Don't ask me about the exact difference, except that strong results in fewer positions. I used wider for my tournament due to the number of games.My question is this..Isn't it pointless to ask if, you want the book to be of a certain length or ply or depth or move for a certain variation, since as I understand it, the setting depth/ply whatever is across the board..
For example..
if you decided that Evan gambit lines should be up to 15 ply, and Ruy lopez lines should be up to 20 ply, how could you do it??
That's what I thought.You can't with the book creation possibilities I'm aware of. E.g. Crafty, >Bringer and Yace use a set ply depth. I can't remember what the Little Goliath >bookmaker does, but you can choose between strong and wide. Don't ask me about >the exact difference, except that strong results in fewer positions. I used >wider for my tournament due to the number of games.My question is this..Isn't it pointless to ask if, you want the book to be of >>a certain length or ply or depth or move for a certain variation, since as I >>understand it, the setting depth/ply whatever is across the board..
For example..
if you decided that Evan gambit lines should be up to 15 ply, and Ruy lopez >>lines should be up to 20 ply, how could you do it??
Please start winboard with /debug and send me the winboard.deb and yace.logApparantly, after resigning and then resetting the board, yace has problems
restarting..
Return to Archive (Old Parsimony Forum)
Users browsing this forum: No registered users and 20 guests