Author: Robert Hyatt
Date: 12:58:54 12/18/03
Go up one level in this thread
On December 18, 2003 at 15:45:17, Peter Berger wrote:
>For information about setup and rules:
>http://www.talkchess.com/forums/1/message.html?336001
>
>As in every good human match there has to be some upset - the 17th game provided
>some.
>
>*Finally* Crafty chose another opening to escape the fruitless Ruy/Nimzo
>discussions. A quiet Guico this time.
>Crafty's 15. Qf3 looks somewhat useless - by move 19 black has probably reached
>easy equality.
>
>But this time Junior didn't find an appropriate plan IMHO and tried to play
>(commit suicide?) on the queenside. 19. ..Qa4 already looks somewhat dubious, as
>does 21. ..c5 , 24. ..c4, and then 25. ...b5 probably was the one mistake too
>much.
>Crafty reacted calmly without doing much itself.
>
>The position was still very tactical and complicated, and for some time it
>looked as if there were still hope for Junior. But it didn't work out. (Analysis
>left as exercise for the reader ;), maybe 27. ..Qxc3 still keeps some chances?
>).
>
>Anyway - at move 41 with Rxg7+ the game had been decided, both programs had
>white up by other three pawns, when .. - Junior and its computer .. froze.
>As this case had not been foreseen in the rules, I chose to use modified ICGA
>rules. I had to stop Crafty too to get the movelist from its gamefile.
>
>The game was resumed with 42 10 for Junior and 60 10 for Crafty by entering
>moves 1-41 in analysis mode on their respective computers (these the times on
>their clocks at time of crash).
>Crafty won the resumed game some 10 moves later.
>
>Now flame the referee ;O).
>
>The end of the match is approaching fast - will Junior manage to fight back
>another time ??
Did I misunderstand your strategy? IE when you said "first to six wins" I
was thinking the first to get 6 points ahead. Perhaps you meant the first
to win six games even if the match ends 6-5?
>
>(Btw, with the 3% chance Crafty has to play 2. Bc4 after 1. e4 e5 freezing might
>have been a useful "idea" ;)). What would the booklearner have done after an
>ordinary loss? Maybe given up on the succesful 1. ..e5 ? I don't know. As I
>don't know yet how the learner will deal with the resumed game. To not spoil it
>for me I won't have a look either.)
>
If you want to see the book learner work effectively, here is what to do
(if you play another match):
1. create a PGN file with just the PGN headers from one game, but no
moves at all.
2. "import pgnfile clear" which will clear all learning data, starting
everything off at zero.
3. Try these in your .craftyrc/crafty.rc:
bookw freq .20
bookw eval .10
bookw CAP .10
bookw learn 1.0
That will _really_ crank up the aggressiveness of the learning code. IE
if it wins a game, it will play the same game again next time unless the
opponent varies. If it loses, it will try a totally new first move. If
the book position ends up good or bad, it will try again or not. The
key is the bookw learn being the dominant term.
I don't play like that on ICC as it is way too strict and I want lots of
variability.
But it will make a difference, if you really think you like aggressive
learning. There is no way to be more aggressive than what the above
will do. :)
>
>I am sorry, but there has to be another short break due to operator holidays
>now. The 18th game will be played/posted either Sunday night or on Monday.
>
>Stay tuned :)
>Peter
>
>(all comments without any warranty at all of course)
>
>Standing: Junior 4 - Crafty 5 ( 8 draws)
>
>[Event "Blitz:120'+10"]
>[Site "Berlin"]
>[Date "2003.12.18"]
>[Round "17"]
>[White "Crafty 19.4"]
>[Black "Junior 8.0.0.2"]
>[Result "1-0"]
>[ECO "C54"]
>[Annotator "P233-PIV2.0GHz"]
>[PlyCount "103"]
>[EventDate "2003.??.??"]
>[TimeControl "2520+10"]
>
>{16MB, Junior8.ctg, P233MMX} 1. e4 e5 2. Bc4 Nf6 3. d3 Nc6 4. Nf3 Bc5 5. c3 d6
>6. O-O O-O 7. Nbd2 a6 8. Bb3 Ba7 9. Nc4 Be6 10. Bg5 h6 11. Bh4 g5 12. Bg3 Ne7
>13. Ne3 Qd7 14. Nd2 Ng6 15. Qf3 Kg7 16. h3 Nf4 17. Bxe6 fxe6 18. Bxf4 gxf4 19.
>Nc2 Qa4 20. Rfc1 Rg8 21. d4 c5 22. dxe5 dxe5 23. Ne1 Rad8 24. Rd1 c4 25. Qe2 b5
>26. b3 Qa5 27. bxc4 Kh8 28. cxb5 axb5 29. Rac1 Rd7 30. Rc2 Qb6 31. Rdc1 Qc5 32.
>Ndf3 Bb8 33. Rb1 Rb7 34. a4 Bd6 35. axb5 Kh7 36. c4 Rbg7 37. Kh1 Nd7 38. Rd1
>Bb8 39. Ng5+ Rxg5 40. Rxd7+ R8g7 41. Rxg7+ {0} Rxg7 {3.09/15 204} 42. Nd3 {187}
>Qb6 {3.55/15 157} 43. Nb4 {(Ta2) 132} Bd6 {3.13/15 240} 44. Nc6 {5} Qc5 {
>3.08/15 81} 45. Qh5 {(Ta2) 155} Qa3 {2.68/14 381} 46. Qd1 {1} Bc7 {3.54/14 235}
>47. c5 {(Td2) 5} Qa4 {3.60/14 94} 48. Qd3 {120} f3 {3.90/15 0} 49. g3 {115} h5
>{4.20/16 0} 50. Kh2 {(h4) 113} h4 {4.81/14 82} 51. g4 {24} Qa1 {5.13/14 27} 52.
>Qxf3 {183} 1-0
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.