If you have the source code than it is legal. If not than it is illegal.
Thomas
Toga II 3.1.2SE
Moderator: Ras
-
Isaksen
Re: A little test
Hi ThomasNice analogy, maybe you should turn it around. It works another way too.
-
Dirt
- Posts: 2851
- Joined: Wed Mar 08, 2006 10:01 pm
- Location: Irvine, CA, USA
Re: Toga II 3.1.2SE
The original authors can release their work under any license they decide to. The GPL puts no restrictions on them, since as the owners they don't need a license for distribution.jdart wrote:How can you close the source to Fruit if the original source was under GPL?
--Jon
Fruit 2.1 will still be GPL. The author could re-release it under a closed license, but that won't invalidate the GPL release.
-
Dirt
- Posts: 2851
- Joined: Wed Mar 08, 2006 10:01 pm
- Location: Irvine, CA, USA
Re: A little test
To keep your actions legal, don't give it to anyone else without including the source. If you don't have the source, don't further distribute the program.Gabinoz wrote:Hi Eelco.
Timecontrol: 5 minutes
A friend sent to me Toga II 3.1.2SE MP. Is this version legal?
Gabino
If you follow this rule I think you are ok using the program yourself in either case. Well, assuming this really is a version of Toga and not a renamed copy of another program.
-
Ovyron
- Posts: 4562
- Joined: Tue Jul 03, 2007 4:30 am
Re: A little test
Well, here come two bookless games played against Rybka 2.3.2a 1CPU 32bit at 1+'14 time control:Eelco de Groot wrote:I did not actually see much of the play yet of this particular version
[White "Toga Checkov Beta 3"]
[Black "Rybka 2.3.2a 32-bit "]
1. e4
{0.34/13 15}1... Nc6{0.06/13 8}2. d4{0.44/13 11}2... d5{0.06/12 16}3. exd5{0.56/12 15}3... Qxd5{0.15/13 10}4. Nf3{0.41/14 23}4... Bg4{0.07/13 16}5. Be2{0.46/13 16}5... Nf6{0.01/12 10}6. Nc3{0.33/13 19}6... Qa5{-0.04/13 10}7. h3{
0.56/12 10}7... Bf5{0.00/13 9}8. O-O{0.53/12 17}8... O-O-O{-0.03/12 8}9. Bb5{0.57/12 11}9... Nb4{0.00/11 3}10. Ne5{0.56/11 9}10... Nxc2{-0.14/12 12}11. Nxf7{0.25/11 15}11... Nxa1{-0.16/12 5}12. Nxh8{0.22/12 15}12... Nc2{0.00/13 14
}13. d5{0.37/13 14}13... Rxd5{0.08/12 4}14. Nxd5{0.37/12 15}14... Qxb5{0.07/11 0
}15. Nxf6{0.16/13 15}15... exf6{0.00/13 5}16. Qh5{0.00/14 16}16... Nd4{0.00/13 7}17. Qf7{0.55/12 14}17... Ne2+{0.00/14 6}18. Kh1{0.81/12 9}18... Bd6{0.01/14 5
}19. Qg8+{0.79/13 16}19... Kd7{0.00/13 0}20. Re1{0.77/13 18}20... Qe5{0.34/12 5
}21. Qxg7+{0.72/12 14}21... Kc6{0.00/13 4}22. f4{0.00/13 9}22... Qc5{0.00/13 4}23. Nf7{0.00/12 11
}23... Qf2{0.00/13 5}24. Rxe2{0.00/11 14}24... Qxe2{0.00/14 4}25. Nxd6{0.00/13 16}25... cxd6{0.00/15 44}26. Qxf6{0.00/14 14}26... Be4{0.00/15 7}27. Qc3+{0.00/14 18}27... Kb5{0.00/16 8}28. Qb3+{0.00/13 14}28... Ka5{0.00/18 8}29. Qa3+{0.00/14 20}29... Kb5{0.00/17 0}30. Qb3+{0.00/14 13}30... Ka5{0.00/22 4}31. Qa3+{0.00/15 20}31... Kb5{0.00/17 0}32. Qb3+{0.00/15 20
}32... Ka5{0.00/23 6}{33. Qa3+{0.00/16 0}{Draw by 3 times repetition.} 1/2-1/2
[White "Rybka 2.3.2a 32-bit "]
[Black "Toga Checkov Beta 3"]
1. Nc3{0.07/13 14}1... Nf6{0.18/14 16}2. d4{0.07/12 14}2... e6{0.19/14 13}3. g3{0.15/13 14}3... d5{-0.05/14 29}4. Bg2{0.15/13 4}4... c5{-0.10/13 14}5. Nf3{0.25/12 16}5... cxd4{-0.19/12 10}6. Nxd4{0.21/14 11}6... Nc6{-0.11/13 15}7. O-O{0.18/11 16}7... Bc5{-0.04/13 27}8. Be3{0.21/12 12}8... Qe7{-0.17/13 19}9. Na4{0.10/12 11}9... Bxd4{-0.19/12 9}10. Bxd4{0.09/10 0}10... Nxd4{
-0.06/13 10}11. Qxd4{0.09/10 0}11... Bd7{-0.01/14 17}12. Nc3{0.03/13 23}12... Bc6{-0.09/13 11}13. Rfd1{0.00/13 21}13... O-O{-0.11/12 14}14. Qe5{0.01/13 13}14... Ng4{-0.09/13 30}15. Qf4{0.00/13 6}15... h5{-0.17/12 9}16. h3{0.07/13 10}16... Nf6{-0.12/13 13}17. Qe5{0.07/12 3}17... h4{-0.21/11 14}18. g4{-0.04/12 24}18... Nd7{-0.26/12 9}19. Qe3{0.00/13 17}19... Qf6{-0.26/13 13}20. Rd4{0.00/13 21}20... Qg6{-0.22/13 14}21. Qd3{0.00/12 5}21... f5{-0.47/12 9}22. Qe3{-0.26/11 13}22... fxg4{-0.51/12 18}23. Rxg4{-0.23/12 6}23... Qf6{-0.50/13 14}24. Qf3{-0.22/12 3}24... Qh6{
-0.39/14 10}25. Qe3{0.00/17 12}25... Qxe3{-0.21/13 9}26. fxe3{0.10/14 7}26... Ne5{-0.24/14 15}27. Rxh4{0.07/13 3}27... Rac8{-0.18/14 17}28. Rf1{0.25/14 19}28... Rxf1+{-0.14/13 14}29. Kxf1{0.27/15 28}29... Nc4{-0.14/15 24}30. Nd1{0.32/13 11}30... g6{-0.08/14 14}31. b3{0.34/14 6}31... Nd6{0.00/14 9}32. Ke1{0.31/15 16}32... Kg7{-0.07/15 15}33. Kd2{0.26/15 27}33... e5{-0.17/14 9}34. a4{0.29/15 24}34... Rd8{-0.14/14 17}35. Nf2{0.31/15 21}35... a6{-0.10/14 18}36. Rg4{0.40/14 12}36... Bd7{-0.07/15 11}37. Rg5{0.38/14 12}37... Bf5{-0.09/14 14}38. Bh1{0.36/15 12}38... e4{-0.11/14 8}39. Rg1{0.27/15 10}39... Rh8{-0.09/14 15}40. Bg2{0.22/14 14}40... Bd7{-0.09/13 14}41. Bf1{0.29/15 17}41... Rf8{-0.30/15 11}42. Ng4{0.26/14 4}42... Bxg4{-0.18/15 11}43. hxg4{0.24/12 0}43... Rh8{-0.15/16 14}44. Kc3{0.27/17 19}44... Kf6{-0.03/16 9}45. Kd4{0.22/18 17}45... Ke6{0.10/17 11}46. c4{0.21/17 7}46... dxc4{-0.13/12 4}47. bxc4{0.19/15 0}47... b6{0.00/15 11}48. Bg2{0.28/16 5}48... Rc8{0.00/16 10}49. Rc1{0.28/15 4}49... Rd8{0.00/15 15}50. g5
{0.31/14 9}50... Nxc4+{0.21/14 37}51. Kxe4{0.28/13 0}51... b5{0.21/14 14}52. axb5{0.32/14 4}52... axb5{0.09/14 14}53. Bh3+{0.30/15 18}53... Ke7{0.18/15 14}54. Bg4{0.29/16 27}54... Re8{0.22/13 10}
55. Ra1{0.28/15 23}55... b4{0.24/12 14}56. Ra7+{0.27/15 18}56... Kf8+{0.07/15 14}57. Kd4{0.27/16 5}57... Nxe3{0.06/15 14}58. Bf3{0.20/15 17}58... Nf5+{0.01/16 18}59. Kc4{0.20/15 4}59... Rb8{0.02/16 15}60. Rh7{0.16/16 16}60... Kg8{0.00/17 19}61. Rh3{0.10/18 21}61... b3{0.00/17 9}62. Bd5+{0.08/18 4}62... Kg7{0.00/18 3}63. Rxb3{0.06/19 5}63... Rxb3{0.00/21 14}64. Kxb3{0.05/18 0}64... Nd4+{0.00/26 10}65. Kc4{0.00/26 12}65... Nxe2{0.00/30 11}66. Kb4{0.00/28 12}66... Nf4{0.00/33 11}67. Kc4{0.00/27 4}67... Ne2{0.00/32 12}68. Kb5{0.00/31 15}68... Nf4{0.00/34 10}69. Kc4{0.00/33 13}69... Ne2{0.00/34 11}70. Kc5{0.00/32 13}70... Nf4{0.00/25 0}{Draw by 3 times repetition.} 1/2-1/2
Rybka couldn't beat it
Your beliefs create your reality, so be careful what you wish for.
-
Edward German
Re: Toga Checkov Beta 3f 'prerelease'
Hello Eelco!Eelco de Groot wrote:Hello Eduard, thanks for the interest but what are you doing on that pirate site? You know the best clones can be found right here! And we have the best arguments about it too. Who is this Tony T. character anyway, sounds like a false name to me... Did he not play in one of the Godfather movies? Or was that Tony Danza?Edward German wrote:Hi Eelco!
Your Engine "Toga Checkov Beta 3f 'prerelease" (Prerelease) on the site by Tony T., does not work on my Intel Core 2 Duo. Can You it new compile? I do not know what ist here wrong. UCI does not work.
Best!
But more seriously, thanks to Denis Mendoza there are now I think much faster compiles and I hope the Core 2 Duo version he compiled with the actual Intel compiler will be to your liking. Thanks Denis! Great Job!
Eelco
First: I really does not know that the one site is a pirate site.
An friend tell me yesterday per phone every Things about this site. I was very surprise about the things that he4 say to me.
I would such Toga only, not more, not the other things there...
Second: Now I found by CCRL Forum a Core 2 Duo "Toga_Checkov_Core2_SSE3!!!" Version (Single) with Checkov Settings. It works now fine on my Core 2 Duo (of Cource Single only). Many Thanks to Mr. Mendoza for his nice Work.
The Engine is very fast but enyway it does not blundered. It will be nice to see this version as MP too. Is that possible?
Many Thanks!
-
Edward German
Re: Toga Checkov Beta 3f 'prerelease'
Hi Tony!Tony Thomas wrote:Why do you guys keep saying that the guys name is Tony T. His name Jonas Conas or something. Again, its a pirate site, stop visiting it.Edward German wrote:Hi Eelco!
Your Engine "Toga Checkov Beta 3f 'prerelease" (Prerelease) on the site by Tony T., does not work on my Intel Core 2 Duo. Can You it new compile? I do not know what ist here wrong. UCI does not work.
Best!
Read my answer here:
http://www.talkchess.com/forum/viewtopi ... 06&t=18813
And sorry for my bad english.
Yesterday I thinks that the one site XXX are a site from You. Is not or yes?
Eduard.
-
Eelco de Groot
- Posts: 4697
- Joined: Sun Mar 12, 2006 2:40 am
- Full name: Eelco de Groot
Re: Toga Checkov Beta 3f 'prerelease'
Hello Eduard,Edward German wrote: Hello Eelco!
First: I really does not know that the one site is a pirate site.
An friend tell me yesterday per phone every Things about this site. I was very surprise about the things that he4 say to me.
I would such Toga only, not more, not the other things there...
Second: Now I found by CCRL Forum a Core 2 Duo "Toga_Checkov_Core2_SSE3!!!" Version (Single) with Checkov Settings. It works now fine on my Core 2 Duo (of Cource Single only). Many Thanks to Mr. Mendoza for his nice Work.
The Engine is very fast but enyway it does not blundered. It will be nice to see this version as MP too. Is that possible?
Many Thanks!
You are welcome! It was really much fun creating 'my own' Toga version engine but my programming efforts so far in it have been negligible. Still strugling with the basic concepts. I think Anthony would refer to me as one of the "pointerless" people
Pointers, Ugghh! He is so right of course.I have not made any changes that could not be transferred to an SMP Toga or something like that.
I believe that Thomas Gaksch is working on the Toga 1.4 MP and wants to release it very soon. It does not scale very well but many people would like to test more cores of course. I can't think of a fundamental reason why Toga should not be scaling well with a well chosen MP algorithm but you never know it. Maybe the author of Cluster Toga could tell us more.
Eelco
-
Eelco de Groot
- Posts: 4697
- Joined: Sun Mar 12, 2006 2:40 am
- Full name: Eelco de Groot
Re: A little test
Hi Ulysses, yes, I had seen the post, I had not realized that could be you. Thank you very much, it is sort of my favourite post on the forumOvyron wrote:Well, here come two bookless games played against Rybka 2.3.2a 1CPU 32bit at 1+'14 time control:Eelco de Groot wrote:I did not actually see much of the play yet of this particular version
[White "Toga Checkov Beta 3"]
[Black "Rybka 2.3.2a 32-bit "]
1. e4
{0.34/13 15}1... Nc6{0.06/13 8}2. d4{0.44/13 11}2... d5{0.06/12 16}3. exd5{0.56/12 15}3... Qxd5{0.15/13 10}4. Nf3{0.41/14 23}4... Bg4{0.07/13 16}5. Be2{0.46/13 16}5... Nf6{0.01/12 10}6. Nc3{0.33/13 19}6... Qa5{-0.04/13 10}7. h3{
0.56/12 10}7... Bf5{0.00/13 9}8. O-O{0.53/12 17}8... O-O-O{-0.03/12 8}9. Bb5{0.57/12 11}9... Nb4{0.00/11 3}10. Ne5{0.56/11 9}10... Nxc2{-0.14/12 12}11. Nxf7{0.25/11 15}11... Nxa1{-0.16/12 5}12. Nxh8{0.22/12 15}12... Nc2{0.00/13 14
}13. d5{0.37/13 14}13... Rxd5{0.08/12 4}14. Nxd5{0.37/12 15}14... Qxb5{0.07/11 0
}15. Nxf6{0.16/13 15}15... exf6{0.00/13 5}16. Qh5{0.00/14 16}16... Nd4{0.00/13 7}17. Qf7{0.55/12 14}17... Ne2+{0.00/14 6}18. Kh1{0.81/12 9}18... Bd6{0.01/14 5
}19. Qg8+{0.79/13 16}19... Kd7{0.00/13 0}20. Re1{0.77/13 18}20... Qe5{0.34/12 5
}21. Qxg7+{0.72/12 14}21... Kc6{0.00/13 4}22. f4{0.00/13 9}22... Qc5{0.00/13 4}23. Nf7{0.00/12 11
}23... Qf2{0.00/13 5}24. Rxe2{0.00/11 14}24... Qxe2{0.00/14 4}25. Nxd6{0.00/13 16}25... cxd6{0.00/15 44}26. Qxf6{0.00/14 14}26... Be4{0.00/15 7}27. Qc3+{0.00/14 18}27... Kb5{0.00/16 8}28. Qb3+{0.00/13 14}28... Ka5{0.00/18 8}29. Qa3+{0.00/14 20}29... Kb5{0.00/17 0}30. Qb3+{0.00/14 13}30... Ka5{0.00/22 4}31. Qa3+{0.00/15 20}31... Kb5{0.00/17 0}32. Qb3+{0.00/15 20
}32... Ka5{0.00/23 6}{33. Qa3+{0.00/16 0}{Draw by 3 times repetition.} 1/2-1/2
[White "Rybka 2.3.2a 32-bit "]
[Black "Toga Checkov Beta 3"]
1. Nc3{0.07/13 14}1... Nf6{0.18/14 16}2. d4{0.07/12 14}2... e6{0.19/14 13}3. g3{0.15/13 14}3... d5{-0.05/14 29}4. Bg2{0.15/13 4}4... c5{-0.10/13 14}5. Nf3{0.25/12 16}5... cxd4{-0.19/12 10}6. Nxd4{0.21/14 11}6... Nc6{-0.11/13 15}7. O-O{0.18/11 16}7... Bc5{-0.04/13 27}8. Be3{0.21/12 12}8... Qe7{-0.17/13 19}9. Na4{0.10/12 11}9... Bxd4{-0.19/12 9}10. Bxd4{0.09/10 0}10... Nxd4{
-0.06/13 10}11. Qxd4{0.09/10 0}11... Bd7{-0.01/14 17}12. Nc3{0.03/13 23}12... Bc6{-0.09/13 11}13. Rfd1{0.00/13 21}13... O-O{-0.11/12 14}14. Qe5{0.01/13 13}14... Ng4{-0.09/13 30}15. Qf4{0.00/13 6}15... h5{-0.17/12 9}16. h3{0.07/13 10}16... Nf6{-0.12/13 13}17. Qe5{0.07/12 3}17... h4{-0.21/11 14}18. g4{-0.04/12 24}18... Nd7{-0.26/12 9}19. Qe3{0.00/13 17}19... Qf6{-0.26/13 13}20. Rd4{0.00/13 21}20... Qg6{-0.22/13 14}21. Qd3{0.00/12 5}21... f5{-0.47/12 9}22. Qe3{-0.26/11 13}22... fxg4{-0.51/12 18}23. Rxg4{-0.23/12 6}23... Qf6{-0.50/13 14}24. Qf3{-0.22/12 3}24... Qh6{
-0.39/14 10}25. Qe3{0.00/17 12}25... Qxe3{-0.21/13 9}26. fxe3{0.10/14 7}26... Ne5{-0.24/14 15}27. Rxh4{0.07/13 3}27... Rac8{-0.18/14 17}28. Rf1{0.25/14 19}28... Rxf1+{-0.14/13 14}29. Kxf1{0.27/15 28}29... Nc4{-0.14/15 24}30. Nd1{0.32/13 11}30... g6{-0.08/14 14}31. b3{0.34/14 6}31... Nd6{0.00/14 9}32. Ke1{0.31/15 16}32... Kg7{-0.07/15 15}33. Kd2{0.26/15 27}33... e5{-0.17/14 9}34. a4{0.29/15 24}34... Rd8{-0.14/14 17}35. Nf2{0.31/15 21}35... a6{-0.10/14 18}36. Rg4{0.40/14 12}36... Bd7{-0.07/15 11}37. Rg5{0.38/14 12}37... Bf5{-0.09/14 14}38. Bh1{0.36/15 12}38... e4{-0.11/14 8}39. Rg1{0.27/15 10}39... Rh8{-0.09/14 15}40. Bg2{0.22/14 14}40... Bd7{-0.09/13 14}41. Bf1{0.29/15 17}41... Rf8{-0.30/15 11}42. Ng4{0.26/14 4}42... Bxg4{-0.18/15 11}43. hxg4{0.24/12 0}43... Rh8{-0.15/16 14}44. Kc3{0.27/17 19}44... Kf6{-0.03/16 9}45. Kd4{0.22/18 17}45... Ke6{0.10/17 11}46. c4{0.21/17 7}46... dxc4{-0.13/12 4}47. bxc4{0.19/15 0}47... b6{0.00/15 11}48. Bg2{0.28/16 5}48... Rc8{0.00/16 10}49. Rc1{0.28/15 4}49... Rd8{0.00/15 15}50. g5
{0.31/14 9}50... Nxc4+{0.21/14 37}51. Kxe4{0.28/13 0}51... b5{0.21/14 14}52. axb5{0.32/14 4}52... axb5{0.09/14 14}53. Bh3+{0.30/15 18}53... Ke7{0.18/15 14}54. Bg4{0.29/16 27}54... Re8{0.22/13 10}
55. Ra1{0.28/15 23}55... b4{0.24/12 14}56. Ra7+{0.27/15 18}56... Kf8+{0.07/15 14}57. Kd4{0.27/16 5}57... Nxe3{0.06/15 14}58. Bf3{0.20/15 17}58... Nf5+{0.01/16 18}59. Kc4{0.20/15 4}59... Rb8{0.02/16 15}60. Rh7{0.16/16 16}60... Kg8{0.00/17 19}61. Rh3{0.10/18 21}61... b3{0.00/17 9}62. Bd5+{0.08/18 4}62... Kg7{0.00/18 3}63. Rxb3{0.06/19 5}63... Rxb3{0.00/21 14}64. Kxb3{0.05/18 0}64... Nd4+{0.00/26 10}65. Kc4{0.00/26 12}65... Nxe2{0.00/30 11}66. Kb4{0.00/28 12}66... Nf4{0.00/33 11}67. Kc4{0.00/27 4}67... Ne2{0.00/32 12}68. Kb5{0.00/31 15}68... Nf4{0.00/34 10}69. Kc4{0.00/33 13}69... Ne2{0.00/34 11}70. Kc5{0.00/32 13}70... Nf4{0.00/25 0}{Draw by 3 times repetition.} 1/2-1/2
Rybka couldn't beat it
Vasik Rajlich has not posted anymore since this debacle! I was really a bit worried but now I'm more philosophical about it. My thinking is if Vasik really succeeded in translating ideas of Fruit into a bitboard version, any new idea he could probably also translate. So it is really up to the basic efficiency of bitboards how much more Vasik could get out of it, at least in my simple theory. So not so worried anymore. We are going to beat them Russians 
A very strange thing has occurred in the latest version. I thought I had made the evaluation function more complex, it now counts pawns attacking the King again, and I can accept that this would translate to some better tactical results. But what I also see is that unexpectedly I get a higher nodes per second, significantly higher, running the Arasan 7 test
How can that be? Unless my computer is playing up, it is not impossible, I had some problems that could be caused by the compiling efforts with MSVC. The way I compiled the sources is still the same. I mean it is not unwelcome but still, it reminded me a bit of that Terminator movie where the general at Cyberdyne presses a button to fight of a global virus that theatens to invade key defense computers and with that press of a button, the computers are allowed to take control, creating our worst nightmare; the Skynet trying to terminate man! Brrr!Code: Select all
Toga Checkov Beta (3.2.1 SE) 128 Mb HT 20 sec per pos Version of 14-01-08 16:13, as before but
// Tempo
if (COLOUR_IS_WHITE(board->turn)){
opening += 25;
endgame += 15;
}
else{
opening -= 20;
endgame -= 15;
Right until now: 170 of 227 ; 29:53m
I/O Fault!
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
-------------------------------------------------------------------------------------
0 | 0 0 0 0 2 4 3 0 2 2 3 0 0 0 0 - 15 1 0 2
20 | 0 - 5 - - 6 1 0 2 - 4 0 6 2 4 1 11 1 0 7
40 | 2 - 2 0 0 - 0 1 10 0 0 0 11 0 8 5 0 - - 0
60 | 0 13 0 2 - 1 19 - 1 - 7 0 11 0 4 0 - 5 - 17
80 | - 3 - - 0 4 0 3 0 1 14 0 0 - 3 1 - 12 0 1
100 | 0 0 0 0 - 3 0 1 7 0 6 6 0 2 - 14 - 3 19 0
120 | 0 - 7 15 1 - 10 5 6 - - 3 1 5 1 1 - 0 - 0
140 | 13 0 3 0 0 8 11 - - - - 0 3 0 - 1 0 - 1 -
160 | 1 2 4 - 12 - 1 19 - 0 2 0 - - - 6 0 - 11 7
180 | 0 10 - 0 7 3 - - - - 1 0 2 4 7 - - 0 - 1
200 | 0 16 4 0 0 - - 9 0 2 0 - 3 18 2 1 - 0 - -
220 | 14 3 - 6 3 0 0
1 sec -> 66/227
2 sec -> 88/227
3 sec -> 103/227
4 sec -> 117/227
5 sec -> 125/227
6 sec -> 130/227
7 sec -> 137/227
8 sec -> 144/227
9 sec -> 146/227
10 sec -> 147/227
11 sec -> 150/227
12 sec -> 155/227
13 sec -> 157/227
14 sec -> 159/227
15 sec -> 162/227
20 sec -> 170/227
n/s: 555.965
TotTime: 76:04m SolTime: 29:53m
Toga Checkov Beta 4 (3.2.1 SE) 128 Mb HT 20 sec per pos. Version from 17-01-08 4:50, almost as previous Beta 3 but new way of adding enemy pawnattacks on the own King.
Right until now: 176 of 227 ; 27:26m
I/O Fault!
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
-------------------------------------------------------------------------------------
0 | 0 0 3 0 5 0 2 0 2 0 1 0 0 0 0 - 15 1 0 2
20 | 0 - 4 - - 4 1 1 1 - 4 0 7 1 3 0 9 1 0 5
40 | 1 - 0 0 0 2 0 1 10 0 0 0 - 0 7 5 0 - 10 0
60 | 0 11 0 1 - 1 - - 4 - 1 0 8 0 10 0 - 4 16 10
80 | - 2 - - 0 3 0 2 0 0 - 0 0 - 2 0 - 9 0 0
100 | 0 0 0 18 15 2 0 1 1 0 5 4 0 2 6 11 - 2 - 0
120 | 0 - 3 12 0 - 5 4 6 - - 1 0 3 1 1 - 0 16 0
140 | - 0 2 0 2 1 9 - - - 12 0 2 0 - 1 0 - 1 -
160 | 1 1 6 - 9 4 1 2 17 0 - 0 - - 19 4 0 - 10 10
180 | 0 0 - 0 17 0 - 1 - - 1 0 2 1 5 - - 0 - 1
200 | 0 11 4 0 0 - - 8 0 1 0 18 3 14 2 1 - 0 18 -
220 | - 16 - 2 1 0 0
1 sec -> 73/227
2 sec -> 103/227
3 sec -> 120/227
4 sec -> 126/227
5 sec -> 136/227
6 sec -> 142/227
7 sec -> 145/227
8 sec -> 147/227
9 sec -> 149/227
10 sec -> 153/227
11 sec -> 159/227
12 sec -> 162/227
13 sec -> 164/227
14 sec -> 164/227
15 sec -> 165/227
20 sec -> 176/227
n/s: 691.610
TotTime: 75:59m SolTime: 27:26mBeta 4: n/s: 691.610
Is that not very strange? I have no explanation for it.
Regards, Eelco
-
Eelco de Groot
- Posts: 4697
- Joined: Sun Mar 12, 2006 2:40 am
- Full name: Eelco de Groot
Re: A little test
(The above differences in NPS probably had to do with an opened Internet Explorer 7 and totally messed with my testing, by taking out big chunks of nodes while there was nothing supposed to be going on in the browser. M$&&%^*[edited by Moderation]!... 
)
I don't have very interesting news, but just wanted to tell a little about the latest experiments with the code. I just started experimenting a bit with Toga's extended history pruning and I am probably introducing horrible horizon effects. But I found it interesting that with little changes you can get big differences in the PV and in times for new iterations. There goes little knowledge in the History reductions so many of the effects may well be random, but still they might have some function in shaping the tree.
If I find something that actually works I think I will call it Fischer Random Reductions
I am trying it out on Will Singleton's testposition, I don't get very fast times for finding 1. e5 because the hardware is much slower than what Johan has I think. I think I will follow a chronological order of new Beta versions:
The regular 3.1.2 Toga is fastest but it may be because it does not rate Rxd7 very high. Probably just because of the different settings. I let the computer calculate a fairly long time:
[d]r1b2q1k/2Qn1p1p/1p1Rpp2/p6B/4P2P/6N1/P4PP1/6K1 w - -
Engine: Toga II 3.1.2SE (256 MB, Athlon 2009 MHz)
by Thomas Gaksch and Fabien Letouzey
9/27 0:00 +1.39 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qc7 Qd6
4.Qxf7 Rf8 5.Qb7 Rb8 6.Qa7 Rg8 (171.458)
10/36 0:00 +1.44 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qxf7 Qxf7
4.Bxf7 e5 5.Nf5 Rd2 6.g4 b5 7.Kg2 b4
8.Bc4 a4 (447.814)
11/36 0:01 +1.50 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qxf7 Qxf7
4.Bxf7 Rd1+ 5.Kh2 Rd2 6.f3 e5 7.Nh5 b5
8.Kg3 Rd6 9.Bd5 f5 (834.403) 820
12/36 0:01 +1.47 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qxf7 Qxf7
4.Bxf7 Rd1+ 5.Kh2 e5 6.Nf5 Rd2 7.Kg3 b5
8.Kf3 Rc2 9.Bb3 Rd2 10.Bd5 (1.449.584) 812
13/36 0:03 +1.44 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qxf7 Qxf7
4.Bxf7 Rd1+ 5.Kh2 Rd6 6.Nh5 e5 7.Kg3 b5
8.Kf3 h6 9.Bd5 Kh7 10.g4 Kg6 (2.940.927) 823
14/44 0:06 +1.45 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qxf7 Qxf7
4.Bxf7 Rd1+ 5.Kh2 Rd6 6.Nh5 e5 7.Kg3 b5
8.Kf3 h6 9.Bd5 Kh7 10.Ng3 Kg6 11.Nf5 (5.604.835) 828
15/48 0:26 +1.34 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Kh2 Rd8
4.Qb7 Rd2 5.Qxb6 Rxa2 6.Bg4 a4
7.Nh5+ Kh8 8.Qd4 a3 9.Qxf6+ Kg8
10.Kg1 Rb2 11.Qg5+ Kh8 12.Qe5+ Kg8
13.Bd1 h6 14.Nf6+ (21.739.056) 831
15/48 0:53 +1.85++ 1.e5 fxe5 2.Ne4 Kg7 3.Rxd7 Bxd7
4.Qxd7 Rd8 5.Qc7 Rc8 6.Qxe5+ Kg8
7.Ng5 Rc1+ 8.Kh2 f6 9.Qxe6+ Kh8
10.Nf7+ Kg8 11.Qxb6 Rf1 12.Qxa5 (42.880.861) 800
16/62 2:54 +2.50++ 1.e5 fxe5 2.Ne4 h6 3.Rd3 Kh7 4.Nd6 e4
5.Rg3 Ra7 6.Qxa7 Qxd6 7.Qa8 Qf8
8.Qxe4+ Kh8 9.Rf3 f5 10.Qxe6 (138.515.322) 795
17/62 3:32 +2.65 1.e5 fxe5 2.Ne4 h6 3.Rd3 f6 4.Nd6 e4
5.Nf7+ Kh7 6.Rxd7 Bxd7 7.Qxd7 Qg7
8.Qxe6 Rf8 9.Qxe4+ Kg8 10.Nd6 Qc7
11.Qf4 Kh7 12.Bf3 (168.928.249) 795
18/62 4:45 +2.73 1.e5 fxe5 2.Ne4 h6 3.Rd3 f6 4.Nd6 e4
5.Nf7+ Kh7 6.Rxd7 Bxd7 7.Qxd7 Qg7
8.Qxe6 Rf8 9.Qxe4+ Kg8 10.Nd6 Qc7
11.Qf4 Qc2 12.Qxh6 Qb1+ 13.Kh2 Qxa2 (227.605.827) 796
19/68 14:05 +3.27++ 1.e5 fxe5 2.Ne4 h6 3.Rd3 f6 4.Nd6 e4
5.Rc3 f5 6.Nxc8 Nf6 7.Qb7 e3 8.Qxa8 exf2+
9.Kf1 Nxh5 10.Nxb6 Qxa8 11.Nxa8 Nf6
12.Rc8+ Kg7 13.Kxf2 (684.184.547) 808
20/68 25:46 +3.40 1.e5 fxe5 2.Ne4 h6 3.Rd3 Kh7 4.Rg3 f5
5.Nd6 Ra7 6.Qxa7 Qxd6 7.Rc3 Qf8
8.Qc7 Kg7 9.Qxc8 Qxc8 10.Rxc8 Kf6
11.Rg8 e4 12.Rg6+ Ke5 13.Rxh6 b5 (1.251.309.144) 809
21/74 49:45 +3.28 1.e5 fxe5 2.Ne4 Rb8 3.Rxd7 Bxd7
4.Qxe5+ f6 5.Nxf6 Rc8 6.Nxd7+ Qg7
7.Qxe6 b5 8.Ne5 Rf8 9.g3 b4 10.Kg2 a4
11.Nf7+ Kg8 12.Ng5+ Kh8 13.Qe4 Rb8
14.Bd1 (2.425.537.205) 812
22/78 156:14 +4.01++ 1.e5 fxe5 2.Ne4 h6 3.Rd3 f6 4.Nd6 a4
5.Bf7 Nc5 6.Rg3 Bd7 7.Rg8+ Qxg8
8.Bxg8 Rxg8 9.Qxb6 Nd3 10.Nf7+ Kh7
11.Qa7 Nf4 12.g3 Nh3+ 13.Kg2 Nf4+
14.Kh2 (7.592.749.433) 809
53 seconds is the fastest time and after 2½ hours the evaluation goes past +4
This is Toga Beta 4 with no changes in the History reductions, needs 3 minutes 45 seconds which is a bit slow:
r1b2q1k/2Qn1p1p/1p1Rpp2/p6B/4P2P/6N1/P4PP1/6K1 w - -
Engine: Toga Checkov Beta 4 (256 MB)
by Thomas Gaksch and Fabien Letouzey
9/28 0:00 +1.72++ 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Rd2
4.Bxf7 Qc5 5.Qa8+ Kg7 6.Qg8+ Kh6
7.Kh2 Rxf2 8.Bxe6 (225.745)
10/32 0:01 +1.70 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Qa4 Kg8 6.Be2 Kh8 7.Qa3 Qd2 (767.513) 700
11/32 0:01 +1.60 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Rd8
7.Bf5 Rd1+ 8.Kh2 Rd2 9.Nxf6 Rxf2
10.Nxh7+ Ke7 (1.298.913) 699
12/38 0:03 +1.67 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rc8 6.Nf1 Rd8 7.e5 fxe5
8.Ne3 Rd7 (2.325.144) 670
13/41 0:05 +1.73 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rc8 6.Nf1 Rd8 7.Ne3 Rc8
8.g4 Qe5 9.Qd7 Rc1+ 10.Nd1 (3.748.755) 688
14/41 0:08 +1.66 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.e5 fxe5 7.Ne4 Qd7
8.Qxd7 Rxd7 9.Bxe6 Rd1+ 10.Kh2 Kg7
11.Kg3 (5.561.364) 666
15/52 0:44 +1.56 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Rd8
7.Bf5 Rd1+ 8.Kh2 Ke7 9.a4 Rd4 10.Nf4 Rxa4
11.Nd5+ Kd6 12.Bxh7 b5 13.Nxf6 Ra2 (32.640.962) 729
15/52 0:58 +1.72 1.Rd1 Rb8 2.Rxd7 Bxd7 3.Qxd7 Kg7
4.Bg4 Qd8 5.Nh5+ Kf8 6.Nxf6 Rc8
7.Nxh7+ Kg7 8.Qxd8 Rxd8 9.Ng5 Rd2
10.Bh5 Rxa2 11.Bxf7 (42.511.906) 725
16/52 1:09 +1.79 1.Rd1 Rb8 2.Rxd7 Bxd7 3.Qxd7 Kg7
4.Bg4 Qd8 5.Nh5+ Kf8 6.Nxf6 Rc8
7.Nxh7+ Kg7 8.Qxd8 Rxd8 9.Ng5 Rc8
10.g3 Rc2 11.a4 Kg6 12.Kg2 (50.314.633) 724
17/52 1:31 +1.75 1.Rd1 Rb8 2.Rxd7 Bxd7 3.Qxd7 Kg7
4.Bg4 Rd8 5.Nh5+ Kg6 6.Qa7 Qd6 7.g3 Rd7
8.Qa8 Rd8 9.Qb7 Rd7 10.Qc8 Rd8
11.Qc4 (66.675.412) 726
17/56 3:45 +2.19++ 1.e5 fxe5 2.Ne4 Kg7 3.Rd1 a4 4.Nd6 Nf6
5.Bxf7 Kh8 6.Ne4 Kg7 7.Bh5+ Kg8
8.Nxf6+ Qxf6 9.Qxb6 Kf8 10.Rd8+ Kg7
11.Qc7+ Kh6 12.Bg4 Qxh4 (163.031.929) 721
18/76 22:41 +2.86++ 1.e5 fxe5 2.Ne4 h6 3.Rd1 f6 4.Nd6 f5
5.Qc6 Rb8 6.Nf7+ Kg7 7.Qxe6 Nf6
8.Qxe5 Rb7 9.Qg3+ Ng4 10.Nd6 Rd7
11.f3 Rxd6 12.Rxd6 (976.271.108) 716
The 4f version is the main testing version, there are no changes in Extended History Pruning yet but some changes in counting pawn attacks on the own King. Luckily it is a bit faster here, 1 minute 51 seconds. Rxd7 gets higher evals than in the Beta 4.
r1b2q1k/2Qn1p1p/1p1Rpp2/p6B/4P2P/6N1/P4PP1/6K1 w - -
Engine: Toga Checkov Beta 4f (256 MB)
by Thomas Gaksch and Fabien Letouzey
9/28 0:00 +1.70++ 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Rd2
4.Bxf7 Qc5 5.Qa8+ Kg7 6.Qg8+ Kh6
7.Kh2 Rxf2 8.Bxe6 (254.154)
10/32 0:00 +1.66 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rf8 5.Nf1 Rb8 6.Qa7 Rc8 7.a4 (437.458)
11/38 0:01 +1.63 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.h5 Rd7 7.Qa8+ Rd8
8.Qb7 Qd1+ 9.Nf1 (803.545) 720
12/42 0:01 +1.57 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.h5 Rd7 7.Qa8+ Rd8
8.Qb7 Qe5 9.Qxb6 Rd1+ 10.Nf1 (1.403.676) 706
13/46 0:19 +1.62 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.Nxb6 Rxf2 12.Nc4+ Kc5 (14.265.133) 732
14/46 0:24 +1.62 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.Nxb6 Rxf2 12.Nc4+ Kc5 (18.010.830) 736
15/48 0:35 +1.64 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.a4 Rxf2 12.Bxh7 Kc5
13.h5 Rd2 (26.100.890) 743
16/50 1:13 +1.66 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.a3 Rxf2 12.Nxb6 h5
13.Nc4+ Kc5 14.Nxa5 (54.345.814) 738
16/58 1:51 +2.04++ 1.e5 fxe5 2.Ne4 Kg8 3.Ng5 Ra7
4.Bxf7+ Kh8 5.Qc6 Nf6 6.Bxe6 Bxe6
7.Nxe6 Qf7 8.Rd8+ Ng8 9.Qxb6 Rd7
10.Rxd7 Qxd7 11.Nc5 Ne7 (81.117.239) 729
17/66 7:31 +2.69++ 1.e5 fxe5 2.Ne4 h6 3.Rd1 f6 4.Nd6 f5
5.a4 e4 6.Qc3+ Kg8 7.Bf7+ Kh7 8.Qc6 Rb8
9.Bxe6 Ne5 10.Qc7+ Qg7 11.Bxf5+ Bxf5
12.Qxb8 (326.650.527) 723
In Beta 4g I tried some changes in extended History Pruning, probably far too much, with double reductions, but this version is very quick in reaching 19 plies compared with default 3.1.2 SE
r1b2q1k/2Qn1p1p/1p1Rpp2/p6B/4P2P/6N1/P4PP1/6K1 w - -
Engine: Toga Checkov Beta 4g (256 MB)
by Thomas Gaksch and Fabien Letouzey
9/28 0:00 +1.70++ 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Rd2
4.Bxf7 Qc5 5.Qa8+ Kg7 6.Qg8+ Kh6
7.Kh2 Rxf2 8.Bxe6 (254.154)
10/32 0:00 +1.66 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rf8 5.Nf1 Rb8 6.Qa7 Rc8 7.a4 (437.458)
11/38 0:01 +1.63 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.h5 Rd7 7.Qa8+ Rd8
8.Qb7 Qd1+ 9.Nf1 (803.545) 600
12/42 0:02 +1.57 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.h5 Rd7 7.Qa8+ Rd8
8.Qb7 Qe5 9.Qxb6 Rd1+ 10.Nf1 (1.403.676) 595
13/46 0:19 +1.62 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.Nxb6 Rxf2 12.Nc4+ Kc5 (11.925.715) 603
14/46 0:25 +1.62 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.Nxb6 Rxf2 12.Nc4+ Kc5 (15.518.833) 606
15/46 0:40 +1.77 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.Kg3 Rxa2 12.Bxh7 Rb2
13.h5 a4 14.Nxf6 (25.177.180) 615
16/48 1:26 +1.56 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.a4 Rd1+ 9.Kh2 Ra1
10.Bxh7 Rxa4 11.Nf4 Kd6 12.Nd5 b5
13.Nxf6 Ra2 (53.879.012) 624
16/50 2:46 +2.17++ 1.e5 fxe5 2.Ne4 Kg8 3.Ng5 Ra7
4.Bxf7+ Kh8 5.Qc6 Nf6 6.Bxe6 Ne8
7.Rd8 Kg7 (100.927.144) 607
17/70 14:52 +2.81++ 1.e5 fxe5 2.Ne4 h6 3.Rd1 f5 4.Nd6 e4
5.Qc3+ Qg7 6.Qc6 Rb8 7.Nxc8 Nc5
8.Rd8+ Kh7 9.Qd6 Rb7 10.Nxb6 Qa1+
11.Bd1 (556.644.438) 623
18/72 21:20 +3.30++ 1.e5 fxe5 2.Ne4 h6 3.Rd1 f6 4.Nd6 e4
5.Nxe4 Qg7 6.Nd6 Nf8 7.Qc6 Rb8
8.Nxc8 Qb7 9.Qxb7 Rxb7 10.Bf3 Rb8
11.Nd6 (843.819.035) 659
19/72 24:17 +3.29 1.e5 fxe5 2.Ne4 h6 3.Rd1 f6 4.Nd6 e4
5.Nxe4 Qg7 6.Nd6 Nf8 7.Qc6 Rb8
8.Nxc8 Qb7 9.Qxb7 Rxb7 10.Bf3 Rb8
11.Nd6 Kg7 (972.594.733) 667
These were the changes in the code:
if (UseExtendedHistory && value < HistoryValue / 2 && depth >= 8){
new_depth--;
if (value < HistoryValue / 3 && depth >= 12) new_depth--;
//[EdG 19-01-2008: Experimental extra reduction]
}
The next version also has double extra reductions, I'm just trying some things, Beta 4h already starts Extended History at 6 ply deep:
r1b2q1k/2Qn1p1p/1p1Rpp2/p6B/4P2P/6N1/P4PP1/6K1 w - -
Engine: Toga Checkov Beta 4h (256 MB)
by Thomas Gaksch and Fabien Letouzey
9/28 0:00 +1.70++ 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Rd2
4.Bxf7 Qc5 5.Qa8+ Kg7 6.Qg8+ Kh6
7.Kh2 Rxf2 8.Bxe6 (253.272)
10/32 0:00 +1.66 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rf8 5.Nf1 Rb8 6.Qa7 Rc8 7.a4 (434.703)
11/36 0:01 +1.63 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.h5 Rd7 7.Qa8+ Rd8
8.Qb7 Qd1+ 9.Nf1 (771.343) 720
12/42 0:06 +1.50 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Qb7 Rc8
4.Ne2 Rc2 5.Nf4 Kg8 6.a4 Rc4 7.Qxb6 Rxe4
8.Qc7 Rxa4 (4.518.123) 728
13/46 0:11 +1.50 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.a4 Rd8
4.Qb7 Rc8 5.Ne2 Rc4 6.Nf4 Kg8 7.Qxb6 Rxe4
8.Qc7 Rxa4 (8.656.789) 720
14/46 0:20 +1.46 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Rd8
4.Nh5+ Kg6 5.Qc7 Qd6 6.Qc3 Qe5 7.Qe3 Qa1+
8.Kh2 Qd4 9.Qg3 Qe5 10.f4 (15.115.960) 730
14/46 0:30 +1.54 1.Rd3 Rb8 2.Rxd7 Bxd7 3.Qxd7 Kg7
4.Bg4 Rd8 5.Nh5+ Kg6 6.Qc7 Qd6 7.Qc3 Qe5
8.Qe3 Qa1+ 9.Kh2 Qd4 10.Qf4 (21.666.947) 719
15/46 0:37 +1.55 1.Rd3 Rb8 2.Rxd7 Bxd7 3.Qxd7 Kg7
4.Bg4 Rd8 5.Nh5+ Kg6 6.Qc7 Qd6 7.Qc3 Qe5
8.Qe3 h6 9.f4 Qd4 10.f5+ Kh7 (26.635.633) 715
15/52 1:02 +1.94++ 1.e5 fxe5 2.Ne4 Kg7 3.Bf3 Kh8 4.Ng5 Rb8
5.Rxd7 Bxd7 6.Qxd7 Kg8 7.Bh5 Rd8
8.Bxf7+ Kh8 9.Qxe6 Qc5 (44.271.694) 713
16/68 5:05 +2.76++ 1.e5 fxe5 2.Ne4 Qg7 3.Ng5 e4 4.Nxf7+ Kg8
5.Rxe6 Nf8 6.Re8 Bb7 7.Re5 Ng6
8.Qxb7 Rf8 9.Rxe4 Qxf7 10.Qxb6 Qxa2
11.Bxg6 hxg6 12.Qxg6+ Kh8 (216.561.325) 708
17/68 7:45 +3.07 1.e5 fxe5 2.Ne4 Qg7 3.Ng5 e4 4.Nxf7+ Kg8
5.Qc6 Rb8 6.Rxe6 Nf8 7.Re7 Be6
8.Rxe6 Nxe6 9.Qxe6 Rf8 10.Ne5+ Kh8
11.Qd5 Qf6 12.Nf7+ Kg7 (332.253.147) 713
18/68 11:22 +3.19 1.e5 fxe5 2.Ne4 Qg7 3.Ng5 e4 4.Nxf7+ Kg8
5.Qc6 Rb8 6.Rxe6 Nf8 7.Re7 Be6
8.Rxe6 Nxe6 9.Qxe6 Rf8 10.Ne5+ Kh8
11.a4 Rf6 12.Nf7+ Kg8 13.Nh6+ Kh8
14.Qe8+ (486.380.280) 712
19/73 33:04 +3.87++ 1.e5 fxe5 2.Ne4 h6 3.Rd3 Kh7 4.Nd6 e4
5.Rg3 Ra7 6.Qc3 e5 7.Nxc8 Ra8 8.Nxb6 Nxb6
9.Qxe5 f6 10.Qc7+ Kh8 11.Qxb6 Rc8
12.Qd4 Rc1+ 13.Bd1 (1.451.883.464) 731
20/73 56:22 +3.34-- 1.e5 a4 2.Ne4 Ra5 3.Nxf6 Rc5 4.Rxd7 Rxc7
5.Rxc7 b5 6.Rxf7 Qxf7 7.Bxf7 b4
8.Bg8 b3 9.a3 h6 10.f4 b2 11.Bh7 Bb7
12.Kf2 Bd5 13.Bb1 Bc4 14.Ke3 (2.494.074.089) 737
21/73 70:50 +3.25 1.e5 a4 2.Ne4 Ra5 3.Nxf6 Rc5 4.Rxd7 Rxc7
5.Rxc7 b5 6.Rxf7 Qxf7 7.Bxf7 b4
8.Bg8 b3 9.a3 h6 10.f4 b2 11.Bh7 Bb7
12.Kf2 Kg7 13.Bd3 Ba6 14.Bc2 (3.132.744.387) 737
22/73 99:38 +3.36 1.e5 a4 2.Ne4 Ra5 3.Nxf6 Rc5 4.Rxd7 Rxc7
5.Rxc7 b5 6.Rxf7 Qxf7 7.Bxf7 b4
8.Bg8 b3 9.a3 h6 10.f4 b2 11.Bh7 Bb7
12.Kf2 Kg7 13.g4 Bd5 14.Nxd5 (4.407.831.475) 737
It is faster again especially compared with Beta 4 but what also is interesting is that it chooses 2... Ng7 for a while which is also what Rybka 2.3.2a has in mind. It switches to 2... h6 but the the variation becomes even more interesting; Toga does not accept the pawn but instead begins an advance with the a pawn with the sole purpose it seems of bringing it to c5 and then exchanging a majority of the heavy pieces. It seems an intricate maneuver, I wonder if it is correct and if heavy History reductions can lead to a selective search that is maybe not totally correct but sometimes seems like a plan and could maybe be backed up with some verification search somehow? Anyway I thought the 1... a5 answer seemed a nice find.
This was the change in code now, it is getting more complex:
if (UseHistory && depth >= HistoryDepth && node_type != NodePV) {
if (!in_check && played_nb >= HistoryMoveNb && new_depth < depth) {
ASSERT(best_value!=ValueNone);
ASSERT(played_nb>0);
ASSERT(sort->pos>0&&move==LIST_MOVE(sort->list,sort->pos-1));
value = sort->value; // history score
if (value < HistoryValue) {
ASSERT(value>=0&&value<16384);
ASSERT(move!=trans_move);
ASSERT(!move_is_tactical(move,board));
ASSERT(!move_is_check(move,board));
new_depth--;
reduced = true;
if (UseExtendedHistory && value < HistoryValue / 3 && depth >= 6){
new_depth--;
if (value < (HistoryValue * 2 / depth) && depth >= 12) new_depth--;
//[EdG 19-01-2008: Experimental extra reduction]
}
}
}
}
Eelco
)I don't have very interesting news, but just wanted to tell a little about the latest experiments with the code. I just started experimenting a bit with Toga's extended history pruning and I am probably introducing horrible horizon effects. But I found it interesting that with little changes you can get big differences in the PV and in times for new iterations. There goes little knowledge in the History reductions so many of the effects may well be random, but still they might have some function in shaping the tree.
If I find something that actually works I think I will call it Fischer Random Reductions
I am trying it out on Will Singleton's testposition, I don't get very fast times for finding 1. e5 because the hardware is much slower than what Johan has I think. I think I will follow a chronological order of new Beta versions:
The regular 3.1.2 Toga is fastest but it may be because it does not rate Rxd7 very high. Probably just because of the different settings. I let the computer calculate a fairly long time:
[d]r1b2q1k/2Qn1p1p/1p1Rpp2/p6B/4P2P/6N1/P4PP1/6K1 w - -
Engine: Toga II 3.1.2SE (256 MB, Athlon 2009 MHz)
by Thomas Gaksch and Fabien Letouzey
9/27 0:00 +1.39 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qc7 Qd6
4.Qxf7 Rf8 5.Qb7 Rb8 6.Qa7 Rg8 (171.458)
10/36 0:00 +1.44 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qxf7 Qxf7
4.Bxf7 e5 5.Nf5 Rd2 6.g4 b5 7.Kg2 b4
8.Bc4 a4 (447.814)
11/36 0:01 +1.50 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qxf7 Qxf7
4.Bxf7 Rd1+ 5.Kh2 Rd2 6.f3 e5 7.Nh5 b5
8.Kg3 Rd6 9.Bd5 f5 (834.403) 820
12/36 0:01 +1.47 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qxf7 Qxf7
4.Bxf7 Rd1+ 5.Kh2 e5 6.Nf5 Rd2 7.Kg3 b5
8.Kf3 Rc2 9.Bb3 Rd2 10.Bd5 (1.449.584) 812
13/36 0:03 +1.44 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qxf7 Qxf7
4.Bxf7 Rd1+ 5.Kh2 Rd6 6.Nh5 e5 7.Kg3 b5
8.Kf3 h6 9.Bd5 Kh7 10.g4 Kg6 (2.940.927) 823
14/44 0:06 +1.45 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qxf7 Qxf7
4.Bxf7 Rd1+ 5.Kh2 Rd6 6.Nh5 e5 7.Kg3 b5
8.Kf3 h6 9.Bd5 Kh7 10.Ng3 Kg6 11.Nf5 (5.604.835) 828
15/48 0:26 +1.34 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Kh2 Rd8
4.Qb7 Rd2 5.Qxb6 Rxa2 6.Bg4 a4
7.Nh5+ Kh8 8.Qd4 a3 9.Qxf6+ Kg8
10.Kg1 Rb2 11.Qg5+ Kh8 12.Qe5+ Kg8
13.Bd1 h6 14.Nf6+ (21.739.056) 831
15/48 0:53 +1.85++ 1.e5 fxe5 2.Ne4 Kg7 3.Rxd7 Bxd7
4.Qxd7 Rd8 5.Qc7 Rc8 6.Qxe5+ Kg8
7.Ng5 Rc1+ 8.Kh2 f6 9.Qxe6+ Kh8
10.Nf7+ Kg8 11.Qxb6 Rf1 12.Qxa5 (42.880.861) 800
16/62 2:54 +2.50++ 1.e5 fxe5 2.Ne4 h6 3.Rd3 Kh7 4.Nd6 e4
5.Rg3 Ra7 6.Qxa7 Qxd6 7.Qa8 Qf8
8.Qxe4+ Kh8 9.Rf3 f5 10.Qxe6 (138.515.322) 795
17/62 3:32 +2.65 1.e5 fxe5 2.Ne4 h6 3.Rd3 f6 4.Nd6 e4
5.Nf7+ Kh7 6.Rxd7 Bxd7 7.Qxd7 Qg7
8.Qxe6 Rf8 9.Qxe4+ Kg8 10.Nd6 Qc7
11.Qf4 Kh7 12.Bf3 (168.928.249) 795
18/62 4:45 +2.73 1.e5 fxe5 2.Ne4 h6 3.Rd3 f6 4.Nd6 e4
5.Nf7+ Kh7 6.Rxd7 Bxd7 7.Qxd7 Qg7
8.Qxe6 Rf8 9.Qxe4+ Kg8 10.Nd6 Qc7
11.Qf4 Qc2 12.Qxh6 Qb1+ 13.Kh2 Qxa2 (227.605.827) 796
19/68 14:05 +3.27++ 1.e5 fxe5 2.Ne4 h6 3.Rd3 f6 4.Nd6 e4
5.Rc3 f5 6.Nxc8 Nf6 7.Qb7 e3 8.Qxa8 exf2+
9.Kf1 Nxh5 10.Nxb6 Qxa8 11.Nxa8 Nf6
12.Rc8+ Kg7 13.Kxf2 (684.184.547) 808
20/68 25:46 +3.40 1.e5 fxe5 2.Ne4 h6 3.Rd3 Kh7 4.Rg3 f5
5.Nd6 Ra7 6.Qxa7 Qxd6 7.Rc3 Qf8
8.Qc7 Kg7 9.Qxc8 Qxc8 10.Rxc8 Kf6
11.Rg8 e4 12.Rg6+ Ke5 13.Rxh6 b5 (1.251.309.144) 809
21/74 49:45 +3.28 1.e5 fxe5 2.Ne4 Rb8 3.Rxd7 Bxd7
4.Qxe5+ f6 5.Nxf6 Rc8 6.Nxd7+ Qg7
7.Qxe6 b5 8.Ne5 Rf8 9.g3 b4 10.Kg2 a4
11.Nf7+ Kg8 12.Ng5+ Kh8 13.Qe4 Rb8
14.Bd1 (2.425.537.205) 812
22/78 156:14 +4.01++ 1.e5 fxe5 2.Ne4 h6 3.Rd3 f6 4.Nd6 a4
5.Bf7 Nc5 6.Rg3 Bd7 7.Rg8+ Qxg8
8.Bxg8 Rxg8 9.Qxb6 Nd3 10.Nf7+ Kh7
11.Qa7 Nf4 12.g3 Nh3+ 13.Kg2 Nf4+
14.Kh2 (7.592.749.433) 809
53 seconds is the fastest time and after 2½ hours the evaluation goes past +4
This is Toga Beta 4 with no changes in the History reductions, needs 3 minutes 45 seconds which is a bit slow:
r1b2q1k/2Qn1p1p/1p1Rpp2/p6B/4P2P/6N1/P4PP1/6K1 w - -
Engine: Toga Checkov Beta 4 (256 MB)
by Thomas Gaksch and Fabien Letouzey
9/28 0:00 +1.72++ 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Rd2
4.Bxf7 Qc5 5.Qa8+ Kg7 6.Qg8+ Kh6
7.Kh2 Rxf2 8.Bxe6 (225.745)
10/32 0:01 +1.70 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Qa4 Kg8 6.Be2 Kh8 7.Qa3 Qd2 (767.513) 700
11/32 0:01 +1.60 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Rd8
7.Bf5 Rd1+ 8.Kh2 Rd2 9.Nxf6 Rxf2
10.Nxh7+ Ke7 (1.298.913) 699
12/38 0:03 +1.67 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rc8 6.Nf1 Rd8 7.e5 fxe5
8.Ne3 Rd7 (2.325.144) 670
13/41 0:05 +1.73 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rc8 6.Nf1 Rd8 7.Ne3 Rc8
8.g4 Qe5 9.Qd7 Rc1+ 10.Nd1 (3.748.755) 688
14/41 0:08 +1.66 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.e5 fxe5 7.Ne4 Qd7
8.Qxd7 Rxd7 9.Bxe6 Rd1+ 10.Kh2 Kg7
11.Kg3 (5.561.364) 666
15/52 0:44 +1.56 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Rd8
7.Bf5 Rd1+ 8.Kh2 Ke7 9.a4 Rd4 10.Nf4 Rxa4
11.Nd5+ Kd6 12.Bxh7 b5 13.Nxf6 Ra2 (32.640.962) 729
15/52 0:58 +1.72 1.Rd1 Rb8 2.Rxd7 Bxd7 3.Qxd7 Kg7
4.Bg4 Qd8 5.Nh5+ Kf8 6.Nxf6 Rc8
7.Nxh7+ Kg7 8.Qxd8 Rxd8 9.Ng5 Rd2
10.Bh5 Rxa2 11.Bxf7 (42.511.906) 725
16/52 1:09 +1.79 1.Rd1 Rb8 2.Rxd7 Bxd7 3.Qxd7 Kg7
4.Bg4 Qd8 5.Nh5+ Kf8 6.Nxf6 Rc8
7.Nxh7+ Kg7 8.Qxd8 Rxd8 9.Ng5 Rc8
10.g3 Rc2 11.a4 Kg6 12.Kg2 (50.314.633) 724
17/52 1:31 +1.75 1.Rd1 Rb8 2.Rxd7 Bxd7 3.Qxd7 Kg7
4.Bg4 Rd8 5.Nh5+ Kg6 6.Qa7 Qd6 7.g3 Rd7
8.Qa8 Rd8 9.Qb7 Rd7 10.Qc8 Rd8
11.Qc4 (66.675.412) 726
17/56 3:45 +2.19++ 1.e5 fxe5 2.Ne4 Kg7 3.Rd1 a4 4.Nd6 Nf6
5.Bxf7 Kh8 6.Ne4 Kg7 7.Bh5+ Kg8
8.Nxf6+ Qxf6 9.Qxb6 Kf8 10.Rd8+ Kg7
11.Qc7+ Kh6 12.Bg4 Qxh4 (163.031.929) 721
18/76 22:41 +2.86++ 1.e5 fxe5 2.Ne4 h6 3.Rd1 f6 4.Nd6 f5
5.Qc6 Rb8 6.Nf7+ Kg7 7.Qxe6 Nf6
8.Qxe5 Rb7 9.Qg3+ Ng4 10.Nd6 Rd7
11.f3 Rxd6 12.Rxd6 (976.271.108) 716
The 4f version is the main testing version, there are no changes in Extended History Pruning yet but some changes in counting pawn attacks on the own King. Luckily it is a bit faster here, 1 minute 51 seconds. Rxd7 gets higher evals than in the Beta 4.
r1b2q1k/2Qn1p1p/1p1Rpp2/p6B/4P2P/6N1/P4PP1/6K1 w - -
Engine: Toga Checkov Beta 4f (256 MB)
by Thomas Gaksch and Fabien Letouzey
9/28 0:00 +1.70++ 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Rd2
4.Bxf7 Qc5 5.Qa8+ Kg7 6.Qg8+ Kh6
7.Kh2 Rxf2 8.Bxe6 (254.154)
10/32 0:00 +1.66 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rf8 5.Nf1 Rb8 6.Qa7 Rc8 7.a4 (437.458)
11/38 0:01 +1.63 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.h5 Rd7 7.Qa8+ Rd8
8.Qb7 Qd1+ 9.Nf1 (803.545) 720
12/42 0:01 +1.57 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.h5 Rd7 7.Qa8+ Rd8
8.Qb7 Qe5 9.Qxb6 Rd1+ 10.Nf1 (1.403.676) 706
13/46 0:19 +1.62 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.Nxb6 Rxf2 12.Nc4+ Kc5 (14.265.133) 732
14/46 0:24 +1.62 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.Nxb6 Rxf2 12.Nc4+ Kc5 (18.010.830) 736
15/48 0:35 +1.64 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.a4 Rxf2 12.Bxh7 Kc5
13.h5 Rd2 (26.100.890) 743
16/50 1:13 +1.66 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.a3 Rxf2 12.Nxb6 h5
13.Nc4+ Kc5 14.Nxa5 (54.345.814) 738
16/58 1:51 +2.04++ 1.e5 fxe5 2.Ne4 Kg8 3.Ng5 Ra7
4.Bxf7+ Kh8 5.Qc6 Nf6 6.Bxe6 Bxe6
7.Nxe6 Qf7 8.Rd8+ Ng8 9.Qxb6 Rd7
10.Rxd7 Qxd7 11.Nc5 Ne7 (81.117.239) 729
17/66 7:31 +2.69++ 1.e5 fxe5 2.Ne4 h6 3.Rd1 f6 4.Nd6 f5
5.a4 e4 6.Qc3+ Kg8 7.Bf7+ Kh7 8.Qc6 Rb8
9.Bxe6 Ne5 10.Qc7+ Qg7 11.Bxf5+ Bxf5
12.Qxb8 (326.650.527) 723
In Beta 4g I tried some changes in extended History Pruning, probably far too much, with double reductions, but this version is very quick in reaching 19 plies compared with default 3.1.2 SE
r1b2q1k/2Qn1p1p/1p1Rpp2/p6B/4P2P/6N1/P4PP1/6K1 w - -
Engine: Toga Checkov Beta 4g (256 MB)
by Thomas Gaksch and Fabien Letouzey
9/28 0:00 +1.70++ 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Rd2
4.Bxf7 Qc5 5.Qa8+ Kg7 6.Qg8+ Kh6
7.Kh2 Rxf2 8.Bxe6 (254.154)
10/32 0:00 +1.66 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rf8 5.Nf1 Rb8 6.Qa7 Rc8 7.a4 (437.458)
11/38 0:01 +1.63 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.h5 Rd7 7.Qa8+ Rd8
8.Qb7 Qd1+ 9.Nf1 (803.545) 600
12/42 0:02 +1.57 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.h5 Rd7 7.Qa8+ Rd8
8.Qb7 Qe5 9.Qxb6 Rd1+ 10.Nf1 (1.403.676) 595
13/46 0:19 +1.62 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.Nxb6 Rxf2 12.Nc4+ Kc5 (11.925.715) 603
14/46 0:25 +1.62 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.Nxb6 Rxf2 12.Nc4+ Kc5 (15.518.833) 606
15/46 0:40 +1.77 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.Nf4 Rd1+ 9.Kh2 Rd2
10.Nd5+ Kd6 11.Kg3 Rxa2 12.Bxh7 Rb2
13.h5 a4 14.Nxf6 (25.177.180) 615
16/48 1:26 +1.56 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Qd8
4.Nh5+ Kf8 5.Bxe6 Qxd7 6.Bxd7 Ke7
7.Bf5 Rd8 8.a4 Rd1+ 9.Kh2 Ra1
10.Bxh7 Rxa4 11.Nf4 Kd6 12.Nd5 b5
13.Nxf6 Ra2 (53.879.012) 624
16/50 2:46 +2.17++ 1.e5 fxe5 2.Ne4 Kg8 3.Ng5 Ra7
4.Bxf7+ Kh8 5.Qc6 Nf6 6.Bxe6 Ne8
7.Rd8 Kg7 (100.927.144) 607
17/70 14:52 +2.81++ 1.e5 fxe5 2.Ne4 h6 3.Rd1 f5 4.Nd6 e4
5.Qc3+ Qg7 6.Qc6 Rb8 7.Nxc8 Nc5
8.Rd8+ Kh7 9.Qd6 Rb7 10.Nxb6 Qa1+
11.Bd1 (556.644.438) 623
18/72 21:20 +3.30++ 1.e5 fxe5 2.Ne4 h6 3.Rd1 f6 4.Nd6 e4
5.Nxe4 Qg7 6.Nd6 Nf8 7.Qc6 Rb8
8.Nxc8 Qb7 9.Qxb7 Rxb7 10.Bf3 Rb8
11.Nd6 (843.819.035) 659
19/72 24:17 +3.29 1.e5 fxe5 2.Ne4 h6 3.Rd1 f6 4.Nd6 e4
5.Nxe4 Qg7 6.Nd6 Nf8 7.Qc6 Rb8
8.Nxc8 Qb7 9.Qxb7 Rxb7 10.Bf3 Rb8
11.Nd6 Kg7 (972.594.733) 667
These were the changes in the code:
if (UseExtendedHistory && value < HistoryValue / 2 && depth >= 8){
new_depth--;
if (value < HistoryValue / 3 && depth >= 12) new_depth--;
//[EdG 19-01-2008: Experimental extra reduction]
}
The next version also has double extra reductions, I'm just trying some things, Beta 4h already starts Extended History at 6 ply deep:
r1b2q1k/2Qn1p1p/1p1Rpp2/p6B/4P2P/6N1/P4PP1/6K1 w - -
Engine: Toga Checkov Beta 4h (256 MB)
by Thomas Gaksch and Fabien Letouzey
9/28 0:00 +1.70++ 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Rd2
4.Bxf7 Qc5 5.Qa8+ Kg7 6.Qg8+ Kh6
7.Kh2 Rxf2 8.Bxe6 (253.272)
10/32 0:00 +1.66 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rf8 5.Nf1 Rb8 6.Qa7 Rc8 7.a4 (434.703)
11/36 0:01 +1.63 1.Rxd7 Bxd7 2.Qxd7 Rd8 3.Qb7 Qd6
4.Bxf7 Rb8 5.Qa7 Rd8 6.h5 Rd7 7.Qa8+ Rd8
8.Qb7 Qd1+ 9.Nf1 (771.343) 720
12/42 0:06 +1.50 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Qb7 Rc8
4.Ne2 Rc2 5.Nf4 Kg8 6.a4 Rc4 7.Qxb6 Rxe4
8.Qc7 Rxa4 (4.518.123) 728
13/46 0:11 +1.50 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.a4 Rd8
4.Qb7 Rc8 5.Ne2 Rc4 6.Nf4 Kg8 7.Qxb6 Rxe4
8.Qc7 Rxa4 (8.656.789) 720
14/46 0:20 +1.46 1.Rxd7 Bxd7 2.Qxd7 Kg7 3.Bg4 Rd8
4.Nh5+ Kg6 5.Qc7 Qd6 6.Qc3 Qe5 7.Qe3 Qa1+
8.Kh2 Qd4 9.Qg3 Qe5 10.f4 (15.115.960) 730
14/46 0:30 +1.54 1.Rd3 Rb8 2.Rxd7 Bxd7 3.Qxd7 Kg7
4.Bg4 Rd8 5.Nh5+ Kg6 6.Qc7 Qd6 7.Qc3 Qe5
8.Qe3 Qa1+ 9.Kh2 Qd4 10.Qf4 (21.666.947) 719
15/46 0:37 +1.55 1.Rd3 Rb8 2.Rxd7 Bxd7 3.Qxd7 Kg7
4.Bg4 Rd8 5.Nh5+ Kg6 6.Qc7 Qd6 7.Qc3 Qe5
8.Qe3 h6 9.f4 Qd4 10.f5+ Kh7 (26.635.633) 715
15/52 1:02 +1.94++ 1.e5 fxe5 2.Ne4 Kg7 3.Bf3 Kh8 4.Ng5 Rb8
5.Rxd7 Bxd7 6.Qxd7 Kg8 7.Bh5 Rd8
8.Bxf7+ Kh8 9.Qxe6 Qc5 (44.271.694) 713
16/68 5:05 +2.76++ 1.e5 fxe5 2.Ne4 Qg7 3.Ng5 e4 4.Nxf7+ Kg8
5.Rxe6 Nf8 6.Re8 Bb7 7.Re5 Ng6
8.Qxb7 Rf8 9.Rxe4 Qxf7 10.Qxb6 Qxa2
11.Bxg6 hxg6 12.Qxg6+ Kh8 (216.561.325) 708
17/68 7:45 +3.07 1.e5 fxe5 2.Ne4 Qg7 3.Ng5 e4 4.Nxf7+ Kg8
5.Qc6 Rb8 6.Rxe6 Nf8 7.Re7 Be6
8.Rxe6 Nxe6 9.Qxe6 Rf8 10.Ne5+ Kh8
11.Qd5 Qf6 12.Nf7+ Kg7 (332.253.147) 713
18/68 11:22 +3.19 1.e5 fxe5 2.Ne4 Qg7 3.Ng5 e4 4.Nxf7+ Kg8
5.Qc6 Rb8 6.Rxe6 Nf8 7.Re7 Be6
8.Rxe6 Nxe6 9.Qxe6 Rf8 10.Ne5+ Kh8
11.a4 Rf6 12.Nf7+ Kg8 13.Nh6+ Kh8
14.Qe8+ (486.380.280) 712
19/73 33:04 +3.87++ 1.e5 fxe5 2.Ne4 h6 3.Rd3 Kh7 4.Nd6 e4
5.Rg3 Ra7 6.Qc3 e5 7.Nxc8 Ra8 8.Nxb6 Nxb6
9.Qxe5 f6 10.Qc7+ Kh8 11.Qxb6 Rc8
12.Qd4 Rc1+ 13.Bd1 (1.451.883.464) 731
20/73 56:22 +3.34-- 1.e5 a4 2.Ne4 Ra5 3.Nxf6 Rc5 4.Rxd7 Rxc7
5.Rxc7 b5 6.Rxf7 Qxf7 7.Bxf7 b4
8.Bg8 b3 9.a3 h6 10.f4 b2 11.Bh7 Bb7
12.Kf2 Bd5 13.Bb1 Bc4 14.Ke3 (2.494.074.089) 737
21/73 70:50 +3.25 1.e5 a4 2.Ne4 Ra5 3.Nxf6 Rc5 4.Rxd7 Rxc7
5.Rxc7 b5 6.Rxf7 Qxf7 7.Bxf7 b4
8.Bg8 b3 9.a3 h6 10.f4 b2 11.Bh7 Bb7
12.Kf2 Kg7 13.Bd3 Ba6 14.Bc2 (3.132.744.387) 737
22/73 99:38 +3.36 1.e5 a4 2.Ne4 Ra5 3.Nxf6 Rc5 4.Rxd7 Rxc7
5.Rxc7 b5 6.Rxf7 Qxf7 7.Bxf7 b4
8.Bg8 b3 9.a3 h6 10.f4 b2 11.Bh7 Bb7
12.Kf2 Kg7 13.g4 Bd5 14.Nxd5 (4.407.831.475) 737
It is faster again especially compared with Beta 4 but what also is interesting is that it chooses 2... Ng7 for a while which is also what Rybka 2.3.2a has in mind. It switches to 2... h6 but the the variation becomes even more interesting; Toga does not accept the pawn but instead begins an advance with the a pawn with the sole purpose it seems of bringing it to c5 and then exchanging a majority of the heavy pieces. It seems an intricate maneuver, I wonder if it is correct and if heavy History reductions can lead to a selective search that is maybe not totally correct but sometimes seems like a plan and could maybe be backed up with some verification search somehow? Anyway I thought the 1... a5 answer seemed a nice find.
This was the change in code now, it is getting more complex:
if (UseHistory && depth >= HistoryDepth && node_type != NodePV) {
if (!in_check && played_nb >= HistoryMoveNb && new_depth < depth) {
ASSERT(best_value!=ValueNone);
ASSERT(played_nb>0);
ASSERT(sort->pos>0&&move==LIST_MOVE(sort->list,sort->pos-1));
value = sort->value; // history score
if (value < HistoryValue) {
ASSERT(value>=0&&value<16384);
ASSERT(move!=trans_move);
ASSERT(!move_is_tactical(move,board));
ASSERT(!move_is_check(move,board));
new_depth--;
reduced = true;
if (UseExtendedHistory && value < HistoryValue / 3 && depth >= 6){
new_depth--;
if (value < (HistoryValue * 2 / depth) && depth >= 12) new_depth--;
//[EdG 19-01-2008: Experimental extra reduction]
}
}
}
}
Eelco