OpenTal - almost new engine

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Ovyron
Posts: 2531
Joined: Tue Jul 03, 2007 2:30 am

Re: OpenTal - almost new engine

Post by Ovyron » Wed Dec 20, 2017 10:56 am

BrendanJNorman wrote:Match ended 6-0 for Thinker 5.3b Inert. :cry:
Can you publish the games? We're not after results here, but trying to produce beautiful games :wink: - Wanna see if Thinker ever won a game while low on Material.

Another idea is to play time-handicap games, just shorten the times that other engines get, then you can see Open Tal defeating the Komodos, Stockfishes and Houdinis at the top :)

I still remember 7 years ago I used to play ProDeo 1.6 Q3 Tactical Engine at Depth 10 against Rybka 4 at depth 5 to equalize their strengths and produce some nice games, OpenTal can play against any opponent with such handicaps.

BrendanJNorman
Posts: 1416
Joined: Sun Feb 07, 2016 11:43 pm
Full name: Brendan J Norman

Re: OpenTal - almost new engine

Post by BrendanJNorman » Wed Dec 20, 2017 2:50 pm

Ovyron wrote:
BrendanJNorman wrote:Match ended 6-0 for Thinker 5.3b Inert. :cry:
Can you publish the games? We're not after results here, but trying to produce beautiful games :wink: - Wanna see if Thinker ever won a game while low on Material.

Another idea is to play time-handicap games, just shorten the times that other engines get, then you can see Open Tal defeating the Komodos, Stockfishes and Houdinis at the top :)

I still remember 7 years ago I used to play ProDeo 1.6 Q3 Tactical Engine at Depth 10 against Rybka 4 at depth 5 to equalize their strengths and produce some nice games, OpenTal can play against any opponent with such handicaps.
I have fantastic news! :)

Although for the most part, OpenTal is too weak to tackle 2800 level engines, my initial test against Thinker 5.3b Inert was misleading.

Thinker is not only the stronger engine (and the Inert version especially has tested strongest against engines!), but had a much better opening book too.

So in those initial test games, OpenTal was getting poor positions out of the opening and just losing like an idiot.

When I switched Thinker to a repertoire book (as OpenTal is using) things became more competitive immediately.

Thinker was using my Boris Spassky opening repertoire book (which is obviously comparable to Tal's given that they are from the same period) and although stronger still, the games were more lively and OpenTal actually won a couple.

Even some of his losses were typical Tal beauty (which humans would likely not have defended successfully).

First I'll show you a game from the REAL Tal from 1958 just to give some background.

[pgn][Event "WchT U26 fin-A 05th"]
[Site "Varna BUL"]
[Date "1958.07.??"]
[EventDate "?"]
[Round "2"]
[Result "0-1"]
[White "Milko Bobotsov"]
[Black "Mikhail Tal"]
[ECO "E81"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "60"]

1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.f3 O-O 6.Nge2 c5 7.Be3
Nbd7 8.Qd2 a6 9.O-O-O Qa5 10.Kb1 b5 11.Nd5 Nxd5 12.Qxa5 Nxe3
13.Rc1 Nxc4 14.Rxc4 bxc4 15.Nc1 Rb8 16.Bxc4 Nb6 17.Bb3 Bxd4
18.Qd2 Bg7 19.Ne2 c4 20.Bc2 c3 21.Qd3 cxb2 22.Nd4 Bd7 23.Rd1
Rfc8 24.Bb3 Na4 25.Bxa4 Bxa4 26.Nb3 Rc3 27.Qxa6 Bxb3 28.axb3
Rbc8 29.Qa3 Rc1+ 30.Rxc1 Rxc1 0-1[/pgn]

How close to this Tal do you think we could get?

Look at this vs Thinker 5.3b Inert...

[pgn][Event "OpenTal vs Thinker 5.3b Inert 3 2"]
[Site "BRENDANNORMD8A2"]
[Date "2017.12.20"]
[Round "8"]
[White "Thinker 5.3b Inert"]
[Black "OpenTal 1.0"]
[Result "1-0"]
[ECO "B36"]
[WhiteElo "2844"]
[BlackElo "2200"]
[PlyCount "99"]
[EventDate "2017.??.??"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nc6 5. c4 Nf6 6. Nc3 g6 7. Be2 Nxd4 8.
Qxd4 Bg7 9. Be3 O-O 10. Qd2 Be6 11. f3 Qa5 12. Rb1 Rac8 13. b3 h5 14. O-O h4
15. Rfc1 h3 16. Nd5 Nxd5 17. Qxa5 Nxe3 18. Qxa7 Nxg2 19. Qxb7 Bd4+ 20. Kh1 Nf4
21. Bf1 Rfe8 22. Qb4 Bd7 23. a4 f5 24. Qd2 e5 25. Bd3 Ra8 26. Bc2 Kg7 27. Rg1
Kf7 28. Rg3 Nh5 29. Rg5 f4 30. a5 Nf6 31. b4 Nh7 32. Rgg1 Nf6 33. b5 Reb8 34.
a6 Be6 35. Rgd1 g5 36. Qa5 Bb6 37. Qa4 Bc5 38. b6 g4 39. b7 g3 40. hxg3 fxg3
41. bxa8=Q Rxa8 42. Rb7+ Kg6 43. Re7 Bf7 44. Qb3 Ra7 45. Rxa7 Bxa7 46. Qb7 Bc5
47. a7 g2+ 48. Kh2 Bxc4 49. Rg1 Bxa7 50. Qxa7 1-0[/pgn]

Do you think a human, especially Bobotsov would have defended as Thinker did? I very much doubt it. :)

Now look at one of OpenTal's wins!

This game seems impossible against a 2840 beast, what amazing tactical play.

[pgn][Event "OpenTal vs Thinker 5.3b Inert 3 2"]
[Site "BRENDANNORMD8A2"]
[Date "2017.12.20"]
[Round "9"]
[White "OpenTal 1.0"]
[Black "Thinker 5.3b Inert"]
[Result "1-0"]
[ECO "B84"]
[WhiteElo "2200"]
[BlackElo "2844"]
[PlyCount "82"]
[EventDate "2017.??.??"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 e6 6. Be3 a6 7. f4 Qc7 8. Be2
Be7 9. O-O O-O 10. Qe1 b5 11. a3 Bb7 12. Bf3 Nc6 13. Nb3 Nd7 14. Qg3 Bf6 15.
Rad1 Rad8 16. Rd3 Nb6 17. Nd2 b4 18. Ne2 a5 19. axb4 axb4 20. b3 Rb8 21. Rc1
Rfd8 22. Bh5 Ba6 23. c4 bxc3 24. Rdxc3 Bxc3 25. Nxc3 Nd7 26. Nd5 Qa5 27. b4
Nxb4 28. Nb3 Qa4 29. Bd4 e5 30. Ne7+ Kf8 31. fxe5 g6 32. e6 Kxe7 33. Qh4+ Ke8
34. Bxg6 hxg6 35. Rc7 Nd5 36. exd5 Qxd4+ 37. Nxd4 Rb1+ 38. Kf2 fxe6 39. dxe6
Rf1+ 40. Kg3 Bb5 41. exd7+ Rxd7 1-0[/pgn]

Jaw dropping stuff.

Pawel, please make no changes (besides the opening book and Fritz GUI fixes) because this is truly Tal.

Anything stronger is just another aggressive engine.

Canoike
Posts: 95
Joined: Tue Jan 17, 2012 7:08 pm

Re: OpenTal - almost new engine

Post by Canoike » Wed Dec 20, 2017 8:10 pm

The source code is provided but there is no makefile for LInux. It is too sad to see such a good chess program running only in a bloated OS.

PK
Posts: 820
Joined: Mon Jan 15, 2007 10:23 am
Location: Warsza
Contact:

Re: OpenTal - almost new engine

Post by PK » Wed Dec 20, 2017 10:07 pm

You can modify this makefile: https://github.com/nescitus/Rodent_III/ ... s/Makefile

You don't need personality paths of course (lines 19-20).

I'd be happy to provide a third-party makefile, but unfortunately cannot prepare it by myself, as I don't use Linux. Call it an allergy if you want. A couple of years ago friends convinced me to install Linux as a second operating system on my laptop, then offered to run this perfectly safe process, then managed to blow up a big and important part of file system.

User avatar
Ovyron
Posts: 2531
Joined: Tue Jul 03, 2007 2:30 am

Re: OpenTal - almost new engine

Post by Ovyron » Thu Dec 21, 2017 1:25 am

BrendanJNorman wrote:When I switched Thinker to a repertoire book (as OpenTal is using) things became more competitive immediately.
Ah ha! So this is workable. Really nice games, indeed!

I'm thinking about going back to game testing and play some games myself with OpenTal, I have the urge to do something (years ago I had the idea to measure the style of an engine by awarding points for sacrifices and material imbalances - but this requires setting opponents to the same strength first...)

royb
Posts: 518
Joined: Wed Mar 08, 2006 11:53 pm

Re: OpenTal - almost new engine

Post by royb » Thu Dec 21, 2017 3:33 am

Canoike wrote:The source code is provided but there is no makefile for LInux. It is too sad to see such a good chess program running only in a bloated OS.
I was able to compile it on Linux Mint 18.2 with no issues using the makefile provided.

carldaman
Posts: 1708
Joined: Sat Jun 02, 2012 12:13 am

Re: OpenTal - almost new engine

Post by carldaman » Fri Dec 22, 2017 5:05 am

BrendanJNorman wrote:
Ovyron wrote:
BrendanJNorman wrote:Match ended 6-0 for Thinker 5.3b Inert. :cry:
Can you publish the games? We're not after results here, but trying to produce beautiful games :wink: - Wanna see if Thinker ever won a game while low on Material.

Another idea is to play time-handicap games, just shorten the times that other engines get, then you can see Open Tal defeating the Komodos, Stockfishes and Houdinis at the top :)

I still remember 7 years ago I used to play ProDeo 1.6 Q3 Tactical Engine at Depth 10 against Rybka 4 at depth 5 to equalize their strengths and produce some nice games, OpenTal can play against any opponent with such handicaps.
I have fantastic news! :)

Although for the most part, OpenTal is too weak to tackle 2800 level engines, my initial test against Thinker 5.3b Inert was misleading.

Thinker is not only the stronger engine (and the Inert version especially has tested strongest against engines!), but had a much better opening book too.

So in those initial test games, OpenTal was getting poor positions out of the opening and just losing like an idiot.

When I switched Thinker to a repertoire book (as OpenTal is using) things became more competitive immediately.

Thinker was using my Boris Spassky opening repertoire book (which is obviously comparable to Tal's given that they are from the same period) and although stronger still, the games were more lively and OpenTal actually won a couple.

Even some of his losses were typical Tal beauty (which humans would likely not have defended successfully).

First I'll show you a game from the REAL Tal from 1958 just to give some background.

[pgn][Event "WchT U26 fin-A 05th"]
[Site "Varna BUL"]
[Date "1958.07.??"]
[EventDate "?"]
[Round "2"]
[Result "0-1"]
[White "Milko Bobotsov"]
[Black "Mikhail Tal"]
[ECO "E81"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "60"]

1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.f3 O-O 6.Nge2 c5 7.Be3
Nbd7 8.Qd2 a6 9.O-O-O Qa5 10.Kb1 b5 11.Nd5 Nxd5 12.Qxa5 Nxe3
13.Rc1 Nxc4 14.Rxc4 bxc4 15.Nc1 Rb8 16.Bxc4 Nb6 17.Bb3 Bxd4
18.Qd2 Bg7 19.Ne2 c4 20.Bc2 c3 21.Qd3 cxb2 22.Nd4 Bd7 23.Rd1
Rfc8 24.Bb3 Na4 25.Bxa4 Bxa4 26.Nb3 Rc3 27.Qxa6 Bxb3 28.axb3
Rbc8 29.Qa3 Rc1+ 30.Rxc1 Rxc1 0-1[/pgn]

How close to this Tal do you think we could get?

Look at this vs Thinker 5.3b Inert...

[pgn][Event "OpenTal vs Thinker 5.3b Inert 3 2"]
[Site "BRENDANNORMD8A2"]
[Date "2017.12.20"]
[Round "8"]
[White "Thinker 5.3b Inert"]
[Black "OpenTal 1.0"]
[Result "1-0"]
[ECO "B36"]
[WhiteElo "2844"]
[BlackElo "2200"]
[PlyCount "99"]
[EventDate "2017.??.??"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nc6 5. c4 Nf6 6. Nc3 g6 7. Be2 Nxd4 8.
Qxd4 Bg7 9. Be3 O-O 10. Qd2 Be6 11. f3 Qa5 12. Rb1 Rac8 13. b3 h5 14. O-O h4
15. Rfc1 h3 16. Nd5 Nxd5 17. Qxa5 Nxe3 18. Qxa7 Nxg2 19. Qxb7 Bd4+ 20. Kh1 Nf4
21. Bf1 Rfe8 22. Qb4 Bd7 23. a4 f5 24. Qd2 e5 25. Bd3 Ra8 26. Bc2 Kg7 27. Rg1
Kf7 28. Rg3 Nh5 29. Rg5 f4 30. a5 Nf6 31. b4 Nh7 32. Rgg1 Nf6 33. b5 Reb8 34.
a6 Be6 35. Rgd1 g5 36. Qa5 Bb6 37. Qa4 Bc5 38. b6 g4 39. b7 g3 40. hxg3 fxg3
41. bxa8=Q Rxa8 42. Rb7+ Kg6 43. Re7 Bf7 44. Qb3 Ra7 45. Rxa7 Bxa7 46. Qb7 Bc5
47. a7 g2+ 48. Kh2 Bxc4 49. Rg1 Bxa7 50. Qxa7 1-0[/pgn]

Do you think a human, especially Bobotsov would have defended as Thinker did? I very much doubt it. :)

Now look at one of OpenTal's wins!

This game seems impossible against a 2840 beast, what amazing tactical play.

[pgn][Event "OpenTal vs Thinker 5.3b Inert 3 2"]
[Site "BRENDANNORMD8A2"]
[Date "2017.12.20"]
[Round "9"]
[White "OpenTal 1.0"]
[Black "Thinker 5.3b Inert"]
[Result "1-0"]
[ECO "B84"]
[WhiteElo "2200"]
[BlackElo "2844"]
[PlyCount "82"]
[EventDate "2017.??.??"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 e6 6. Be3 a6 7. f4 Qc7 8. Be2
Be7 9. O-O O-O 10. Qe1 b5 11. a3 Bb7 12. Bf3 Nc6 13. Nb3 Nd7 14. Qg3 Bf6 15.
Rad1 Rad8 16. Rd3 Nb6 17. Nd2 b4 18. Ne2 a5 19. axb4 axb4 20. b3 Rb8 21. Rc1
Rfd8 22. Bh5 Ba6 23. c4 bxc3 24. Rdxc3 Bxc3 25. Nxc3 Nd7 26. Nd5 Qa5 27. b4
Nxb4 28. Nb3 Qa4 29. Bd4 e5 30. Ne7+ Kf8 31. fxe5 g6 32. e6 Kxe7 33. Qh4+ Ke8
34. Bxg6 hxg6 35. Rc7 Nd5 36. exd5 Qxd4+ 37. Nxd4 Rb1+ 38. Kf2 fxe6 39. dxe6
Rf1+ 40. Kg3 Bb5 41. exd7+ Rxd7 1-0[/pgn]

Jaw dropping stuff.

Pawel, please make no changes (besides the opening book and Fritz GUI fixes) because this is truly Tal.

Anything stronger is just another aggressive engine.
Actually, since OpenTal is really an anti-human player, a better idea might be to match up OpenTal against one of the Thinker Active personalities, since they were also meant to be used against humans, primarily.

Great games, btw. :D

CL

carldaman
Posts: 1708
Joined: Sat Jun 02, 2012 12:13 am

sample game: OpenTal vs Thinker Active

Post by carldaman » Sat Dec 23, 2017 11:58 pm

I've only had time to run a 4-game match between OpenTal 1.0 and Thinker 5.4D active. Tal won 3-1. :) Here's the best game with some annotations:

[pgn]
[Event "Tourn08o-10m8s"]
[Date "2017.12.22"]
[Round "2"]
[White "Thinker5.4D_64-bit Active"]
[Black "OpenTal 1.0"]
[Result "0-1"]
[ECO "A45"]
[WhiteElo "2650"]
[BlackElo "2650"]
[PlyCount "114"]
[TimeControl "600+8"]

1. d4 Nf6 2. e3 g6 3. Bd3 Bg7 4. e4 $5 O-O {[%eval 89,21] [%emt 0:00:23]} 5.
Nf3 d5 $5 {[%eval 85,21] [%emt 0:00:23]} (5... d6) 6. exd5 (6. e5 Ne4 $5 $13)
6... c5 $5 $146 {[%eval 97,20] [%emt 0:00:22]} (6... --) 7. dxc6 Nxc6 $44 {
[%eval 115,21] [%emt 0:00:22]} 8. Be2 Bf5 $1 $36 {[%eval 135,18] [%emt 0:00:21]
} 9. O-O Qc7 $5 {[%eval 105,18] [%emt 0:00:21]} (9... Qe8 $5) (9... Re8) (9...
Rc8) 10. Na3 Rfd8 {[%eval 112,19] [%emt 0:00:21]} (10... Rad8 $36) 11. Qe1 a6 {
[%eval 139,18] [%emt 0:00:20]} 12. c3 b5 {[%eval 110,20] [%emt 0:00:20]} 13.
Nh4 Be4 {[%eval 125,20] [%emt 0:00:20]} 14. f3 Bd5 {[%eval 134,19] [%emt 0:00:
19]} 15. Qf2 b4 {[%eval 164,19] [%emt 0:00:19]} 16. Nc2 bxc3 {[%eval 166,20]
[%emt 0:00:19]} 17. bxc3 Na5 $5 {[%eval 154,19] [%emt 0:00:19]} (17... Rab8 $36
) 18. Bd2 Nc4 {[%eval 164,20] [%emt 0:00:18]} 19. Bxc4 Bxc4 {[%eval 193,19]
[%emt 0:00:18] after this trade light square weaknesses are bound to
haunt White for the rest of the game} 20. Rfc1 Rab8 {[%eval 191,20] [%emt 0:00:
18]} 21. Ne3 Rb2 {[%eval 195,21] [%emt 0:00:18]} 22. a3 Rdb8 {[%eval 234,20]
[%emt 0:00:17]} 23. Nd1 R2b3 {[%eval 268,21] [%emt 0:00:17]} 24. a4 a5 {[%eval
274,20] [%emt 0:00:17]} 25. g3 Nd5 {[%eval 277,18] [%emt 0:00:17]} 26. Ng2 Qb7
{[%eval 242,18] [%emt 0:00:16]} 27. Nge3 Nxe3 {[%eval 236,21] [%emt 0:00:16]}
28. Nxe3 Bd3 {[%eval 230,22] [%emt 0:00:16]} 29. Re1 Rc8 {[%eval 218,19] [%emt
0:00:16]} 30. Red1 Rb2 {[%eval 237,19] [%emt 0:00:15]} 31. Rdc1 e5 {[%eval 239,
17] [%emt 0:00:15]} 32. Nd1 Rb3 {[%eval 247,20] [%emt 0:00:15]} 33. Be3 e4 {
[%eval 257,19] [%emt 0:00:15]} 34. fxe4 Bxe4 {[%eval 286,19] [%emt 0:00:15]}
35. Qf1 Bf5 {[%eval 289,17] [%emt 0:00:15]} 36. Bd2 h5 {[%eval 282,19] [%emt 0:
00:14]} 37. Ne3 Be4 {[%eval 236,19] [%emt 0:00:14]} 38. Qe1 h4 $1 {[%eval 420,
18] [%emt 0:00:14]} 39. gxh4 f5 {[%eval 411,18] [%emt 0:00:14]} 40. Nd1 Re8 {
[%eval 394,18] [%emt 0:00:14]} 41. Bg5 Bc6 {[%eval 386,17] [%emt 0:00:14]} 42.
Qf1 Bg2 $3 {[%eval 364,20] [%emt 0:00:13] the B can't really be taken :)} 43.
Qf2 (43. Qxg2 Re1+ 44. Kf2 Re2+ $1) 43... Bf3 {[%eval 343,18] [%emt 0:00:13]}
44. Ne3 f4 $1 {[%eval 390,21] [%emt 0:00:13]} 45. Bxf4 Rf8 {[%eval 284,19]
[%emt 0:00:13]} 46. Bg5 Be4 {[%eval 454,21] [%emt 0:00:13]} 47. Qa2 Kh7 {
[%eval 428,21] [%emt 0:00:13]} 48. Qe2 Rb2 {[%eval 411,22] [%emt 0:00:13]} 49.
Qg4 Rff2 {[%eval 546,18] [%emt 0:00:12]} 50. Qg3 Rf3 {[%eval 688,20] [%emt 0:
00:12]} 51. Qg4 Bf8 {[%eval 596,18] [%emt 0:00:12]} 52. Bf4 Rxf4 $3 {[%eval
1056,20] [%emt 0:00:12] just deadly} 53. Qxf4 Bh6 $1 $19 {[%eval 1117,24]
[%emt 0:00:12] now White can't avert heavy loss of material, since it would
otherwise get mated} 54. Rf1 (54. Qg3 Bxe3+ 55. Qxe3 Rg2+ 56. Kf1 Qf7+ 57. Ke1
Bd3) 54... Bxf4 {[%eval 1222,24] [%emt 0:00:12]} 55. Rxf4 Qd7 {[%eval 1301,26]
[%emt 0:00:12]} 56. Rf2 Rxf2 {[%eval 1493,25] [%emt 0:00:12]} 57. Kxf2 Qh3 {
[%eval 1765,27] [%emt 0:00:12] User Adjudication} 0-1
[/pgn]

Not the shortest win, but saccing three pawns, then the exchange, plus a sac that had to be declined, all amount to a great win. :)

carldaman
Posts: 1708
Joined: Sat Jun 02, 2012 12:13 am

Re: sample game: OpenTal vs Thinker Active

Post by carldaman » Sun Dec 24, 2017 12:31 am

If you liked that last game you might also enjoy the following effort. Not to take anything away from the brilliant OpenTal, which truly resembles the real Tal very closely. :) I fully appreciate how important an achievement is to have an engine playing almost exactly like the great Tal.

My goal has been somewhat similar, but I've wanted to get to a point where the machine not only plays like Morphy, Anderssen, Tal, Nezhmetdinov or Shirov, but also can rise above them in strength while retaining the same style, and being (why not?) even stronger than any human alive, past or present.

The results have been amazing, as far as I can tell. I only wish I had more time and better health to devote to such a project. The "entity" currently goes by the slightly humorous name Zero-the-Hero, to poke some fun at the recent hoopla involving AZ beating Stockfish. At this point, I'm not yet at liberty to reveal more than just games and analysis.

Here's its best win vs OT1:


[pgn]
[Event "Tourn08o-10m8s"]
[Date "2017.12.23"]
[Round "3"]
[White "Zero-the-Hero"]
[Black "OpenTal 1.0"]
[Result "1-0"]
[ECO "D15"]
[WhiteElo ""]
[BlackElo "2650"]
[PlyCount "95"]
[TimeControl "600+8"]

1. d4 {[%eval 113,25] [%emt 0:00:49]} d5 2. c4 {[%eval 118,25] [%emt 0:00:17]}
dxc4 3. Nf3 {[%eval 151,26] [%emt 0:00:35]} Nf6 4. Nc3 {[%eval 136,25] [%emt 0:
00:38]} c6 5. e4 $5 {[%eval 172,22] [%emt 0:00:03]} b5 6. Be2 $5 {[%eval 142,
26] [%emt 0:00:19]} e6 {[%eval -11,22] [%emt 0:00:23]} 7. O-O {[%eval 140,25]
[%emt 0:00:34]} Be7 {[%eval 0,23] [%emt 0:00:23]} 8. Qc2 $5 {[%eval 153,24]
[%emt 0:00:23]} (8. a4 b4 9. e5) 8... O-O {[%eval 38,20] [%emt 0:00:23] Black
does not know what dangers await...} (8... Bb7) 9. e5 $5 {[%eval 176,24]
[%emt 0:00:11]} Nd5 {[%eval 46,21] [%emt 0:00:22]} (9... Ne8 10. b3 $36) 10.
Ne4 $5 $146 {[%eval 168,25] [%emt 0:00:12] What's fascinating is that this
move is a novelty, while in a previous game between two corr heavy weights on
ICCF, b3 was played, which petered out to a draw} (10. b3 cxb3 11. axb3 Nxc3
12. Qxc3 Bb7 13. Qc2 Na6 14. Qe4 Nb4 15. Qg4 Kh8 16. Bd2 a5 17. Rfd1 Qb6 18.
Rac1 Na2 19. Rc2 Nb4 20. Rc3 Nd5 21. Rd3 Qc7 22. Nh4 f5 23. exf6 Nxf6 24. Qh3
Kg8 25. Nf3 Bd6 26. Rc3 Qe7 27. Bd3 Bb4 28. Rcc1 Bxd2 29. Rxd2 g6 30. Ng5 Bc8
31. Rxc6 Bd7 32. Rc5 a4 33. bxa4 Rxa4 34. Bf1 Nd5 35. Nf3 Rxd4 36. Nxd4 Qxc5
37. Nxb5 Qc1 38. Rxd5 Qe1 39. f3 Qe3+ 40. Kh1 Qe1 41. Kg1 Qe3+ {1/2-1/2 (41)
Dhanish,P (2600)-Neto,H (2590) ICCF 2014}) 10... Nd7 {[%eval 48,20] [%emt 0:00:
22]} (10... h6 11. Ng3 $5 (11. Bd2 $5 Bb4 $5 12. Bxb4 Nxb4 13. Qd2 $1 Nd5 14.
g3) 11... Nd7 12. Qe4 c5 13. a4 cxd4 (13... Bb7 14. Qg4) 14. Qxd4 f5 $5 15.
exf6 Bc5 16. Qg4 (16. Qh4) 16... N7xf6 17. Qg6 Ne7) 11. Neg5 $40 {[%eval 267,
25] [%emt 0:00:23]} g6 {[%eval 37,20] [%emt 0:00:21]} 12. a4 $36 {[%eval 226,
26] 95} Bb7 {[%eval 57,19] [%emt 0:00:21]} 13. Qe4 $1 {[%eval 175,26] 73} c3 $2
{[%eval 65,19] [%emt 0:00:21]} (13... Kg7 14. Qg4 h6 15. Ne4 Rh8 (15... h5 16.
Qg3) 16. Bd2 $1 (16. h4 Rh7 (16... Kf8 $5 17. Qg3 $1) 17. Qg3 a6 18. Nd6 Bxd6
19. exd6 N7f6 20. Bd1 $5 Nh5 (20... Kg8 21. Bc2) 21. Qh2 Kg8 22. Re1 Ng7 23.
Bc2 Nf5 24. g4 Nb4 25. Bb1 Nxd6 26. Ne5 $44) 16... Kg8 17. h4 $1 Rh7 18. Rfb1
$5 c5 (18... a5 19. b3 c3 20. Nxc3 Nxc3 21. Bxc3 b4 22. Bb2 Rc8 23. Rd1 h5 (
23... Kh8 24. Rac1) 24. Qg3 $16) 19. axb5 cxd4 20. Bxc4 Qb8 21. Nd6 $16 Bxd6 (
21... Ne3 $5 22. Qxd4 $1 Nxc4 23. Nxc4 $16) 22. exd6 Qxd6 23. Qxd4 e5 24. Qd3
$18 Nc5 (24... -- 25. Bc3 $140 $18 Nc5 26. Bxe5) 25. Qa3 Nb6 26. Bf1 Bxf3 27.
Bb4 $1 Bd5 28. Bxc5 $18 Qe6 29. Qe3 $1) 14. Qh4 {[%eval 376,23] [%emt 0:00:07]
Zero already sees itself as winning} h5 {[%eval 65,21] [%emt 0:00:20]} 15. Qg3
$1 {[%eval 417,24] [%emt 0:00:13]} b4 {[%eval 53,20] [%emt 0:00:20]} 16. Nxe6
$1 {[%eval 465,24] [%emt 0:00:19] let the fireworks begin} Qe8 {[%eval 57,20]
[%emt 0:00:20]} 17. Nf4 $3 {[%eval 582,24] [%emt 0:00:05] not taking the Rf8}
c5 {[%eval 71,18] [%emt 0:00:19]} 18. Nxh5 {[%eval 691,20] [%emt 0:00:05]} Rc8
{[%eval 76,18] [%emt 0:00:19]} 19. Bd3 {[%eval 736,24] [%emt 0:00:14]} Rc6 {
[%eval 48,17] [%emt 0:00:19]} 20. Re1 {[%eval 780,24] [%emt 0:00:36]} cxb2 {
[%eval 0,19] [%emt 0:00:19]} 21. Bxb2 {[%eval 789,24] [%emt 0:00:03]} c4 {
[%eval 0,21] [%emt 0:00:18]} 22. Bf5 $1 {[%eval 900,22] [%emt 0:00:05] the
board is catching on fire!} c3 {[%eval 0,19] [%emt 0:00:18]} 23. Bc1 {[%eval
966,22] [%emt 0:00:12]} N7b6 {[%eval 0,20] [%emt 0:00:18]} 24. e6 $1 {[%eval
1055,22] [%emt 0:00:31]} Nc4 {[%eval 0,20] [%emt 0:00:18]} 25. Bh6 {[%eval
1012,22] [%emt 0:00:19]} b3 {[%eval 0,16] [%emt 0:00:17] OpenTal is still
assessing this as 0.00 (dynamically balanced?), but White (Zero) can see past
the zeroes, lol} 26. Rab1 {[%eval 998,23] [%emt 0:00:49]} b2 {[%eval -116,17]
[%emt 0:00:17]} 27. Bd3 $5 {[%eval 1020,21] [%emt 0:00:32]} (27. Bg7 {also was
considered} Nd6 28. exf7+ Nxf7 (28... Rxf7 29. Bxg6) 29. Bc2 $18) 27... Bd6 {
[%eval 0,19] [%emt 0:00:17]} 28. Qh3 {[%eval 626,26] 79} Na3 {[%eval 32,16]
[%emt 0:00:17]} 29. Ng7 {[%eval 999,21] [%emt 0:00:19]} Qd8 {[%eval 61,17]
[%emt 0:00:16]} 30. Qg4 {[%eval 1036,22] [%emt 0:00:30]} Bc8 {[%eval 71,17]
[%emt 0:00:16]} 31. Nh4 {[%eval 802,20] [%emt 0:00:13]} Ne7 {[%eval 100,13]
[%emt 0:00:16]} 32. Nh5 $1 {[%eval 1245,18] [%emt 0:00:02]} Bxe6 {[%eval 0,18]
[%emt 0:00:16]} 33. Nf6+ {[%eval 1002,24] [%emt 0:00:13]} Kh8 {[%eval 0,8]
[%emt 0:00:00] OpenTal only seeing 0.00 here, just like SF vs AlphaZero} 34.
Qg5 {[%eval 1061,25] [%emt 0:00:07]} Rg8 {[%eval -78,16] [%emt 0:00:16]} 35.
Nxg6+ $1 {[%eval 1092,22] [%emt 0:00:03] burnin' away :)} Rxg6 {[%eval -124,21]
[%emt 0:00:15]} 36. Qh5 $1 {[%eval 1057,24] [%emt 0:00:12]} Ng8 {[%eval -142,
19] [%emt 0:00:15]} ({if} 36... Rxf6 37. Bf8+ Kg8 38. Bxe7 Bxe7 39. Qh7+) 37.
Bg5+ {[%eval 1075,23] [%emt 0:00:01]} Rh6 $8 {[%eval -176,21] [%emt 0:00:15]} (
{if} 37... Kg7 38. Qh7+ Kf8 39. Bxg6 fxg6 40. Rxe6 $18 {is crushing}) 38. Bxh6
{[%eval 1082,26] [%emt 0:00:04]} Nxf6 {[%eval -190,24] [%emt 0:00:15]} 39. Qg5
{[%eval 1082,29] [%emt 0:00:08]} Bf8 $8 {[%eval -217,25] [%emt 0:00:15]} 40.
Bxf8 {[%eval 1082,31] [%emt 0:00:12]} Qxf8 {[%eval -325,24] [%emt 0:00:15]} 41.
Qxf6+ {[%eval 1082,30] [%emt 0:00:06]} Qg7 {[%eval -383,25] [%emt 0:00:14]} 42.
Qd8+ $1 {[%eval 1100,29] [%emt 0:00:04]} Qg8 {[%eval -423,23] [%emt 0:00:14]}
43. Qh4+ {[%eval 1152,27] [%emt 0:00:08]} Kg7 {[%eval -459,23] [%emt 0:00:14]}
44. d5 $1 {[%eval 1181,28] [%emt 0:00:03]} Bxd5 {[%eval -476,22] [%emt 0:00:14]
} 45. Qg5+ {[%eval 1425,24] [%emt 0:00:14]} Rg6 {[%eval -397,21] [%emt 0:00:14]
} 46. Qxd5 {[%eval 1455,25] [%emt 0:00:02]} Nxb1 {[%eval -469,21] [%emt 0:00:
14]} 47. Bxg6 $1 {[%eval 2244,25] [%emt 0:00:02] some nifty winning technique
at the end} Kxg6 {[%eval -596,22] [%emt 0:00:13]} 48. Re3 {[%eval 5673,31]
[%emt 0:00:03] User Adjudication} 1-0
[/pgn]

A monumental attacking effort, carried out against a worthy opponent, itself capable of mounting a similar kind of attack.

BrendanJNorman
Posts: 1416
Joined: Sun Feb 07, 2016 11:43 pm
Full name: Brendan J Norman

Re: sample game: OpenTal vs Thinker Active

Post by BrendanJNorman » Sun Dec 24, 2017 1:20 am

carldaman wrote:If you liked that last game you might also enjoy the following effort. Not to take anything away from the brilliant OpenTal, which truly resembles the real Tal very closely. :) I fully appreciate how important an achievement is to have an engine playing almost exactly like the great Tal.

My goal has been somewhat similar, but I've wanted to get to a point where the machine not only plays like Morphy, Anderssen, Tal, Nezhmetdinov or Shirov, but also can rise above them in strength while retaining the same style, and being (why not?) even stronger than any human alive, past or present.

The results have been amazing, as far as I can tell. I only wish I had more time and better health to devote to such a project. The "entity" currently goes by the slightly humorous name Zero-the-Hero, to poke some fun at the recent hoopla involving AZ beating Stockfish. At this point, I'm not yet at liberty to reveal more than just games and analysis.

Here's its best win vs OT1:


[pgn]
[Event "Tourn08o-10m8s"]
[Date "2017.12.23"]
[Round "3"]
[White "Zero-the-Hero"]
[Black "OpenTal 1.0"]
[Result "1-0"]
[ECO "D15"]
[WhiteElo ""]
[BlackElo "2650"]
[PlyCount "95"]
[TimeControl "600+8"]

1. d4 {[%eval 113,25] [%emt 0:00:49]} d5 2. c4 {[%eval 118,25] [%emt 0:00:17]}
dxc4 3. Nf3 {[%eval 151,26] [%emt 0:00:35]} Nf6 4. Nc3 {[%eval 136,25] [%emt 0:
00:38]} c6 5. e4 $5 {[%eval 172,22] [%emt 0:00:03]} b5 6. Be2 $5 {[%eval 142,
26] [%emt 0:00:19]} e6 {[%eval -11,22] [%emt 0:00:23]} 7. O-O {[%eval 140,25]
[%emt 0:00:34]} Be7 {[%eval 0,23] [%emt 0:00:23]} 8. Qc2 $5 {[%eval 153,24]
[%emt 0:00:23]} (8. a4 b4 9. e5) 8... O-O {[%eval 38,20] [%emt 0:00:23] Black
does not know what dangers await...} (8... Bb7) 9. e5 $5 {[%eval 176,24]
[%emt 0:00:11]} Nd5 {[%eval 46,21] [%emt 0:00:22]} (9... Ne8 10. b3 $36) 10.
Ne4 $5 $146 {[%eval 168,25] [%emt 0:00:12] What's fascinating is that this
move is a novelty, while in a previous game between two corr heavy weights on
ICCF, b3 was played, which petered out to a draw} (10. b3 cxb3 11. axb3 Nxc3
12. Qxc3 Bb7 13. Qc2 Na6 14. Qe4 Nb4 15. Qg4 Kh8 16. Bd2 a5 17. Rfd1 Qb6 18.
Rac1 Na2 19. Rc2 Nb4 20. Rc3 Nd5 21. Rd3 Qc7 22. Nh4 f5 23. exf6 Nxf6 24. Qh3
Kg8 25. Nf3 Bd6 26. Rc3 Qe7 27. Bd3 Bb4 28. Rcc1 Bxd2 29. Rxd2 g6 30. Ng5 Bc8
31. Rxc6 Bd7 32. Rc5 a4 33. bxa4 Rxa4 34. Bf1 Nd5 35. Nf3 Rxd4 36. Nxd4 Qxc5
37. Nxb5 Qc1 38. Rxd5 Qe1 39. f3 Qe3+ 40. Kh1 Qe1 41. Kg1 Qe3+ {1/2-1/2 (41)
Dhanish,P (2600)-Neto,H (2590) ICCF 2014}) 10... Nd7 {[%eval 48,20] [%emt 0:00:
22]} (10... h6 11. Ng3 $5 (11. Bd2 $5 Bb4 $5 12. Bxb4 Nxb4 13. Qd2 $1 Nd5 14.
g3) 11... Nd7 12. Qe4 c5 13. a4 cxd4 (13... Bb7 14. Qg4) 14. Qxd4 f5 $5 15.
exf6 Bc5 16. Qg4 (16. Qh4) 16... N7xf6 17. Qg6 Ne7) 11. Neg5 $40 {[%eval 267,
25] [%emt 0:00:23]} g6 {[%eval 37,20] [%emt 0:00:21]} 12. a4 $36 {[%eval 226,
26] 95} Bb7 {[%eval 57,19] [%emt 0:00:21]} 13. Qe4 $1 {[%eval 175,26] 73} c3 $2
{[%eval 65,19] [%emt 0:00:21]} (13... Kg7 14. Qg4 h6 15. Ne4 Rh8 (15... h5 16.
Qg3) 16. Bd2 $1 (16. h4 Rh7 (16... Kf8 $5 17. Qg3 $1) 17. Qg3 a6 18. Nd6 Bxd6
19. exd6 N7f6 20. Bd1 $5 Nh5 (20... Kg8 21. Bc2) 21. Qh2 Kg8 22. Re1 Ng7 23.
Bc2 Nf5 24. g4 Nb4 25. Bb1 Nxd6 26. Ne5 $44) 16... Kg8 17. h4 $1 Rh7 18. Rfb1
$5 c5 (18... a5 19. b3 c3 20. Nxc3 Nxc3 21. Bxc3 b4 22. Bb2 Rc8 23. Rd1 h5 (
23... Kh8 24. Rac1) 24. Qg3 $16) 19. axb5 cxd4 20. Bxc4 Qb8 21. Nd6 $16 Bxd6 (
21... Ne3 $5 22. Qxd4 $1 Nxc4 23. Nxc4 $16) 22. exd6 Qxd6 23. Qxd4 e5 24. Qd3
$18 Nc5 (24... -- 25. Bc3 $140 $18 Nc5 26. Bxe5) 25. Qa3 Nb6 26. Bf1 Bxf3 27.
Bb4 $1 Bd5 28. Bxc5 $18 Qe6 29. Qe3 $1) 14. Qh4 {[%eval 376,23] [%emt 0:00:07]
Zero already sees itself as winning} h5 {[%eval 65,21] [%emt 0:00:20]} 15. Qg3
$1 {[%eval 417,24] [%emt 0:00:13]} b4 {[%eval 53,20] [%emt 0:00:20]} 16. Nxe6
$1 {[%eval 465,24] [%emt 0:00:19] let the fireworks begin} Qe8 {[%eval 57,20]
[%emt 0:00:20]} 17. Nf4 $3 {[%eval 582,24] [%emt 0:00:05] not taking the Rf8}
c5 {[%eval 71,18] [%emt 0:00:19]} 18. Nxh5 {[%eval 691,20] [%emt 0:00:05]} Rc8
{[%eval 76,18] [%emt 0:00:19]} 19. Bd3 {[%eval 736,24] [%emt 0:00:14]} Rc6 {
[%eval 48,17] [%emt 0:00:19]} 20. Re1 {[%eval 780,24] [%emt 0:00:36]} cxb2 {
[%eval 0,19] [%emt 0:00:19]} 21. Bxb2 {[%eval 789,24] [%emt 0:00:03]} c4 {
[%eval 0,21] [%emt 0:00:18]} 22. Bf5 $1 {[%eval 900,22] [%emt 0:00:05] the
board is catching on fire!} c3 {[%eval 0,19] [%emt 0:00:18]} 23. Bc1 {[%eval
966,22] [%emt 0:00:12]} N7b6 {[%eval 0,20] [%emt 0:00:18]} 24. e6 $1 {[%eval
1055,22] [%emt 0:00:31]} Nc4 {[%eval 0,20] [%emt 0:00:18]} 25. Bh6 {[%eval
1012,22] [%emt 0:00:19]} b3 {[%eval 0,16] [%emt 0:00:17] OpenTal is still
assessing this as 0.00 (dynamically balanced?), but White (Zero) can see past
the zeroes, lol} 26. Rab1 {[%eval 998,23] [%emt 0:00:49]} b2 {[%eval -116,17]
[%emt 0:00:17]} 27. Bd3 $5 {[%eval 1020,21] [%emt 0:00:32]} (27. Bg7 {also was
considered} Nd6 28. exf7+ Nxf7 (28... Rxf7 29. Bxg6) 29. Bc2 $18) 27... Bd6 {
[%eval 0,19] [%emt 0:00:17]} 28. Qh3 {[%eval 626,26] 79} Na3 {[%eval 32,16]
[%emt 0:00:17]} 29. Ng7 {[%eval 999,21] [%emt 0:00:19]} Qd8 {[%eval 61,17]
[%emt 0:00:16]} 30. Qg4 {[%eval 1036,22] [%emt 0:00:30]} Bc8 {[%eval 71,17]
[%emt 0:00:16]} 31. Nh4 {[%eval 802,20] [%emt 0:00:13]} Ne7 {[%eval 100,13]
[%emt 0:00:16]} 32. Nh5 $1 {[%eval 1245,18] [%emt 0:00:02]} Bxe6 {[%eval 0,18]
[%emt 0:00:16]} 33. Nf6+ {[%eval 1002,24] [%emt 0:00:13]} Kh8 {[%eval 0,8]
[%emt 0:00:00] OpenTal only seeing 0.00 here, just like SF vs AlphaZero} 34.
Qg5 {[%eval 1061,25] [%emt 0:00:07]} Rg8 {[%eval -78,16] [%emt 0:00:16]} 35.
Nxg6+ $1 {[%eval 1092,22] [%emt 0:00:03] burnin' away :)} Rxg6 {[%eval -124,21]
[%emt 0:00:15]} 36. Qh5 $1 {[%eval 1057,24] [%emt 0:00:12]} Ng8 {[%eval -142,
19] [%emt 0:00:15]} ({if} 36... Rxf6 37. Bf8+ Kg8 38. Bxe7 Bxe7 39. Qh7+) 37.
Bg5+ {[%eval 1075,23] [%emt 0:00:01]} Rh6 $8 {[%eval -176,21] [%emt 0:00:15]} (
{if} 37... Kg7 38. Qh7+ Kf8 39. Bxg6 fxg6 40. Rxe6 $18 {is crushing}) 38. Bxh6
{[%eval 1082,26] [%emt 0:00:04]} Nxf6 {[%eval -190,24] [%emt 0:00:15]} 39. Qg5
{[%eval 1082,29] [%emt 0:00:08]} Bf8 $8 {[%eval -217,25] [%emt 0:00:15]} 40.
Bxf8 {[%eval 1082,31] [%emt 0:00:12]} Qxf8 {[%eval -325,24] [%emt 0:00:15]} 41.
Qxf6+ {[%eval 1082,30] [%emt 0:00:06]} Qg7 {[%eval -383,25] [%emt 0:00:14]} 42.
Qd8+ $1 {[%eval 1100,29] [%emt 0:00:04]} Qg8 {[%eval -423,23] [%emt 0:00:14]}
43. Qh4+ {[%eval 1152,27] [%emt 0:00:08]} Kg7 {[%eval -459,23] [%emt 0:00:14]}
44. d5 $1 {[%eval 1181,28] [%emt 0:00:03]} Bxd5 {[%eval -476,22] [%emt 0:00:14]
} 45. Qg5+ {[%eval 1425,24] [%emt 0:00:14]} Rg6 {[%eval -397,21] [%emt 0:00:14]
} 46. Qxd5 {[%eval 1455,25] [%emt 0:00:02]} Nxb1 {[%eval -469,21] [%emt 0:00:
14]} 47. Bxg6 $1 {[%eval 2244,25] [%emt 0:00:02] some nifty winning technique
at the end} Kxg6 {[%eval -596,22] [%emt 0:00:13]} 48. Re3 {[%eval 5673,31]
[%emt 0:00:03] User Adjudication} 1-0
[/pgn]

A monumental attacking effort, carried out against a worthy opponent, itself capable of mounting a similar kind of attack.
Great work (and games) mate, can't wait to see how you develop this guy!

Myself, I have 3 new ProDeo personalities and about 5 new Rodents (Kasparov, Positional Vlad, Positional Sergei, Capablanca, etc) still coming, plus a Fruit Reloaded personality.

Too much fun...Alpha who? :lol: :wink:

Post Reply