Millenium new release of chess computer

[Yarc]

Game 3 was a win for MCG in 84 moves:

[Event "40 moves in 2 hours"]
[Date "2015.12.29"]
[Round "3"]
[White "MCG"]
[Black "Cosmos"]
[Result "1-0"]
[ECO "C43"]

1.e4 e5 2.Nf3 Nf6 3.d4 Nxe4 4.Bd3 d5 5.Nxe5 Bd6 6.O-O O-O 7.c4 c6 8.Nc3 Bxe5
9.dxe5 Nxc3 10.bxc3 dxc4 11.Bxc4 Qa5 12.Qd4 Rd8 13.Qf4 Be6 14.Bxe6 fxe6 15.c4
Nd7 16.Qg4 Qxe5 17.Rb1 Nf6 18.Qg5 Qxg5 19.Bxg5 b6 20.Rfe1 Kf7 21.Rb2 c5
22.Rbe2 Rd4 23.Bxf6 gxf6 24.Rxe6 Rxc4 25.Re7+ Kg6 26.a3 Rc2 27.h4 Rd8 28.Rxa7
Rdd2 29.Re3 Rxf2 30.Rg3+ Kf5 31.Rxh7 Rfd2 32.Rb7 Rb2 33.Kh2 Rd4 34.Kh3 Rb1
35.Rc3 Rh1+ 36.Kg3 Rg4+ 37.Kf2 Rf4+ 38.Kg3 Rfxh4 39.Rxb6 Rg4+ 40.Kf2 Rf4+
41.Ke3 Re1+ 42.Kd2 Re5 43.Rc6 Rf2+ 44.Kd3 Rxg2 45.Kc4 Rg4+ 46.Kb5 Rg1 47.Rf3+
Ke4 48.Rcxf6 Rc1 49.R6f4+ Kd5 50.Rd3+ Ke6 51.Kc6 c4 52.Rd6+ Ke7 53.Rd7+ Ke8
54.Rb7 Rh5 55.Re4+ Kf8 56.Rb2 Kg7 57.Ra2 Kf6 58.a4 c3 59.a5 c2 60.Re2 Rh8
61.Rd2 Rc8+ 62.Kd7 Ra8 63.Kc7 Ra7+ 64.Kc6 Ra8 65.a6 Ke5 66.a7 Ke6 67.Rg2 Kf7
68.Rf2+ Ke6 69.Re2+ Kf5 70.Re7 Kf4 71.Rc7 Ke5 72.Kb7 Rf8 73.Rcxc2 Rb1+
74.Rcb2 Rf7+ 75.Ka6 Rf6+ 76.Ka5 Rxb2 77.Rxb2 Rf8 78.Rb8 Rf1 79.Rb5+ Kd6
80.Kb6 Rf8 81.Ra5 Ra8 82.Kb7 Rd8 83.a8=Q Rxa8 84.Kxa8 1-0

The MCG draws level with the Cosmos.

Final position:
[fen]K7/8/3k4/R7/8/8/8/8 b - - 0 84[/fen]

[Yarc]

Game 4 and another win for MCG in 81 moves!

[Event "40 moves in 2 hours"]
[Date "2016.01.01"]
[Round "4"]
[White "Cosmos"]
[Black "MCG"]
[Result "0-1"]
[ECO "D02"]

1.d4 Nf6 2.Nf3 d5 3.Bf4 c5 4.e3 e6 5.c3 Qb6 6.Qb3 Nc6 7.Qxb6 axb6 8.Bd3 Be7
9.O-O Ne4 10.Bxe4 dxe4 11.Nfd2 f5 12.Nc4 Ra6 13.Nd6+ Bxd6 14.Bxd6 cxd4
15.cxd4 Kd7 16.Bg3 g5 17.Nc3 h5 18.h4 gxh4 19.Bxh4 Nb4 20.Rfc1 Ra5 21.Kf1 Nd3
22.Rc2 Rg8 23.Bg3 Nb4 24.Rd2 Rg4 25.a3 h4 26.Bh2 Rg6 27.Ne2 Rg4 28.Rc1 Nc6
29.Nf4 Ne7 30.Rdc2 Nc6 31.Ke2 Kd8 32.Rc3 Bd7 33.Kd2 e5 34.dxe5 Rxe5 35.Ke2
Be6 36.R1c2 Bf7 37.b4 b5 38.Rd2+ Ke8 39.Rd6 Bc4+ 40.Kd2 Re7 41.a4 Ne5 42.axb5
Bxb5 43.Rc8+ Kf7 44.Kc3 Nd3 45.Rd5 Bd7 46.Rc7 Nxf4 47.Bxf4 Be6 48.Rxe7+ Kxe7
49.Rb5 Bc8 50.g3 h3 51.Rb6 Be6 52.Rxb7+ Bd7 53.Rb8 Be8 54.Ra8 Rg8 55.Ra1 Bb5
56.Rh1 Rh8 57.Bg5+ Kd6 58.Bf6 Rh7 59.Bh4 Rc7+ 60.Kb3 Bd7 61.Kb2 Rc4 62.Rxh3
Rxb4+ 63.Kc3 Ra4 64.Bf6 Ra2 65.Rh2 Be6 66.Kb4 Rc2 67.Be5+ Kxe5 68.f4+ exf3
69.Rxc2 Ke4 70.Kc5 Bb3 71.Rb2 Bd1 72.Rh2 Kxe3 73.Rh1 Be2 74.Kd6 f2 75.Ke5 Bd3
76.Rc1 Kf3 77.Rd1 Kg4 78.Ke6 f1=Q 79.Rxf1 Bxf1 80.Ke5 Bh3 81.Kd4 Kxg3 0-1

I resigned on behalf of the Cosmos.

Final position:
[fen]8/8/8/5p2/3K4/6kb/8/8 w - - 0 82[/fen]

[Yarc]

On to Gamer 5 which ended in a draw in just 38 moves:

[Event "40 moves in 2 hours"]
[Date "2016.01.09"]
[Round "5"]
[White "MCG"]
[Black "Cosmos"]
[Result "1/2-1/2"]
[ECO "B10"]

1.e4 c6 2.Nf3 d5 3.Nc3 e6 4.exd5 cxd5 5.Bb5+ Nc6 6.Ne5 Qc7 7.d4 Bd6 8.Bf4 Ne7
9.O-O O-O 10.Nxc6 bxc6 11.Bxd6 Qxd6 12.Bd3 Rb8 13.b3 c5 14.Qh5 g6 15.Qe5 Qxe5
16.dxe5 Nc6 17.f4 Bb7 18.Na4 Rfc8 19.Nxc5 Nxe5 20.fxe5 Rxc5 21.Rf6 Ra5
22.Raf1 Ba6 23.Bxa6 Rxa6 24.Rxf7 Rxa2 25.Rc7 a6 26.Rf6 Rf8 27.Kf2 Rxf6+
28.exf6 e5 29.Ke3 h5 30.Kd3 Ra1 31.Re7 Rd1+ 32.Ke2 Rg1 33.Kf2 Rc1 34.Rc7 Kf8
35.Ke2 Rg1 36.Kf2 Rc1 37.Ke3 Re1+ 38.Kf2 Rc1 1/2-1/2

Final position:
[fen]5k2/2R5/p4Pp1/3pp2p/8/1P6/2P2KPP/2r5 w - - 18 39[/fen]

[Yarc]

Game 6 was a victory for Cosmos in 52 moves:

[Event "40 moves in 2 hours"]
[Date "2016.01.10"]
[Round "6"]
[White "Cosmos"]
[Black "MCG"]
[Result "1-0"]
[ECO "D24"]

1.d4 Nf6 2.c4 e6 3.Nc3 d5 4.Nf3 dxc4 5.e3 c5 6.Bxc4 a6 7.O-O b5 8.Bb3 Bb7
9.a4 c4 10.Ba2 b4 11.Ne2 Bxf3 12.gxf3 b3 13.Bb1 Bd6 14.e4 Nfd7 15.Be3 O-O
16.Qc1 Qc7 17.e5 Be7 18.Be4 Ra7 19.d5 Bc5 20.d6 Qb6 21.a5 Qb4 22.Bd2 c3
23.Nxc3 Qc4 24.Na4 Qxc1 25.Rfxc1 Bd4 26.Rc4 Bxe5 27.Be3 f5 28.Bxa7 fxe4
29.fxe4 Bxd6 30.Rd1 Ne5 31.Rc3 Be7 32.Bc5 Nf3+ 33.Kg2 Nh4+ 34.Kg3 Ng6 35.Bxe7
Nxe7 36.Rc7 Nbc6 37.Rdd7 Rf6 38.Nb6 e5 39.Nc8 Ng6 40.Rxg7+ Kf8 41.Rxh7 Ke8
42.Rhd7 Kf8 43.Kg4 Re6 44.Kf5 Nf4 45.h4 Nd4+ 46.Rxd4 exd4 47.Kxf4 Rf6+ 48.Kg3
Re6 49.f3 Rg6+ 50.Kf2 Rh6 51.Ne7 Rd6 52.Nf5 1-0

I resigned on behalf of the MCG.

Final position:
[fen]5k2/2R5/p2r4/P4N2/3pP2P/1p3P2/1P3K2/8 b - - 6 52[/fen]

[Yarc]

In Game 7 the MCG wins well in 43 moves:

[Event "40 moves in 2 hours"]
[Date "2016.01.17"]
[Round "7"]
[White "MCG"]
[Black "Cosmos"]
[Result "1-0"]
[ECO "B01"]

1.e4 d5 2.exd5 Nf6 3.Bb5+ Bd7 4.Bc4 Bf5 5.Nc3 g6 6.Nf3 Bg7 7.d3 O-O 8.Qe2 a6
9.Bb3 b5 10.a3 Bg4 11.Bg5 Nbd7 12.Qe3 Bxf3 13.Qxf3 Nc5 14.Ba2 Rb8 15.d4 Ncd7
16.O-O Re8 17.Rae1 Nb6 18.Re5 b4 19.axb4 Nbd7 20.Re2 Rxb4 21.Rd1 h6 22.Bc1 a5
23.Bb3 Rb8 24.Ba4 Rb4 25.b3 Rb6 26.Bc6 h5 27.Rde1 Rb8 28.Nb5 Rb6 29.Bf4 Qc8
30.Nxc7 Rd8 31.Rxe7 Nb8 32.Be8 Nxe8 33.Nxe8 Bf8 34.Bc7 Bxe7 35.Rxe7 f5 36.d6
Rdxd6 37.Nxd6 Rxd6 38.Bxd6 Qc6 39.Qg3 Qe8 40.Rxe8+ Kg7 41.Rg8+ Kxg8 42.Qxg6+
Kh8 43.Be5# 1-0

The MCG announced mate in 5 in this game.

Final position:
[fen]1n5k/8/6Q1/p3Bp1p/3P4/1P6/2P2PPP/6K1 b - - 2 43[/fen]

[Yarc]

Game 8 was a win for the Cosmos in 55 moves:

[Event "40 moves in 2 hours"]
[Date "2016.01.24"]
[Round "8"]
[White "Cosmos"]
[Black "MCG"]
[Result "1-0"]
[ECO "B08"]

1.e4 d6 2.d4 Nf6 3.Nc3 g6 4.Nf3 Bg7 5.Be2 O-O 6.O-O Bg4 7.h3 Bxf3 8.Bxf3 c6
9.Re1 Qb6 10.e5 Nfd7 11.exd6 exd6 12.Ne2 Re8 13.c3 d5 14.a4 Na6 15.a5 Qd8
16.Bf4 Nf8 17.Qd2 Nc7 18.Bg5 Qd6 19.b4 Nfe6 20.Bh4 Nb5 21.a6 b6 22.Bg3 Qd8
23.Kh1 Ng5 24.Ng1 Rxe1 25.Rxe1 Nxf3 26.Nxf3 Qc8 27.Ra1 Qf5 28.Re1 Rd8 29.Bh4
Rf8 30.Re7 Qb1+ 31.Kh2 Qb3 32.Ne5 Nd6 33.Bg3 Bxe5 34.Bxe5 Nc4 35.Qe1 Nxe5
36.dxe5 c5 37.Rxa7 cxb4 38.cxb4 d4 39.Rb7 Qd5 40.Rxb6 Qa2 41.Qf1 Rc8 42.Rb7
Ra8 43.b5 Qd5 44.f4 d3 45.e6 fxe6 46.Qa1 e5 47.fxe5 h5 48.Qc3 Qe4 49.e6 Qf4+
50.Kg1 Qh6 51.a7 Rf8 52.Qxd3 Qc1+ 53.Kh2 Qf4+ 54.Qg3 Qxg3+ 55.Kxg3 Kh8 1-0

I resigned on behalf of the MCG. So now they are 4 points each! Well done Cosmos.

Final position:
[fen]5r1k/PR6/4P1p1/1P5p/8/6KP/6P1/8 w - - 1 56[/fen]

[Yarc]

On to Game 9 and a draw in 41 moves:

[Event "40 moves in 2 hours"]
[Date "2016.01.26"]
[Round "9"]
[White "MCG"]
[Black "Cosmos"]
[Result "1/2-1/2"]
[ECO "A38"]

1. c4 Nf6 2. Nc3 c5 3. g3 g6 4. Bg2 Bg7 5. Nf3 O-O 6. d3 d6 7. O-O Nc6
8. Qb3 b6 9. Ng5 Bb7 10. Qa4 Qc7 11. Bf4 e5 12. Be3 Nd4 13. Bxb7 Qxb7
14. Rad1 Rfe8 15. Bxd4 exd4 16. Nce4 Nxe4 17. Nxe4 Re7 18. Rfe1 d5
19. cxd5 Qxd5 20. Rc1 f5 21. Ng5 Bh6 22. h4 Rf8 23. Qc4 Qxc4 24. Rxc4 Rfe8
25. Kf1 a5 26. a3 Bg7 27. b4 cxb4 28. axb4 a4 29. b5 Ra7 30. Rc6 Rb7
31. Ra1 Ra8 32. Ne6 a3 33. Rxa3 Rxa3 34. Rc8+ Kf7 35. Nd8+ Ke8 36. Nxb7+ Kd7
37. Rd8+ Kc7 38. Re8 Kd7 39. Rd8+ Kc7 40. Re8 Kd7 41. Rd8+ 1/2-1/2

Final position:
[fen]3R4/1N1k2bp/1p4p1/1P3p2/3p3P/r2P2P1/4PP2/5K2 b - - 10 41[/fen]

[Yarc]

The final one, Game 10 and this was a win for the MCG in 40 moves:

[Event "40 moves in 2 hours"]
[Date "2016.01.30"]
[Round "10"]
[White "Cosmos"]
[Black "MCG"]
[Result "0-1"]

1. e4 c5 2. Nf3 Nc6 3. Nc3 g6 4. d4 cxd4 5. Nxd4 Bg7 6. Be3 Nf6 7. Be2 d6
8. Nxc6 bxc6 9. e5 dxe5 10. Qxd8+ Kxd8 11. Rd1+ Kc7 12. Bc4 e6 13. O-O Nd5
14. a3 Rb8 15. Na4 Nxe3 16. fxe3 f5 17. b4 Bh6 18. Kf2 e4 19. Nc5 e5 20. Be6 Bf8
21. Bxc8 Bxc5 22. Bxf5 gxf5 23. bxc5 Rb5 24. Ke2 Rf8 25. g4 f4 26. exf4 exf4
27. Rd4 Rxc5 28. Rxe4 Rxc2+ 29. Kd3 Rxh2 30. Rexf4 Rxf4 31. Rxf4 a5 32. Kc4 Kb6
33. a4 Rh3 34. Kd4 Rb3 35. Re4 Rb4+ 36. Kd3 Rxe4 37. Kxe4 Kc5 38. g5 Kb4 39. Ke5 c5
40. Kf4 c4 0-1

I resigned on behalf of the Cosmos.

Final position:
[fen]8/7p/8/p5P1/Pkp2K2/8/8/8 w - - 0 41[/fen]

[Yarc]

Well, I think you can see from the result of the 10 game match these two machines are fairly equally matched with the MCG maybe having a slight edge.

My observations and ideas about MCG’s play

In the end game it was able to see much further ahead than the Cosmos, often by 4 or more moves. When just a few pawns remained, it often was looking an extra 6 moves ahead! Probably, the hash tables were kicking in giving it a boost, not to mention the extra processing power.

Despite the MCG’s extra grunt it can still make inferior moves and lose badly to the Cosmos. Considering the imbalanced processing power of these machines (Cosmos 10 MHZ 8-bit vs MCG 48MHZ 32-bit), I have to think that the Cosmos might have the better position evaluation, and or less aggressive tree pruning, thus not missing key moves. This is speculation as I’m no chess programmer, but there must be something to explain why the Cosmos does so well.

Watching the MCG’s thought process, I noticed that once it had started considering moves at a particular ply-level it would always complete its analysis for all moves at that ply. I wondered what impact this may have had on its time control during the match. The Cosmos is different in this respect. I would often see it part-way through analysing moves at a ply-level and then stop with a move before having evaluated all those available. Is the Cosmos maximising its time allocation using this approach? How is the MCG able to determine how much time is required to examine all moves at a ply-level? If it thinks there is not enough time to evaluate all moves, does it stop at the ply-level it has reached and therefore miss out on looking at even some moves at the next ply? This might explain why it sometimes loses its positional advantage in a game.

Also worth noting is that the MCG has a book library of 20,000 moves compared to only 6000 for the Cosmos. I often found that the MCG gained much time in the beginning as a result, but this also shows that the Cosmos was working out recognised book moves. Despite using more clock time, the Cosmos still did very well in the games as can be seen.

These are just some of my thoughts. In any case, the MCG probably does have a rating higher than 2000 but much less than 2100 at least against computers that is. This is still a respectable playing strength for the price! Well done Millenium for releasing this great little machine.

Hope you enjoy examining these games, I certainly enjoyed watching them!

[paulwise3]

Yarc,

Thank you for these games!
The Cosmos has the same program as the GK 2100 (Frans Morsch), so we have seen a new Lang - Morsch duel .
I would have thought the ChessGenius could easily outplay the Cosmos.
Soon I hope to have a Mephisto Chess Challenger (also GK 2100 clone) at 40 MHz instead of 10 MHz (sorry Steve...), and am anxious to see them play against each other .

Duel regards,
Paul

[Yarc]

Paul,

That's okay, I hope people find them interesting.

I had discovered the Cosmos has the same program as the GK2100.
It will be interesting to hear more about your Mephisto Chess Challenger running at 40 MHZ. I suspect this would win over the MCG.

[paulwise3]

Yarc,

I am not that patient to watch them play at tournament level, so I will start playing at 30 sec/move. But when I finish tidying the attic, I may have more room and time to play at higher levels.

@Administrator: How is it with the work on the PGN-tag? That will make it much easier to publish games here!

Will keep you informed regards,
Paul

[Yarc]

Paul,

Yes, playing a 10 game match at 40 game in 2 hours takes a long time. I chose this as it was the only tournament level that both machines supported.

I have heard that some programs only play at their best on tournament levels, don't know if this is the case with either the GK2100 or MCG programs.

I think a match with a faster time control will be interesting all the same.

[IanO]

Paul,

Yes, playing a 10 game match at 40 game in 2 hours takes a long time. I chose this as it was the only tournament level that both machines supported.

I have heard that some programs only play at their best on tournament levels, don't know if this is the case with either the GK2100 or MCG programs.

I think a match with a faster time control will be interesting all the same.

I agree. I also own both the Cosmos and and MCG and they are both superior at blitz controls. Cosmos due to the Morsch null-move secret sauce which gains at least a ply depth in the middle-game, and Genius due to Lang's finely-tuned selective search. Like you, I find that the Cosmos and MCG are evenly matched, even at blitz controls. Cosmos tends to play sharper and riskier, and MCG wants to trade pieces off to get to safe positions.

In fact, I would be quite interested to hear what innovations from the last two decades made it into the MCG engine. One thing to realize, the programs in the dedicated models we hold near and dear represent a time capsule of the development of chess programming from the eighties and early nineties, before innovations such as the null-move heuristic (except for Morsch), automated evaluation tuning, and late-move reductions. Even models produced in the 2000s were using the older engines because the companies had no need to fund further development. The MCG is unique as far as I know for incorporating the latest generation of a dedicated engine, thanks to Lang's continued development of Chess Genius on smartphones.

[Yarc]

Paul,

Yes, playing a 10 game match at 40 game in 2 hours takes a long time. I chose this as it was the only tournament level that both machines supported.

I have heard that some programs only play at their best on tournament levels, don't know if this is the case with either the GK2100 or MCG programs.

I think a match with a faster time control will be interesting all the same.

I agree. I also own both the Cosmos and and MCG and they are both superior at blitz controls. Cosmos due to the Morsch null-move secret sauce which gains at least a ply depth in the middle-game, and Genius due to Lang's finely-tuned selective search. Like you, I find that the Cosmos and MCG are evenly matched, even at blitz controls. Cosmos tends to play sharper and riskier, and MCG wants to trade pieces off to get to safe positions.

In fact, I would be quite interested to hear what innovations from the last two decades made it into the MCG engine. One thing to realize, the programs in the dedicated models we hold near and dear represent a time capsule of the development of chess programming from the eighties and early nineties, before innovations such as the null-move heuristic (except for Morsch), automated evaluation tuning, and late-move reductions. Even models produced in the 2000s were using the older engines because the companies had no need to fund further development. The MCG is unique as far as I know for incorporating the latest generation of a dedicated engine, thanks to Lang's continued development of Chess Genius on smartphones.

What I find interesting, as mentioned in my earlier post, is that the Cosmos hardware seems to be much slower than that of the MCG. I wonder how well the Cosmos would play with MCG's hardware, unless I'm missing something about the Cosmos's hardware. Despite the advances in Computer chess, the MCG is still only drawing (near enough!) with the Cosmos. I think this shows just how good the earlier machines could be. I wonder what version two of the MCG will be like if it comes to fruition?