Impressive - Star Beryl finds Marshall Levitsky Queen sac

[JonP01]

As some of you might know, I have been very impressed with my new Novag Star Beryl. Not only is it's playing strength and style very surprising (if not amazing) for a 4K program, but it also has an excellent openings library full of variety. I've yet to see it repeat an opening after many games with it. Additionally, it does not seem prone to the afflictions that plague it's bigger brother (Star Opal / Carnelian II) - namely the ocassional bizarre blunders where it drops major pieces for no reason at all. Add to that a fine finish in exquisite textuered grey plastic and this machine is one of my firm favourites. What more could one ask for, except perhaps for a Deluxe version in simulated walnut a la Carnelian II (plastic, of course).

Yesterday I decided to see how long it would take to solve one of my favourite positions of all time - the aesthetically very pretty but deadly move 23...Qg3!! played by Frank Marshall against Stepan Levitsky in 1912:

[d] 5rk1/pp4pp/4p3/2R3Q1/3n4/2q4r/P1P2PPP/5RK1 b - - 0 23

To be honest I was wondering if the machine could solve this at all, since the longest variation is something like around 10 ply before all the smoke clears and the pieces are off the board.

I was amazed, however, to find that Star Beryl solved this position in a mere 12 seconds! Yes, that's 12 seconds for a mere 4K program running on an incredibly modest processor that can't even hope to compete with a 6301Y, let alone a 6502

No question about it - this is amazing. I am starting to think that for practical play, this machine might even be better than the current Novag 16K programs, since (touch wood), it seems to be only slightly weaker, but it does not make those ocassional horrific blunders that the larger program makes (and which spoil otherwise good games).

That said, like even the best of the current Novags (Citrine, Star Ruby), the selective search can sometimes miss very obvious moves that other programs can pick up on quite easily. In this 40 moves in 2 hour game played yesterday, Beryl commits an inaccuracy in the opening leaving it fatally vulnerable on the A1 - H8 diagonal. The Saitek Maestro did not waste time inflicting a fatal blow. That makes the current match score between these two 1.5 - 0.5 in favour of Maestro.

[Event "40 in 2 Blair Athol"]
[Site "?"]
[Date "2007.05.12"]
[Round "?"]
[White "Saitek Maestro"]
[Black "Novag Star Beryl"]
[Result "1-0"]
[ECO "D04"]
[PlyCount "107"]
[EventDate "2007.??.??"]

1. Nf3 d5 2. d4 Nf6 3. e3 g6 4. Nbd2 Nc6 5. Be2 e6 6. O-O Be7 7. b3 O-O 8. c4
Bd7 9. a3 Ne4 10. Nxe4 dxe4 11. Nd2 f5 12. Bb2 Re8 13. Rc1 Qc8 14. Qc2 Qd8 15.
d5 exd5 16. cxd5 f4 17. dxc6 Bxc6 18. Qc3 Bf6 19. Qxf6 Qxf6 20. Bxf6 Kf7 21.
Bg5 fxe3 22. fxe3+ Kg7 23. Bf6+ Kh6 24. g4 g5 25. Rc5 Rf8 26. Bb5 Bxb5 27. Rxb5
b6 28. Nxe4 c5 29. Rd1 Rae8 30. Bxg5+ Kg6 31. Rd6+ Kf7 32. Rf6+ Kg8 33. Bh6 Rf7
34. Rxf7 Kxf7 35. Nd6+ Kg6 36. Nxe8 Kxh6 37. e4 Kg5 38. b4 Kxg4 39. bxc5 bxc5
40. Rxc5 Kf3 41. e5 h5 42. e6 Ke3 43. Nd6 h4 44. e7 Kd4 45. Ra5 Kc3 46. Rxa7
Kd3 47. a4 Kd4 48. a5 Kd5 49. Rd7 h3 50. a6 Kc5 51. a7 Kc6 52. e8=Q Kb6 53. Qh5
Kc6 54. Qb5# 1-0

[Steve B]

Hi Jon

Bravo to the Star Beryl!
seems like you uncovered a little known "Best Buy" here

could also be the most underrated chess computer released for quite some time

nice to see you actually pleased with a dedicated computer at last regards
Steve

[tiger]

As some of you might know, I have been very impressed with my new Novag Star Beryl. Not only is it's playing strength and style very surprising (if not amazing) for a 4K program, but it also has an excellent openings library full of variety. I've yet to see it repeat an opening after many games with it. Additionally, it does not seem prone to the afflictions that plague it's bigger brother (Star Opal / Carnelian II) - namely the ocassional bizarre blunders where it drops major pieces for no reason at all. Add to that a fine finish in exquisite textuered grey plastic and this machine is one of my firm favourites. What more could one ask for, except perhaps for a Deluxe version in simulated walnut a la Carnelian II (plastic, of course).

Yesterday I decided to see how long it would take to solve one of my favourite positions of all time - the aesthetically very pretty but deadly move 23...Qg3!! played by Frank Marshall against Stepan Levitsky in 1912:

[d] 5rk1/pp4pp/4p3/2R3Q1/3n4/2q4r/P1P2PPP/5RK1 b - - 0 23

To be honest I was wondering if the machine could solve this at all, since the longest variation is something like around 10 ply before all the smoke clears and the pieces are off the board.

I was amazed, however, to find that Star Beryl solved this position in a mere 12 seconds! Yes, that's 12 seconds for a mere 4K program running on an incredibly modest processor that can't even hope to compete with a 6301Y, let alone a 6502

No question about it - this is amazing. I am starting to think that for practical play, this machine might even be better than the current Novag 16K programs, since (touch wood), it seems to be only slightly weaker, but it does not make those ocassional horrific blunders that the larger program makes (and which spoil otherwise good games).

That said, like even the best of the current Novags (Citrine, Star Ruby), the selective search can sometimes miss very obvious moves that other programs can pick up on quite easily. In this 40 moves in 2 hour game played yesterday, Beryl commits an inaccuracy in the opening leaving it fatally vulnerable on the A1 - H8 diagonal. The Saitek Maestro did not waste time inflicting a fatal blow. That makes the current match score between these two 1.5 - 0.5 in favour of Maestro.

[Event "40 in 2 Blair Athol"]
[Site "?"]
[Date "2007.05.12"]
[Round "?"]
[White "Saitek Maestro"]
[Black "Novag Star Beryl"]
[Result "1-0"]
[ECO "D04"]
[PlyCount "107"]
[EventDate "2007.??.??"]

1. Nf3 d5 2. d4 Nf6 3. e3 g6 4. Nbd2 Nc6 5. Be2 e6 6. O-O Be7 7. b3 O-O 8. c4
Bd7 9. a3 Ne4 10. Nxe4 dxe4 11. Nd2 f5 12. Bb2 Re8 13. Rc1 Qc8 14. Qc2 Qd8 15.
d5 exd5 16. cxd5 f4 17. dxc6 Bxc6 18. Qc3 Bf6 19. Qxf6 Qxf6 20. Bxf6 Kf7 21.
Bg5 fxe3 22. fxe3+ Kg7 23. Bf6+ Kh6 24. g4 g5 25. Rc5 Rf8 26. Bb5 Bxb5 27. Rxb5
b6 28. Nxe4 c5 29. Rd1 Rae8 30. Bxg5+ Kg6 31. Rd6+ Kf7 32. Rf6+ Kg8 33. Bh6 Rf7
34. Rxf7 Kxf7 35. Nd6+ Kg6 36. Nxe8 Kxh6 37. e4 Kg5 38. b4 Kxg4 39. bxc5 bxc5
40. Rxc5 Kf3 41. e5 h5 42. e6 Ke3 43. Nd6 h4 44. e7 Kd4 45. Ra5 Kc3 46. Rxa7
Kd3 47. a4 Kd4 48. a5 Kd5 49. Rd7 h3 50. a6 Kc5 51. a7 Kc6 52. e8=Q Kb6 53. Qh5
Kc6 54. Qb5# 1-0

Chess Tiger 2007 for PalmOS solves it at ply depth 4 in 1s (Palm TX, 768Kb hash tables).


// Christophe

[Uri Blass]

This position is clearly a bad test position because black has many winning moves

I am not sure if Qg3 is the fastest way to win.

New game
[d]5rk1/pp4pp/4p3/2R3Q1/3n4/2q4r/P1P2PPP/5RK1 b - - 0 1

Analysis by Rybka 2.2n2 32-bit :

1...Qc3-b2
-+ (-3.17) Depth: 2 00:00:00
1...Qc3-g3
-+ (-3.37) Depth: 2 00:00:00
1...Qc3-g3
-+ (-3.33) Depth: 3 00:00:00
1...Qc3-g3
-+ (-3.48) Depth: 4 00:00:00
1...Qc3-g3 2.Qg5xg3
-+ (-3.55) Depth: 5 00:00:00
1...Qc3-g3 2.Qg5xg3 Nd4-e2+ 3.Kg1-h1 Ne2xg3+
-+ (-3.55) Depth: 6 00:00:00 4kN
1...Qc3-b4 2.Rf1-d1 Rh3-c3
-+ (-3.86) Depth: 6 00:00:00 10kN
1...Qc3-b4 2.Qg5-e5 Rh3-c3 3.Rc5xc3
-+ (-3.20) Depth: 7 00:00:00 16kN
1...Qc3-g3 2.Qg5xg3 Nd4-e2+ 3.Kg1-h1 Ne2xg3+
-+ (-3.80) Depth: 7 00:00:00 17kN
1...Qc3-g3 2.Qg5xg3 Nd4-e2+ 3.Kg1-h1 Ne2xg3+
-+ (-3.80) Depth: 8 00:00:00 19kN
1...Qc3-g3 2.Qg5xg3 Nd4-e2+ 3.Kg1-h1 Ne2xg3+
-+ (-3.80) Depth: 9 00:00:00 25kN
1...Qc3-g3 2.Qg5xg3 Nd4-e2+ 3.Kg1-h1 Ne2xg3+
-+ (-3.83) Depth: 10 00:00:02 36kN
1...Qc3-g3 2.Qg5xg3 Nd4-e2+ 3.Kg1-h1 Ne2xg3+
-+ (-3.86) Depth: 11 00:00:03 56kN
1...Qc3-g3 2.Qg5xg3 Nd4-e2+ 3.Kg1-h1 Ne2xg3+
-+ (-3.86) Depth: 12 00:00:05 88kN
1...Qc3-g3 2.Qg5xg3 Nd4-e2+ 3.Kg1-h1 Ne2xg3+
-+ (-3.86) Depth: 13 00:00:07 154kN
1...Qc3-b2 2.Kg1-h1 Rh3-c3 3.Rc5xc3 Qb2xc3 4.f2-f3 Nd4xc2 5.Qg5-g4 Qc3-c6 6.Rf1-c1 h7-h6
-+ (-3.88) Depth: 13 00:00:20 1543kN
1...Qc3-b2 2.Kg1-h1 Rh3-c3 3.Rc5xc3 Qb2xc3 4.f2-f3 Nd4xc2 5.Qg5-e7 Qc3-c4 6.Rf1-g1 Qc4xa2 7.Qe7xb7
-+ (-3.98) Depth: 14 00:01:34 7918kN
1...Qc3-b2 2.Kg1-h1 Rh3-c3 3.Rc5xc3 Qb2xc3 4.f2-f3 Qc3-c4 5.Rf1-c1 Nd4-e2 6.Rc1-a1 Qc4xc2 7.a2-a4 Qc2-b2
-+ (-3.92) Depth: 15 00:01:42 8524kN

(, 13.05.2007)

[Steve B]

This position is clearly a bad test position because black has many winning moves

I am not sure if Qg3 is the fastest way to win.

1...Qc3-b2 2.Kg1-h1 Rh3-c3 3.Rc5xc3 Qb2xc3 4.f2-f3 Qc3-c4 5.Rf1-c1 Nd4-e2 6.Rc1-a1 Qc4xc2 7.a2-a4 Qc2-b2
-+ (-3.92) Depth: 15 00:01:42 8524kN

(, 13.05.2007)

Uri
Chess legend and history tells us that the spectators watching the game ..stood up and cheered and SHOWERED the board with GOLD Coins when Marshall placed his Queen En prise into a nest of white pawns

after Qb2 i think the spectators are taking a short break and going to the bathroom

Regards
Steve

[fern]

JJaaaaaaa jajaja
what a wonderful reply!!!!!
Uri maybe is right, but he misses the artistic anf funny part of that specific move.
Chess is also ART.

Laughing regards
Fernando

[rightrook]

Me laugh too...

Novag Sap2 finds Qg3 in less than a second (at 3 ply)
and announces mate!

regards
Rob

Hey Steve..funny reply...good one...

[rightrook]

Sorry...my mistake...there is no forced mate here..)

I forgot to enter the white queen on the board position..

However, 3 ply still finds Qg3 for blacks best move here....etc.

[JonP01]

after Qb2 i think the spectators are taking a short break and going to the bathroom

lol Steve. Yes, much of what attracts me to chess is the aesthetic beauty of many sacrificial combinations. If that move had been played against me, I think at first my jaw would have dropped, then I would have laughed indignantly, then got up out of my chair, held my opponent's arm up and shouted to everyone in the tournament hall to come and take a look Whilst I am not an expert in every classic game ever played, for me it is hard to top this move, even if computers can easily find the solution.

[JonP01]

Novag Sap2 finds Qg3 in less than a second (at 3 ply)
and announces mate!

Hi Robert,

I guess that is part of the explanation why Star Beryl finds the answer so incredibly quickly for the tiny program it is and the miniscule hardware it runs on. Since both programs are by the same author, it makes sense that some of the ideas in the top programs make their way to the basic programs where possible. In this case it's clear that Kittenger programs have a lot of extension capability for captures, checks and quiesence well beyond the basic search depth.

Although Star Beryl does not have an info display, my guess it would be showing 3 ply as well by the time it finds the move.

Still Stunned by Star Beryl Regards

Jonathan