Collectors Corner..On The New Novag Star Beryl

[Steve B]

As a collector i consider the Star Bery highly collectible because it appears that the SB could well be the STRONGEST 4k program ever released for sale..more testing has to be done however to say this with any degree of certainty

the bulk of my comments however will be from the vantage point of a chess player whose main interest in acquiring the SB is to play competitive chess games against the computer

Cosmetically the computer has a very sleek modern design very appealing to the eye
its is thin enough to slip inside a briefcase and it operates on batteries and Adaptor

there is an LCD display screen whose principal purpose is to display the computers move as there are no LEDS on the board
the display screen is an unmitigated disaster almost impossible to see in good light and there is no differentiation between the letters E and F
indeed the _ part of the letter E is simply hidden in the display screen.it makes no difference what angle you try to view the screen. it is simply impossible to make out the letter E..i have never seen this sort of a manufacturing design flaw before in a chess computer.
the computer insists you make the move for it on the sensory playing surface before you can proceed so if the player confuses the E with the F and as a result enters the wrong move for the computer then the computer will signal an" error" tone and not allow the player to enter his move . so while the confusion between E's And F's is annoying it will not allow you to make the wrong moves for the computer
the information shown in the display is extremely limited to only the computers current move and a 1 move hint while the player is on the move which does not change while the player is pondering his move
no clock times,evaluations ,depth search..etc etc.

even though the computer will give the player a 1 ply hint i dont think the computer ponders while you are on the move
having played several games against it i have yet to see it reply instantly after i made a move even if the move was an obvious recapture or even if it is the move shown as a hint

the touch sensory playing surface is not "feather touch" and needs some degree of force for moves to be registered..not really a big problem unless you want to play a fast Blitz game and then it can become distracting

the actual game play is a nice surprise..probably not stronger then 1500 Elo it does have a decent book and the moves it plays are not silly meaningless ones..the computer can beat you if you are not careful because of the typically active playing style of all Kittinger programs
The SB will recognize draws my Stalemate ,insufficient material and threefold repetition(although the manual does not mention this third draw)
As far as the basic mates the computer could not even mate with K+Qvs.K and instead allowed a draw by 3 fold repetition



game play however can sadly be marred because for some reason the SB makes surprise"out of the blue"game losing blunders even after pondering for a fair amount of time totally inconsistent with its play up to that point in the game

to illustrate this i offer these two games
as there is no level for 40/1hr ..i played two games at 90 seconds per move avg:

Game 1:

in the first game i took the white pieces and the SB stayed in book for 9 moves which is a nice surprise for this little computer


[White "Steve B"]
[Black "Star Beryl"]
[TimeControl "90Sec/Avg"]
[Result "1-0"]

1.d4 d5 2.c4 e6 3.Nf3 Nf6 4.Nc3 c6 5.e3 Nbd7 6.Bd3 dxc4 7.Bxc4 b5 8.Bd3
a6 9.O-O c5 10.Re1 Qc7 11.h3 cxd4 12.exd4 Bb7 13.d5 O-O-O 14.dxe6 Nc5
15.e7 Rxd3 16.e8=Q+ Nxe8 17.Rxe8+ Rd8 18.Bf4 Rxe8
19.Bxc7 Kxc7 20.Nd5+ Kc8 21.Rc1Rd8 22.b4 Rxd5 23.Qe2 Bd6 24.bxc5

[d] 2k4r/1b3ppp/p2b4/1pPr4/8/5N1P/P3QPP1/2R3K1 b - - 0 24

the position is double edged with Black having defended nicely up to this point
a reasonable move such as 24..Bf4 would leave the game up for grabs with chances on both sides
sadly however the SB uncorks the honker..

24.. Bxc5 ??

allowing 25.Qe3 which just wins a pieces for nothing

[d]2k4r/1b3ppp/p7/1pbr4/8/4QN1P/P4PP1/2R3K1 b - - 0 25

25..Rd1+ 26.Kh2 Rxc1 27.Qxc1 Kd7 28.Qxc5 Re8 29.Ne5+ Ke6 30.Nd3 Kd7 31.Qa7 Kc8 32.Nc5 Re7 33.Nxa6 Kd8 34.Qb6+ Ke8 35.Nc7+Rxc7 36.Qxc7 Bd5 37.Qe5+ Be6 38.Qxb5+ Ke7 39.a4 f5 40.a5 h5 41.a6 Bd7 42.Qxd7+ Kxd7 43.a7 Ke7 44.a8=Q


in the second game i was black and the SB stayed in book until move 8

Game 2:

[[White "Star Beryl"]
[Black "Steve B"]
[TimeControl "90Sec/Avg"]
[Result "0-1"]

1.d4 d5 2.Nf3 e6 3.c4 Nf6 4.Bg5 Be7 5.e3 O-O 6.Nc3 Nbd7 7.Bd3 h6 8.Bxf6
Nxf6 9.O-O Re8 10.Qb3 c6 11.Bc2 Bd6 12.Rad1 e5 13.cxd5 e4 14.dxc6 exf3
15.cxb7 Bxh2+ 16.Kxh2 Ng4+ 17.Kg1 Qh4 18.b8=Q Rxb8 19.Qxb8 Bd7


[d] 1Q2r1k1/p2b1pp1/7p/8/3P2nq/2N1Pp2/PPB2PP1/3R1RK1 w - - 0 20

in this game i attacked the SB widely sacking pieces left and right to see what kind of defensive ability it had
and up to this position it played quite well
i am down a rook and a pawn and i was about to resign when the SB pulled another rotten apple out of the barrel

20.Qd6 ???

almost any other Queen move (Qf4 or Qg3) and i would be writing this review singing the praises of its defensive ability

instead now i have 20..Re5!
[d] 6k1/p2b1pp1/3Q3p/4r3/3P2nq/2N1Pp2/PPB2PP1/3R1RK1 w - - 0 21
and the SB cannot avoid being mated

21.Qb8+ Be8 22.Qxe5 Nxe5 23.dxe5 Qg4 24.Bh7+ Kxh7
25.g3 Qh3 26.b4 Qg2# 0-1

In summary...

given the poor display screen design ,lack of information displayed by the computer and penchant to make wild game ending blunders i cannot recommend the SB for someone looking to play against it to improve their level of play
for a few dollars more you can get a much stronger computer with many more features

however for me personally .. this computer was a gift from a dear friend in order that i might wile away the hours while recuperating from recent surgery and as such..collector or no collector..player or no player...it will forever have a place in my collection as high as the rarest of my computers:

http://new.photos.yahoo.com/cpaa2@sbcgl ... 6891716/87

Many Thanks Fern Regards
Steve

[Georg Langrath]

I myself was also a little disappointed that the buttons were of some kind of very thin plastic that were rather loose, and also crunched when you pressed on it.
In a match against Mephisto ll, I gave Beryl Plus 15 seconds and Mephisto ll one second, the match demonstraded before in this forum. When I analysed the game in Fritz 10, then Fritz couldn't find any serious mistake from Beryl.
I am convinced that Beryl Plus has the same program as Novag Star Beryl. Never that Kittinger should try to polish a 4K program.

Georg

[Steve B]

I myself was also a little disappointed that the buttons were of some kind of very thin plastic that were rather loose, and also crunched when you pressed on it.
In a match against Mephisto ll, I gave Beryl Plus 15 seconds and Mephisto ll one second, the match demonstraded before in this forum. When I analysed the game in Fritz 10, then Fritz couldn't find any serious mistake from Beryl.
I am convinced that Beryl Plus has the same program as Novag Star Beryl. Never that Kittinger should try to polish a 4K program.

Georg

i am not sure what computer you are referring to Georg
do you have the Mephisto II or the Mephisto MMII?

Uncertain Regards
Steve

[Georg Langrath]

I myself was also a little disappointed that the buttons were of some kind of very thin plastic that were rather loose, and also crunched when you pressed on it.
In a match against Mephisto ll, I gave Beryl Plus 15 seconds and Mephisto ll one second, the match demonstraded before in this forum. When I analysed the game in Fritz 10, then Fritz couldn't find any serious mistake from Beryl.
I am convinced that Beryl Plus has the same program as Novag Star Beryl. Never that Kittinger should try to polish a 4K program.

Georg

i am not sure what computer you are referring to Georg
do you have the Mephisto II or the Mephisto MMII?

Uncertain Regards
Steve

Sorry. I meant MMll

Georg

[JonP01]

Hi Steve,

Well it is nice to get consensus of opinion regarding the baffling display issues on the Beryl style of housing. It's been mentioned by both Georg and myself, so it seems that just about anyone who uses this unit is going to have the same gripe. Novag really ought to fix this in a future production run. As I mentioned previously though, I think this is the second or possibly third Novag machine in this style of housing, so the question has to be asked as to why they did not fix the problem when the updated Star Beryl went into production.

As for the games you have illustrated, if I could play devil's advocate, those positions are of the type where a human can see instantly that the replies are losing, hower a computer would need to go the hard slog and calculate all the variations through - unless it had extreme intelligence in terms of chess knowledge (and I woud argue that you just can't / shouldn't necessarily put this knowledge into any program anyway - let alone a tiny 4K one, since there could be specific tactical continuations in similar sorts of scenarios where self-pinning or the threat of mate with delaying moves won't result in a future catastrophe and would actually result in a better position once the smoke clears.

If we look at your first example, one continuation might extend to a total of 10 ply before the material balance is sorted and white (to the machine at any rate) has a clearly won position at least in terms of material. In one continuation, after 10 ply, white has won the bishop and it's just rook against Queen with 4 pawns apiece. On the 9th ply though, it's still Bishop and Rook versus Queen. Of course, black has a terrible position even if the material balance stayed that way - and the undeveloped rook and centralised white Queen would have won any which way. But I would not hold any of above against a tiny 4K program. Remember that we have quoted horrible losing continuations against Star Ruby, where it uncorked an horrific reply that lost material on the 40 moves in 2 hour level. It was a similar situation - easy for a human to see the catastrophe at a glance without calculation, but beyond the Star Ruby's search horizon, or otherwise missed by it's selective search.

In the second position, again there are delaying moves that would eat into the search horizon. After Qd6 the actual calculation required is fairly complex for a basic computer as compared to the human. Even the mighty HIARCS running on my Athlon 3800+ takes about 3 to 4 seconds to definitely score Re5 as winning. That's 3 to 4 seconds on a CPU running at 2400 Mhz that includes major archtectural advantages over the tiny processor within Beryl, let alone the 300 times increase in clock speed (it's probably thousands of times as fast in reality). So again, although it looks terrible for the machine, at least I can understand the blunder. I don't mind a machine blundering if I can get into it's mind and understand why it made the mistake. In both the above cases, as horrible as the moves might have seemed, I can see and understand what went through Beryl's mind. I would imagine that Beryl's prgram decided that any of those Queen moves were "OK" but it came to the decision that Qd6 kept the Queen more active and centralised than moving it to f4 or g3, for example. Perhaps the selective search completely missed the rook move altogether, especially as it puts the rook en prise.

But you are right to put the Beryl into context. It looks like my enthusiastic ravings here have raised expectations as to what the little machine is all about. It's certainly not about playing strength. It's all about the astonishing achievement of cramming such a relatively decent program into 4K whist including such a broad openings book, surpirsing endgame knowledge and - in overall terms - a reasonable game of chess. I'm even thinking of trying to re-acquire the old Scisys Companion II program so that I can play a 10 or 20 game match between the two. It would seem a conclusive way to determine if the Beryl really is the World 4K Champion or not


No Beryl Muddle here Regards


Jonathan

[Georg Langrath]

Hi Steve,

Well it is nice to get consensus of opinion regarding the baffling display issues on the Beryl style of housing. It's been mentioned by both Georg and myself, so it seems that just about anyone who uses this unit is going to have the same gripe. Novag really ought to fix this in a future production run. As I mentioned previously though, I think this is the second or possibly third Novag machine in this style of housing, so the question has to be asked as to why they did not fix the problem when the updated Star Beryl went into production.

Jonathan

What makes thing worse is of course that you must look at display each move from Beryl as it has no lamps that show moves. That would make the quality of display even more important.

In a better house Beryl would be a good choice for a player under 1500 in ELO I think. I am rather sure that a computer plays better against a human than a machine. The reason is easy. The computer sees every single bigger blunder not beyond horizon, a man doesn't. Even grandmasters have put their pieces in no forced situations, so defender can mate in one move or two. Much more can a 1400 player oversee such a blunder. Even the simplest computer sees that. That is a truth, not a belief.
In the match against Mephisto MM2 I showed before, I think Beryl played at a level of at least 1600 in ELO, even if I gave Beryl handicap. And an analyse with Fritz 10 showed that Beryl had won in positionell understanding, not because Mephisto MMll made big blunders. Truth is that Mephisto made one bigger blunder, but there Beryl had already won. That positionell understanding i amazing for a 4k computer, I think.

As I mentioned before even Eric Hallsworth gave Beryl a good judgment, and he knows what he is talking about.

Georg

[Steve B]

In the second position, again there are delaying moves that would eat into the search horizon. After Qd6 the actual calculation required is fairly complex
Jonathan

hi Jon

well not exactly..another 30 seconds of searching and the SBeryl given two minutes a move(rather then the 90 seconds i was playing on) did play Qf4
if the SB had pondering enabled i think alot of these game losing blunders would be avoided
and i dont think enabling pondering requires a larger program size
but i could be wrong about that

while i say again as a COLLECTOR the SB is certainly a collectors item ..as a computer to buy and play against..at $100 bucks..sorry there are just better computers to buy for not very much more money


Pondering Power Regards
Steve

[JonP01]

Hi Steve,

Well the Scysis Companion II program had pondering capability, and it was only 4K. As you know, the earlier Sensor Chess (with or without modules), did not ponder either. I'm not sure why pondering was not built into Beryl. It is a pity - it certainly would have improved the playing strength as you say.

[fern]

Seems like avoiding ponder you save life of the processing unit. With it On the processor never stand still. So companies get lot less complains.

Electric engineer ferdinan

[Georg Langrath]

I don't even think that the goal is to get it as strong as possible.

Georg