Mind blowing moves made by the Star Opal

[JMark]

Hi Monsieur, The more I read about the Star Opal the more I'm uncertain about its true abilities, At some times it seems presented as predictably poor and then in another post it shines with positive attributes. It actually won a game against the H8 Miami !!!!! Bizzare! At times you seem impressed with the Star Opal, unless I'm reading this wrong:

... the rating does not really tell the whole story about GM. I believe it is one of those machine like the current 16K Novags (Carnelian II and Star Opal). I believe these particular machines are stronger against humans than their computer versus computer raings would have us believe.]


Yet other sentiments include the following:

...it is safe for me to estimate the stock version at around 1250 USCF. This would make Star Opal possibly the most over-rated (by the manufacturer) dedicated chess computer ever released, since the difference between the manufacturer's estimate and my own is some 650 points. I cannot think of any other program where there is such a huge discrepancy]


One things is for sure. You must see some promise in the Star Opal, or at least at some point in time you did since you spent considerable time with it. My time with the Star Opal is too limited to evaluate other then to say it plays one heck of an un-boring game! Like you said, never sure what it may do, not necessarily just blunders but just a great game style, in my opinion. I forget that I'm playing a machine which has been my major problem with some computers.

The physical design is indeed beautiful. Playing with LED coordinates is so much more natural, I don't like the distraction of looking at an LCD. I hardly ever glance at one,(this is why I don't mind that I can hardly see the Obsidians without leaning over) So for me and my level of capability the Star Opal is a winner. I have not read much else about the Star Opal regarding it as a computer vs human opponent. If anyone else has evaluations with it let us know. For those advanced players, lets simply demand Novag produce the next generation Star Opal with the guts of the Sapphire II, that should satisfy us all!

Regards, John

PS- By the way, Like you, I have also stocked up on a few Advanced Travel sets! Now if only it had LED's!!!! I discovered too late a new looking Jade II recently sold on Ebay

The Star Opal actually surprised me a little in this brief encounter. I truly wasn't expecting it to win a single game, but it did all the right things to secure the win in the first encounter.

As for the other two games, when I tested the positions at 45 seconds per move, Star Opal made none of the blunders that marred it's play in the original games. I had always known this computer needs a lot of thinking time which was the reason for the change of crystal to begin with.

My feeling is that these two machines - were they to have a proper, long match - are roughly comparable in strength at faster time controls (<= one minute per move) but perhaps the 24 Mhz Star Opal might edge out slightly at slow tournament controls, though this would depend on the frequency of it's "inexplicable" blunders. Of course the only way to tell for sure is over a very long match carried out under very strict operating conditions. I have little doubt, however, that a stock Star Opal would be noticeably weaker against a Mark VI over the course of a formal match.

One thing I did note in the games was that the Mark VI was taking longer than the Star Opal to move in the openings (out of book), plus the middlegame, but sped up in the late middlegame and endgame. But in the second and third games, by then the Star Opal had lost positions.

In the first game, for example, total time - including operator time - used by the Mark VI was around 20 minutes and 30 seconds for 21 moves. In the second game, Mark VI took close to 46 minutes - including operator time - to make it's 40 moves. Obviously Steve would not have been trying to execute the moves in the same way one would under a strict, formal tournament test - nor was I for that matter - but notwithstanding this, the trend was as I noted above.

But Star Opal is still clearly a machine with a propensity to make ocassional bizarre moves - something which never effected any other mid-level (or better) program from Novag apart from the Dynamic "S" from around 1984 and , of course, the Star Opal's predecessors (though the Dynamic "S" was far, far worse than Star Opal in terms of blunders and oversights).

One thing that is clear is the Star Opal really needs to be used at the fixed time levels rather than the average time levels. My own experience suggests that this reduces the move randomisation in a given position by at least 50%, as well as reducing the possibility of poor moves because the computer takes a fraction of the time that it should.

Interestingly, there is actually not a whole lot of difference in timing between the fixed and variable levels in any case, since on the variable time levels, most of the time the deviation from the average time is not hugely significant. It's certainly not like the other Novags where a given move can take twice (or even more) the average time to make. The worst I have seen is maybe 30% more time on ocassions, but critically some moves are made in well under 50% of the average time, despite the fact that position really demands that the full time is used. And not surprisingly, these situations often marry up to where the Star Opal makes some of it's inexplicable blunders. There have been many positions I have tested where the Star Opal blunders on the variable time level, whilst not blundering on the comparable fixed time level.

It is clear that this 16K Star Opal program by necessity lacks the code sophistication found in other Novag 16K programs such as Primo and VIP. Afterall, the latter two programs have far smaller openings libraries (the Star Opal's being well over twice the size) and they also have much more RAM than Star Opal. Something had to give and I think these compromises relate more to chess knowledge and time allocation handling than they do pure tactical ability.

[SirDave]

Steve, looking at the photo, I think the little Star Opal was intimidated by the size of the V5. Try replaying the match with a partition separating them.

Size matters regards,
Dave

[Steve B]

Steve, looking at the photo, I think the little Star Opal was intimidated by the size of the V5. Try replaying the match with a partition separating them.

Size matters regards,
Dave

Better yet Dave
ill try another game with each computer in a separate room
Would do me some good to get the exercise

Running Back N Forth Regards
Steve

[Steve B]

I couldnt get the Star Opal to agree to play the V5 again but the Novag Super VIP(a good 100 Elo stronger then the Star Opal) eagerly agreed to give it a go
SVIP clocks in at 1684(Selective Search).. one has to admire the SVIP's spunk in agreeing to cross swords with the Mighty V5

same fixed depth 5 ply level as the last game
Super VIP had white..
V5 took about 30 seconds to complete a 5 ply search and the SVIP needed about 90 seconds
the game was pretty even until the SVIP went pawn snatching with its Q leaving its K open to a devastating attack


[Date "2011.10.8"]
[White "Novag Super VIP"]
[Black "Fidelity EAG V5"]
[Time Control "5 Ply Fixed Depth"]
[Result "0-1"]

1.c4 e5 2.e3 d5 3.cxd5 Qxd5 4.Nc3 Qa5 5.Nge2 Nf6 6.d4 exd4 7.exd4 Be7
8.Bd2 O-O 9.Nc1 Qb6 10.Nb3 Rd8 11.Bd3 Bg4 12.Ne2 Nc6 13.Bc3 Nb4 14.O-O
a5 15.h3 Be6 16.Bxb4 Bxb4 17.Qc2 Bd6 18.Rfe1 Bxb3 19.axb3 Rd7 20.Bf5 Re7
21.Bd3 Bb4 22.Rec1 Rae8 23.Qc4 Qd6 24.Qc2 g6 25.Bc4 Ne4
26.Bd3 c6 27.Nc3 Nf6 28.Rd1 Qf4 29.Ne2 Qh6 30.Nc3 Qh4 31.Na2 Bd6 32.Qc3
Qf4 33.g3 Qh6 34.Bf1 Ne4 35.Qxa5 Nxf2 36.Re1 Nxh3+ 37.Kg2 Ng5 38.Rxe7
Qh3+ 39.Kf2 Qxg3+ 40.Ke2 Rxe7+ 41.Kd1 Nf3 42.Qd8+ Kg7 43.Qxe7 Bxe7
44.Nc1 Qe1+ 45.Kc2 Nxd4+ 46.Kb1 Nxb3 47.Ra7 Qxc1+


FINAL POSITION
[fen]8/Rp2bpkp/2p3p1/8/8/1n6/1P6/1Kq2B2 w - - 0 48[/fen]
SVIP was showing -9.99 here so i resigned for it

The V5 could be the strongest 5 ply searcher ever released for sale regards
Steve

[Steve B]

The V5 could be the strongest 5 ply searcher ever released for sale regards
Steve

In order to test this theory....
we move up the Novag Portable food chain and next in line is the Jade I
Selective Search rates the J1 a good 60 elo higher then the SVIP clocking it at 1744

This game the V5 gets white..
same 5 ply fixed depth level for both
the J1 averaged about 20 seconds to finish a 5 ply search

in the game The V5 opened with my own favorite Trompowsky
J1 was out of book after 4 moves
J1 sacs a P early on to wreck the V5's pawn structure saddling it with 2 sets of doubled pawns and forcing it to castle long
faced with an irresistible Q side attack against its K the V5.. still up the one P.. is none the less forced to settle for a draw by 3 fold repetition

[Date "2011.10.9"]
[White "Fidelity EAG V5"]
[Black "Novag Jade I"]
[Time Control "5 Ply Fixed Depth]
[Result "1/2-1/2"]

1.d4 Nf6 2.Bg5 Ne4 3.Bh4 g5 4.f3 gxh4 5.fxe4 d6 6.Nf3 e6 7.e3 Bh6 8.Qd3
Nc6 9.Na3 h3 10.gxh3 Rg8 11.O-O-O Bd7 12.e5 Rg7 13.Kb1 dxe5 14.dxe5 Nb4
15.Qb3 Nd5 16.e4 Ne3 17.Rd3 Nxf1 18.Rxf1 Qc8
19.Qc4 Rg2
20.Rfd1 Bc6 21.h4 Rg4 22.Rg1 Rxg1+ 23.Nxg1 Bf4 24.Nf3 a6 25.Qd4 b6
26.Ng5 Bxg5 27.hxg5 Qb7 28.Rd1 Kf8 29.Qb4+ Ke8 30.Rd4 Rd8 31.Rxd8+ Kxd8
32.Nc4 Bb5 33.Nd2 c5 34.Qb3 Ke7 35.Qf3 Qd7 36.Qf4 Qd4 37.a3 c4 38.Ka2 c3
39.Nb3 Bc4 40.Qf6+ Ke8 41.Qh8+ Ke7 42.Qf6+ Ke8 43.Qh8+ Ke7 44.Qf6+

FINAL POSITION
[fen]8/4kp1p/pp2pQ2/4P1P1/2bqP3/PNp5/KPP4P/8 w - - 0 44[/fen]

Well..there goes that theory Regards
Steve

[Monsieur Plastique]

Jon (M.Plastique), I can wield a soldering iron. Is the 'upgrade' to 24mhz difficult to do? Is it something you might be able to give some instructions on?

Hi SirDave,

It is an extremely simple upgrade. The crystal is simply a through the hole mounting, so merely needs to be desoldered (I used a vaccum desolderer), then the new crsytal is soldered in place. So it really is an extremely basic operation.

You really just need to be careful to not solder for too long obviously, and to be careful when unplugging and replugging the ribbon cable between the mainboard and the sensor board.