Questions on Novag Obsidian and some strange behavior

[fourthirty]

Interesting
my manual is coded...85-661-000
manual copyright-2002
perhaps the Serial number can provide a clue...
133883

My manual is coded 85-661-004
manual copyright-2002
Serial number 133883

Sigh.

Features Removed Regards...

[JMark]

Interesting
my manual is coded...85-661-000
manual copyright-2002
perhaps the Serial number can provide a clue...
133883

My manual is coded 85-661-004
manual copyright-2002
Serial number 133883

Sigh.

Features Removed Regards...

HI, Now the big question...what else was changed on the obsidian? I'm sure Novag didn't just take away the 'Next Best' feature. I'm curious what they were tinkering around with. I think Steve is right, the serial numbers might shed some light. My serial number is 216670.

[Steve B]

I don't understand how this machine was released with the lack of variety in the opening responses. Love the design and gameplay of the Obsidian, but really hate the opening book play.

Do other Obsidian's respond in a similar manner? Any other thoughts on changes to default settings to "open" up the openings???

Havent checked this but i wonder what will happen if i follow the suggestion i made in this thread 4 years ago
using the NEXT BEST function during the opening to see if more variety is provided?

Hmmm Regards
Steve

[SirDave]

I don't understand how this machine was released with the lack of variety in the opening responses. Love the design and gameplay of the Obsidian, but really hate the opening book play.

Do other Obsidian's respond in a similar manner? Any other thoughts on changes to default settings to "open" up the openings???

Havent checked this but i wonder what will happen if i follow the suggestion i made in this thread 4 years ago
using the NEXT BEST function during the opening to see if more variety is provided?

Hmmm Regards
Steve

Hi Steve,

Fwiw: I addressed this question in my study of the Openings Diversity of 4 Novag Portables:

http://www.hiarcs.net/forums/viewtopic.php?t=5258

'Incidentally, a small test was also conducted using the Next Best key available with the Jade, Jade II and Amber. It appears that when it comes to openings, all the Next Best key does is act as a toggle: If E2E4 is entered, the Next Best key responds with either E7E5 or C7C5. This was true even in the case of the Jade which is interesting since in Normal mode it by far responds with C7C6 (Caro-Kann) over C7C5 as the second most frequent response to E2E4. In the case of D2D4, the Next Best key toggles between D7D5 and G8F6. In short, the Next Best key allows only two different responses to E2E4 and D2D4 and does not choose from a diverse menu of openings.'

While it can't be assumed that the same would be true of the Obsidian, I can't say for sure, but my guess is that it is.

[Cyberchess]

Greetings Hiarcs Forum Members.I have recently bought a Novag Obsidian chess computer , and I am rather displeased with its lack of variety openings.In response to the Move 1.d4 Obsidian always responds with d5

I decided to run a quick test this morning on the Obsidian's Black responses to White's opening move. It appears the RANDOM function does affect the opening response, but not necessarily in the way that I would expect. I played 20 opening sequences for each setting.

AFTER POWER ON @ DEFAULT SETTINGS (RANDOM = 0)
1.e4
1...e5 (20/20 100%)

then

2.Nf3
2...Nc6 (20/20 100%)

Also, played a Queen's pawn opening:
1.d4
2...d5 (20/20 100%)

RANDOM = 1
1.e4
1...e5 (20/20 100%)

No change in Obsidian's response.

RANDOM = 2
1.e4
1...c5 (18/20 90%)
1...Nc6 (1/20 5%)
1...c6 (1/20 5%)

Slight variety now, but 1...e5 is never played???

RANDOM = 3
1.e4
1...c5 (19/20 95%)
1...a6 (1/20 5%)

Again 1...e5 is never played???

LEVEL CHANGED TO FUN 1 (Fn1) & RANDOM = 0
1.e4
1...e5 (20/20 100%)

I don't understand how this machine was released with the lack of variety in the opening responses. Love the design and gameplay of the Obsidian, but really hate the opening book play.

Do other Obsidian's respond in a similar manner? Any other thoughts on changes to default settings to "open" up the openings???

Thanks for posting your results, Greg.

Personally, I would find the lack of opening variety far more disconcerting than the lack of the [Next Best] function. I was quite disappointed back in ’86 after purchasing a highly touted Fidelity Par Excellence and finding that it responded to 1.) e4 with 1.) …. e5 100% of the time unless an opening choice was specified beforehand. My Novag units, on the other hand, all had pretty decent opening variety as well as other desirable features.

Changing the level of randomness in order to have the machine vary its opening responses will result in more varied albeit weaker play when the machine is out of book. I have a hunch that all of the latter generation Novag units had some type of serious flaw. They may have laid off the engineering dept. in order to reduce production costs in a rapidly shrinking market. It’s really a pity.

Fondly reminiscing about the Novags of yore regards….
John

[SirDave]

Personally, I would find the lack of opening variety far more disconcerting than the lack of the [Next Best] function. I was quite disappointed back in ’86 after purchasing a highly touted Fidelity Par Excellence and finding that it responded to 1.) e4 with 1.) …. e5 100% of the time unless an opening choice was specified beforehand. My Novag units, on the other hand, all had pretty decent opening variety as well as other desirable features.

Changing the level of randomness in order to have the machine vary its opening responses will result in more varied albeit weaker play when the machine is out of book. I have a hunch that all of the latter generation Novag units had some type of serious flaw. They may have laid off the engineering dept. in order to reduce production costs in a rapidly shrinking market. It’s really a pity.

Fondly reminiscing about the Novags of yore regards….
John

I also have little interest in a board that has little or no openings diversity. In the study I mention above, I found that the Novag Jade II (or Zircon II) and Amber (or Emerald Classic Plus) had great diversity and the random feature was particularly effective in enhancing it.

Yet, the Obsidian which is/was the sequel to the Emerald Classic Plus is an openings diversity dud. It makes no sense. All Novag (or what was left of it) had to do was transfer practically the same motherboard & firmware to the new Obsidian external board.

[Monsieur Plastique]

They may have laid off the engineering dept. in order to reduce production costs in a rapidly shrinking market. It’s really a pity.

Fondly reminiscing about the Novags of yore regards….
John

It needs to be borne in mind that even before the time of Obsidians, Star Opals, etc, the actual programmers of the chess engines themselves likely had little to nothing to do with the products or the company producing them. And even in cases where they might have, in an historical context decent opening books were still (and always have been) a specialised discipline separate to the creation of an engine (though a book compiler needs to be aware of the program characteristics so as to avoid the machine "voluntarily" playing lines that weaken it).

There would not have been many programmers who both did the engine and the book, simply because books are an extremely time consuming, demanding discipline. And if they did, I'd bet those books are not exactly the best or most enjoyable to play against. When I was doing the opening book for the aborted Nintendo DSi effort, for example, I spent many months full time on it, of which half of that was plugging holes, managing transpositions and achieving the desired level of variety. I remember spending literally one entire week just on the Kings Gambit and Latvian, so as to ensure that no expert in these openings could ever hope to get the better of the machine and that by the time the machine was thinking for itself, it was out of danger so to speak. I also had to run every single position through a multi-processor chess engine to ensure there were no tactical holes in the lines the program would play of it's own volition. If I had charged $60 an hour for my time, no one would have been able to afford it, let alone a company that is producing mass produced, Chinese plastic products to a shrinking market.

My belief is that once the original programmers were long gone (as is the case with Obsidian), companies such as Novag had engineering contractors who worked with the original code and modified it in certain respects such as level variety, feature tweaking, possible debugging (or introducing new bugs!) and engine tweaks that may or may not have improved play, etc. They would also have needed to modify the code to account for hardware improvements over time - due to changes in processor architecture and speeds. I am pretty sure, for example, that the 16K Novag program with that 8,900 book library has gone through a number of hardware improvements and subtle tweaking long after Kittinger had nothing to do with it, since the last of that generation plays noticeably stronger than the originals.

That being the case, these engineers likely faced the dilemma of force fitting newly compiled code plus an openings book into off the shelf chips or perhaps needing to compile the whole thing from scratch. This in turn may have required books to be modified - perhaps culled in order to force fit them onto an off the shelf hardware solution. Perhaps rather than painstakingly weed out lines as we might have done as enthusiasts, they simply took to it with a virtual butcher's knife. What you see is the result. Alternatively, if the source of the original opening books was not available then when compiling the entire program it would have been necessary to produce a new book from scratch. Unless you are a knowledgeable chess player (and reasonably strong), have book experience and are prepared to spend a lot of time, then creating even a small(ish) book from scratch for a mid-level dedicated machine is still a massive undertaking - well beyond what Novag would have been capable of by even the late 90s.

The books in the last Novags all have the issue we see with the Obsidian (even the Citrine book has more holes than Swiss Cheese), save for the Star Opal book which I believe is an old one probably still done at the time when there was a considerable manual component involved (done by hand at is were). I say this because there are a number of buggy lines that come up if you play enough games, which suggests errors introduced by a hand made process.

That said, it is good to see some decent experiments into the Obsidian book. The book does not seem to be (quite) as fatally flawed as I had thought, though I would love to see what sort of variety we get after the first move. For example, how many Silicians would we get after 2. Nf3. The Star opal book does not have that much variety on move 1 but the more moves made in the game, the more the variety branches out. I'd hope this might be the case with the Obsidian, but like some others I've always been reluctant to go out and buy one on account of my requirement for a wide and "randomised" book. Still even with the experiments shown above, they certainly make me suspect my hypothesis above is correct - namely they no longer had original source, had to tweak the engine to fit with hardware and feature expectations and this in turn required a book to be made by non-experts from scratch.

[SirDave]

Sounds right to me.

[JMark]

...That said, it is good to see some decent experiments into the Obsidian book. The book does not seem to be (quite) as fatally flawed as I had thought

True and I like what I feel is an important point that Steve brought up, that each machine have their own little nuances. Realizing that truth I have become less critical and enjoy each different machine for what it is. The computers 'character' is the combination of its strengths and flaws.

[Reinfeld]

Loving this thread. Observations:

1. Very impressed by RadioSmall's little experiment.

2. Really enlightening from SirDave - I didn't know this was how the labor worked, and it's commendable. Now I want to go find a program he wrote:

I remember spending literally one entire week just on the Kings Gambit and Latvian, so as to ensure that no expert in these openings could ever hope to get the better of the machine and that by the time the machine was thinking for itself, it was out of danger so to speak. I also had to run every single position through a multi-processor chess engine to ensure there were no tactical holes in the lines the program would play of it's own volition.

3. Steve offers two solutions here:

the computer will take back its move and come up with another book move
also set the computer to Book Training Mode
when you get the Out Of Book Signal simply switch to player v player and input some more book lines..you might wind up back in its book
i haven't tried it but its worth a go

4. I remain curious about the effect of opening books on deds, and what that's worth in terms of strength. One experiment I've never tried is the naked engine - test them with books turned off.

5. I'm sort of struck by the unspoken standard that a computer MUST respond with variety straight out of the gate without prompting to be considered diverse. Evidently, some feel the experience is crummy if you have to manually tell it to play a Sicilian or a Benko, which never bothers me. I like doing that - I like probing the thing manually, to see how deep it goes into a line, how far the SirDaves in the world have gone, what their sources are, and play the game from there.

6. All this Obsidian scrutiny - has anyone looked for this sort of thing from the Morsch-GK 2100 group, which provided equally slim opening books? Or the Ex-Par Ex Fidelity variants with their bigger books? I detect little variety from either. Again, you've got to order them to play something different - but they still know what they're doing.

I've always assumed the deds play what they rate as the highest-percentage response to start. It doesn't mean they lack variety. You just have to make them show some leg.

- R.

[Larry]

Interesting
my manual has no such addendum and NEXT BEST works perfectly on mine
i guess there were two versions of the Obsidian released?
my manual is coded...85-661-000
manual copyright-2002
perhaps the Serial number can provide a clue...
133883

Steve

My Obsidian is the same as Steve's, ie, no addendum and the NEXT BEST
works fine, both in and out of book.
my manual is coded...85-661-000
my Serial number.....143887. Note this serial number is WAY higher than
Steve's, unless he read his wrong, in which case our Obsidians were only
four digits apart from each other.
L

[Steve B]

Interesting
my manual has no such addendum and NEXT BEST works perfectly on mine
i guess there were two versions of the Obsidian released?
my manual is coded...85-661-000
manual copyright-2002
perhaps the Serial number can provide a clue...
133883

Steve

My Obsidian is the same as Steve's, ie, no addendum and the NEXT BEST
works fine, both in and out of book.
my manual is coded...85-661-000
my Serial number.....143887. Note this serial number is WAY higher than
Steve's, unless he read his wrong, in which case our Obsidians were only
four digits apart from each other.
L

nope my serial number is as I wrote it
im thinking the first three digits show the increased serial number putting yours 10 above mine and Marks at 83 above mine
I doubt that hundreds of thousands of Obsidians were sold but im just guessing

Breakfast Serial Regards
Steve

[Steve B]

I also have little interest in a board that has little or no openings diversity.

...That said, it is good to see some decent experiments into the Obsidian book. The book does not seem to be (quite) as fatally flawed as I had thought

True and I like what I feel is an important point that Steve brought up, that each machine have their own little nuances. Realizing that truth I have become less critical and enjoy each different machine for what it is. The computers 'character' is the combination of its strengths and flaws.

Exactly

the nuances of the different models from all of the different companies is what makes collecting these babies interesting
god for bid each model was exactly the same in terms of programming and features

this model has a book that sucks while that model plays a great endgame
one model has a memory save feature while another has printer capability
I guess if some folks are only interested in owning a small handful of computers then these nuances matter but I think with the prices these things sell for these days we can all live with the weakness/strengths of the different models..in fact we should celebrate these differences

Vive La Différence Regards
Steve

[Cyberchess]

4. I remain curious about the effect of opening books on deds, and what that's worth in terms of strength. One experiment I've never tried is the naked engine - test them with books turned off.

5. I'm sort of struck by the unspoken standard that a computer MUST respond with variety straight out of the gate without prompting to be considered diverse. Evidently, some feel the experience is crummy if you have to manually tell it to play a Sicilian or a Benko, which never bothers me. I like doing that - I like probing the thing manually, to see how deep it goes into a line, how far the SirDaves in the world have gone, what their sources are, and play the game from there.

6. All this Obsidian scrutiny - has anyone looked for this sort of thing from the Morsch-GK 2100 group, which provided equally slim opening books? Or the Ex-Par Ex Fidelity variants with their bigger books? I detect little variety from either. Again, you've got to order them to play something different - but they still know what they're doing.

I've always assumed the deds play what they rate as the highest-percentage response to start. It doesn't mean they lack variety. You just have to make them show some leg.

- R.

I recently suggested this experiment to Katherine (appleshampogal) who now owns a Par Excellence unit. After disabling the opening book upon startup, I found, to my amazement, that the machine would correctly calculate the book moves of the Petrov Defense to a considerable depth.

Computers excel in the open, tactical positions that typically arise from the symmetrical king pawn openings, and I believe that many models were designed to exploit this fact for the benefit of performance and ELO rating. One of the nice attributes of the pre-Y2K Novag units was the inclusion of a [Tournament Mode] whereby the machine would offer no variety; instead opting for the most promising lines in its opening book.

The Franz Morsch derivatives (Centurion, R.S. Master Chess, Saitek/Mephisto Chess Challenger, etc.) unfortunately come with a woefully inadequate 6K opening book. To their credit, however, they offer a choice of [Tournament] [Active] [Passive] [Gambit] [Normal] modes of play, with the [Normal] option providing the most diversity in opening choices.

Out of book on move 5 regards….
John

[Cyberchess]

They may have laid off the engineering dept. in order to reduce production costs in a rapidly shrinking market. It’s really a pity.

Fondly reminiscing about the Novags of yore regards….
John

It needs to be borne in mind that even before the time of Obsidians, Star Opals, etc, the actual programmers of the chess engines themselves likely had little to nothing to do with the products or the company producing them. And even in cases where they might have, in an historical context decent opening books were still (and always have been) a specialised discipline separate to the creation of an engine (though a book compiler needs to be aware of the program characteristics so as to avoid the machine "voluntarily" playing lines that weaken it).

Yes, I wholeheartedly concur that a chess programmer should not be bogged down with the mundane task of opening book creation. After all, they need to concentrate their efforts on pruning techniques, evaluation weights, dynamic piece assessment and the like.

I recall how Marty Hirsch’s success in the early ‘90s was partially attributed to MChess’s cutting-edge opening book compiled by Sandro Necchi of Italy. Indeed, finding a good opening book author is an essential task for any PC or dedicated chess programmer. Back in the roaring ‘80s, rumors abounded of opening books that were specifically designed to exploit weaknesses in the opening books of leading competitors. This technique, of course, resulted in a plethora of misleading data as to which program was truly better.

Random chess tournaments regards,
John