Systema/Krypton Challenge, Horvath and the Excalibur pit

[Reinfeld]

Friends,
So I snagged this little item today:

http://ebay.eu/1DXbkU7

As always, my motives were pure. I like to find the best available dedicateds by different programmers (excluding modules). I realized this Systema machine was a Gyula Horvath production. I felt certain I didn't own anything by that author, and the elo strength appeared to be decent - this one fits nicely in the 1800 zone - right in between Par Excellence, Super Constellation, Turbo King, etc.

I also knew nothing about Systema. Researching the board more deeply led me back (naturally) to Nick's site, which includes the following statement in the entry on the Systema Challenge and the Krypton Challenge:

Krypton Regency is the same as Krypton Regency, Krypton Comet, Systema Challenge and CXG Legend II. The software is also very closely related to Excalibur Ivan and Excalibur Igor*.

The Challenge also appeared to have a feature I've seen on no other dedicated (more on this later)**.

OK - so Systema and Krypton appear to be the same outfit, using different names in different countries, and Excalibur sold the units in the U.S. - correct me if I'm off.

The last bit in Nick's quote gave me a start, and then I remembered the Excalibur/Nelson debate, which I love, and it dawned on me that maybe I already owned some Horvath machines and hadn't realized it. Curse you, Nick!

I have a handful of Excaliburs, acquired earlier in my collecting forays. Most of them are crap, but I do have Igor (model 711E) and the GM***, which are respectable machines. I tend to assume they're the same, but I've never tried a hard test between the two.

Checking back on wiki-elo, I noticed that authorship on those last two gets a question mark. The entry used to say Nelson.

Yet the entries on variations of the Krypton Challenge leave no doubt that Horvath is the author. And Nick refers to them as "closely related" to the Krypton/Systema models.

With respect to all, the Nelson v Horvath debate matters less to me than the similarity of the machines. In that sense, it certainly seems that harder tests are in order. Nick included a useful rating excerpt from his suite of game tests in an earlier thread on this subject. I have edited it to include only the disputed models (minus tuning) and the so-called Nick-Elo rating:

Excalibur Igor-24 Mhz 30S Gyula Horvath/Ron Nelson - 1991
CXG Sphinx Comet Model CXG-902 Normal Gyula Horvath - 1988
Excalibur Glass Chess Normal Gyula Horvath/Ron Nelson - 1935
Excalibur Phantom Force Gyula Horvath/Ron Nelson - 1916
Excalibur Alexandra The Great Gyula Horvath/Ron Nelson - 1916
Excalibur Igor Gyula Horvath/Ron Nelson - 1875

Let's kick out the overclocked 24 Mhz Igor for starters. You still see an ELO variation - with machines tested against the same moves - of more than 100 points, which suggests that while these machines may be similar, they are not the same.

Also note that the Excalibur GM is not included here (maybe Nick added it later.) I don't know whether the Sphinx Comet is the equivalent of the Krypton/Systema Challenge, but those models don't appear in this group. Wiki-elo isn't particularly helpful here, either. Some entries provide no rating at all.

The authorship question is tough to solve. The similarity question (which might give us hints as to authorship) is not. What is the degree of similarity?

- R.

*Pretty hard to overlook the similarities in housing between Krypton Regency and Igor.

**The rare feature: tuning, or what they call the USER PROGRAMMABLE EVALUATION FUNCTION. You control 7 chessy factors (king attack, doubled pawns) that can be dialed from 0 to 100. I can't think of another dedicated model that allows this degree of control. Anyone else?

Note that the feature is unavailable in the Excalibur models (Igor and GM) that are part of the debate. So we have a distinct feature-based difference in the Challenge. It reminds me of the personality creation function in the Chessmaster series, albeit on a less sophisticated level, and it invites all sorts of amusing scenarios. For instance, is it possible to tune the Challenge to a higher level than its default strength?

***Karpov 2294 GM is the obvious early version of Obsidian (with a Kittinger program). Am I wrong?

[Reinfeld]

OK, I don't want to ruffle feathers, but here's my take on the Nelson/Horvath authorship question, and an archive of earlier statements from others:

A) Since Nelson said he programmed all the Excalibur models (undisputed), the only way for Nick to be right is if Nelson lied/distorted his role.

The charitable interpretation puts Nelson in the Barnes role, "director rather than composer," as another poster put it. But the issue still feels unsettled.

B) I don't entirely buy Nick's argument about job titles. Titles are just that - they don't convey reality, especially in companies catering to small audiences.

- R.

Below, for the sake of convenience, assorted quotes from recent threads:

All the Ron Nelson Excalibur programs are 32K ROM and the programs started off with 6502 processors at varying clock speeds back in the 73 level days (from memory 5 to 6 Mhz). By the time the program had been "refined" towards the end of the Excalibur days, it was still a 32K program but significantly stronger owing to the use of RISC processors and faster clock speeds, much more level variety, etc.
- Monsieur Plastique

Because of the Krypton heritage I lean towards Levy on the early weak computers, Danielsen on the improved computers ie KM3 & KA upwards as well as the LCD's and then of course Horvath on the Igor & Ivan's.
Nick

I think it is fair to say that the chess computer community is split down the middle on the question of who programmed those Excaliburs i.e the ones not obviously cloned and therefore known to be programmed by Kittinger, Taylor, Horvath or Morsch. So this debate is becoming a rerun of an earlier thread -
Mike Watters
http://www.hiarcs.net/forums/viewtopic. ... ght=#66538

Well Dave my theory is:

All the weak ones = Levy origin to about Kingmaster II
Mid Range = Danielsen origin
Top end = Horvath origin

Of course there is going to be some overlap between them as well.
Nick

If these Excalibur machines had all played a couple of classes stronger, I would seriously doubt that Nelson had anything to do with them. But their playing strength just in itself is perhaps the strongest evidence you can have, assuming you are not swayed by the communications from the man himself claiming authorship. It doesn't matter if you are talking about the Fidelity Challenger 8, the earliest King Arthur or the latest Alexandra: the relative capabilities, character and deficiencies of these programs in comparison to their contemporaries is always the same.
MP

I know when I was in contact with him(around 2001 ) he certainly seemed to be involved heavily in programming..
I recall and have posted before in another thread that I noticed a bug in the LCD chess computer when it was first released ..it would not promote to a Knight
when I alerted the Company to this fact I was surprised to get a response back from Nelson Himself..he complained that "Marketing" had rushed him to get the computer out fast and he fixed the bug in later releases

My guess is that at the time the later Excalibur computers were released (say from 1997 onward)Nelson was totally free to work on programming all of the time
Steve

My few tests with the Igor and the Krypton Regency (Horvath) show a clear difference in playing style and the moves. So, actually I'm not convinced in this theory, but as mentioned above - not impossible.

Of course it's possible too that Nelson created a totally new program for the H8 processor.
Fluppio

It is clear that CXG Accolade, Comet, Legend II and Krypton Challenge were Horvath engines. Though it wouldn't surprise me if Levy were involved on the business side. He seemed to be an engine broker.
IanO (July 2014)

As for the GM, I have great difficulty believing it is Ron's work but am happy to stand corrected. The GM is considerably stronger than all other native Excalibur machines yet the difference in hardware between even an Alexandra and the GM is not huge. Why then, would a machine marketed well before the Alexandra play a far stronger game than the Alexandra when the hardware specification differences simply cannot account for that strength difference? They are both 32K programs, they both use an H8 processor and the GM is only about 20% faster. Nowhere near enough to account for almost 200 points difference in my own testing. So to me it has to come down to authorship.
MP (July 2014)

[Steve B]

**The rare feature: tuning, or what they call the USER PROGRAMMABLE EVALUATION FUNCTION. You control 7 chessy factors (king attack, doubled pawns) that can be dialed from 0 to 100. I can't think of another dedicated model that allows this degree of control. Anyone else?



***Karpov 2294 GM is the obvious early version of Obsidian (with a Kittinger program). Am I wrong?

Hi R

you might want to reconsider adding Lang to your ever growing collection
several of his programs have a similar feature whereby you can change the way the engine values(weights) its pawns and each piece
it will also allow you to alter the way the engine evaluates pawn structures ..etc.etc
but you are right.. this is not a feature commonly seen

as to the Karpov 2294...the engine is the Novag Emerald Classic Plus by
Kittinger

Gaining Weight Regards
Steve

[Steve B]

OK, I don't want to ruffle feathers, but here's my take on the Nelson/Horvath authorship question, and an archive of earlier statements from others:

A) Since Nelson said he programmed all the Excalibur models (undisputed), the only way for Nick to be right is if Nelson lied/distorted his role.

The charitable interpretation puts Nelson in the Barnes role, "director rather than composer," as another poster put it. But the issue still feels unsettled.

B) I don't entirely buy Nick's argument about job titles. Titles are just that - they don't convey reality, especially in companies catering to small audiences.

- R.

Well i am not going to be dragged into this debate again for the third time this year so my only post about this in this thread is a question to you..

its not exactly clear where you stand on the issue
care to opine either way?


Getting You On Record Regards
Steve

[Monsieur Plastique]

The charitable interpretation puts Nelson in the Barnes role, "director rather than composer," as another poster put it. But the issue still feels unsettled.

I thought the better analogy would be musician versus composer. It is clear to me at least that during the Saitek Golden Era post Mark V (i.e Kaplan), Craig Barnes was one of the main - if not the main - engine room men so to speak and it is likely that more of his code found it's way into more Scisys / Saitek chips than that of any other person working for them at the time.

I still maintain (and always will unless Nelson speaks for himself strapped to a Polygraph and contradicts me), that all of the programs from Excalibur (excluding clones of other brands such as Novag) are quite simply far too weak for the hardware with the sole exception of the Grandmaster and simply play a lot of dumb and completely pointless moves that used to characterise programs from the late 70s and very early 80s - characteristics that point to one programmer and one programmer only - Nelson - and no one else (sorry to be blunt but it is the truth as I see it).

Even the most Mickey Mouse of the other programmers could produce far better quality on the same hardware. EVERY other "name" professional programmer from the early to mid 80s onward is capable of producing a program around 200 ELO stronger on average than anything Nelson has ever created. That is why the GM sticks out above all the other native Excaliburs. The engine wasn't his work. It couldn't possibly be. It's like saying my chess rating is 1800 and then I go out and win a match against a FIDE Master. Impossible.

[spacious_mind]

Hi Reinfeld,

The Comet is probably a Danielsen and not Horvath.

Regards
Nick

[IanO]

Another way to compare engine authorship: level selection. Horvath's CXG Sphinx Accolade/Legend/Concerto and Krypton Regency/Challenge have the exact same 100 levels as the Excalibur Legend II/Ivan/Igor/Grandmaster/Mirage. This and the choice of LCD display, H8 processor, RAM, ROM makes it pretty clear that these Excaliburs also used Horvath's fine engine.

[Reinfeld]

Hi Reinfeld,

The Comet is probably a Danielsen and not Horvath.

Regards
Nick

Nick,
I relied on your list that labeled it as a Horvath. Was it a typo or has your view changed?

Steve said:

its not exactly clear where you stand on the issue
care to opine either way?

I lean Horvath, for a few reasons:

1) The strength gap between earlier and later Excaliburs, and the playing style of known Nelson machines, as pointed out by players who are stronger than I am.
2) The obvious similarity in housing/design between Krypton Regency (an undisputed Horvath) and Excalibur Igor (authorship disputed)
3) The business/marketing relationship between Excalibur and some of these other, less prominent entities.

The information conflicts - no way around that. If any of the Excaliburs are Horvaths, that would mean Nelson overstated his role when he spoke with you. That's not a comment on your veracity - but it might be a comment on his. He wouldn't be the first businessman to embellish.

The ultimate test is playing. Moves don't lie. I'm waiting for the Systema Challenge in the mail. When it gets here, I'll try matching it up to Igor and Excalibur GM, which (historically) have been listed as identical programs. If so, and if the Challenge comes from the same programmer, the similarity should carry over. If not, we're back to ambiguity.

- R.

[spacious_mind]

Hi Reinfeld,

The Comet is probably a Danielsen and not Horvath.

Regards
Nick

Nick,
I relied on your list that labeled it as a Horvath. Was it a typo or has your view changed?

The ultimate test is playing. Moves don't lie. I'm waiting for the Systema Challenge in the mail. When it gets here, I'll try matching it up to Igor and Excalibur GM, which (historically) have been listed as identical programs. If so, and if the Challenge comes from the same programmer, the similarity should carry over. If not, we're back to ambiguity.

- R.

Hi Reinfeld,

Correct at first I thought it was Horvath, especially since there are conflicting manuals. The Challenge manual also lists Comet on the front page. But the built in play styles resemble Danielsen and the moves tend to as well. Also the amount of levels on the actual Comet are much less.

You are going to be back to ambiguity quite soon after your tests. Igor and Ivan are very hard to pin down. But to me that doesn't mean support for Nelson at all. If you do the math this is the amount of Style settings that you can do with the Horvath's:

100,000,000,000,000 (One hundred thousand billion).

Which makes it pretty much impossible to match the computers, but I think this heightens the argument that it would be quite easy to change style settings, repackage and no one would be the wiser. Especially when you remove this option from your computers. Makes being caught almost impossible.

Best regards

Nick