I take your point re clones v programmers. I spoke too broadly. But I still wonder. Some stray thoughts:Your statement would be acceptable for clones but for identifying same programmers it is not that simple and not really correct in all cases. Take the Horvath Test post for example that I posted and compare Test Game 1 Style 9 with Legend to the standard out of the box style 5. There are 17 move deviations out of 34 moves. That would make it a similarity of 50% coming out of the exact same computer by just changing the style. Based on your comments you would discount it as an absolute certainty that style 9 is a different programmer to style 5.The underlying point still feels right to me, however. A dedicated that matches moves less than 80 percent of the time with a dedicated of roughly equal strength doesn't feel like a clone.
- R.
I had posted these tests so that people can get a complete picture of the difficulty of these discussions.
- Nick
1. With respect to identifying clones, it seems to me that five factors matter, in descending order of relevance:
a. documentation/records (programmer is plainly identified)
b. moves
c. features/housing (not always decisive, but important)
d. Mhz/processing power (not always decisive, but important)
e. chess strength/rating (the least important factor in terms of similarity, because different programmers can reach similar levels of strength)
The original tests you ran on the GK 2100 variants are still the starting point for me when I think about these things. In that scenario, there was little doubt that Morsch was the programmer in question. The genesis/origin of RS 2250XL
was part of your inquiry, and you sought to determine whether it was a GK 2100 variant. You concluded that it was not. Later debate in this forum established (with some disagreement) that 2250 XL is indeed a Morsch machine, though not a direct GK 2100 variant. Its closest relative might be the TC 2100, if memory serves (I still can't find the final tallies from the rating-test thread.)
The difference with the Horvath/Nelson/Excalibur debate is a comparative lack of documentation. Only the Legend/Regency models have the clear Horvath stamp. The Excaliburs, like the Radio Shacks, suffer from limited documentation, which leaves the void and creates the speculative magnet Steve has mentioned.
2. The Excalibur models seem to come in three feature sets. I'm listing them from what appears to be weakest to strongest:
a. The 73-level models (King Arthur, Einstein and others)
b. The 136-level models (Ivan II, Alexandra, others)
c. The 100-level models (GM, Igor, Ivan, Mirage)
I can't find documentation for Avenger, so I don't know what the feature set entails. Others (here and on talkchess) say it's the handheld version of GM, and attribute it to Horvath.
Most of this debate surrounds the origins of the 100-level models, which are clearly the strongest players. There seems to be less interest in identifying the programmer(s) of the weaker models.
3. The 100-level Excalibur models bear more than a passing resemblance to the known Horvath models (Legend/Regency).
Nick suggests that the "tuneability" of the Excalibur models makes it harder to spot clones and identical programmers, and adds that tuning will make a dedicated behave and move very differently. I agree to a point, but it's also clear that we're looking at default settings. When I play Igor in these tests, I change nothing but time controls. Same goes for Legend. I assume that's true when Nick performs his tests.
Under those conditions, Legend and Igor/Mirage/GM (on default settings) are playing very different games. I admit it's hard to know, but is tuning the sole cause for this - or is it possible that different programmers are at work?
- R.