I suspect, apart from Steve and Cyberchess, that I'm the only other member in the forum who possesses one of these babies - and I love it. I think it's one of the better training methods I've ever encountered. I
envy all the extra program packs Cyberchess has acquired. I've never seen them on ebay. I purchased my set at a garage sale, complete with the Pal Benko endorsement, and it came with three standard packs of program cards.
We had a discussion a while back on this, here:
http://www.hiarcs.net/forums/viewtopic. ... 1a74f7231c
As for testing the paper simulator against a dedicated, Cyberchess said this:
Just realized we can do this anyway. For training purposes I used to try to deduce the correct move from the remaining choices if I had picked an incorrect or equally viable move. Since the digital unit requires no training, we will simply input the correct move and proceed to the next array of choices. Come to think of it, this is probably how the analog unit was designed to be used anyway.
Hmmm.
At the PRO level, Cyberchess tests the user's strengths against GM-level games, just like the old Learn Chess from the Masters-style books. It suggests that top-level performance would hit a USCF rating between 2050 and 2200. Too many wrong choices, and you're prompted to lower the skill level. Make the right choices, and you're (supposedly) a master.
I haven't figured this out completely yet, but this much is clear:
1. Cyberchess gives you six possible move choices on each move.
2. Five are wrong - the score drops accordingly (on the wheel, depending on the chosen skill level).
3. Some of those wrong moves are actually deemed acceptable, and lead to plus scores on the Cyberchess wheel.
4. Only one move is "right" - that leads to the next move in the game in the program pack.
5. The program packs include verbal annotations to the move choices, with limited analysis. OK - at least you can see the general reasoning.
6. Opening theory has admittedly evolved since the 70s, but the programmed pack games do not start from the opening position - even when this device was created, the manufacturers understood the need to avoid booked-up opposition.
7. The thing measures the
relative strength of the human player, matched against a professional's moves, for better or worse. You're supposed to tune it down or up, depending on how your choices fare.
So, to play a game against a dedicated (or even something kickass strong like Rev II) doesn't strike me as that hard. You just let the dedicated make the "wrong" moves, and score accordingly, thus developing a Cyberchess score, with the understanding that it's not perfect. Would work like this:
1. If the dedicated chooses a move Cyberchess dislikes, mark the score accordingly.
2. Take back the dedicated's "wrong" move and enter the move Cyberchess prefers.
3. Repeat, scoring accordingly to the end.
It's certainly possible that the dedicated will continue to choose "wrong" moves, but it's not going to choose them every time, or even most of the time. Most of the time, the choices favored by Cyberchess will be decent, and the strong programs will favor them, though rank blunders are possible.
The key bit, from my standpoint, is checking the dedicated's overall score AFTER the test ends, by the simple expedient of running its choices against a good strong engine.
In other words, if you try a match, you're going to have a dedicated play against Cyberchess, and see whether it matches the simulation's top level. If it doesn't, that's the score according to Cyberchess. Period. Post-mortem analysis is the place to test the accuracy of the Cyberchess score. That means a two-part process, but it doesn't seem too cumbersome to me.
Either way, this is a wonderful thread. I appreciate all the background material I did not know, and I would pay real money for some of those additional program packs if I could find them.
- R.
"You have, let us say, a promising politician, a rising artist that you wish to destroy. Dagger or bomb are archaic and unreliable - but teach him, inoculate him with chess."
– H.G. Wells