| |
BGonline.org Forums
Trad to Nactation conversion algorithm
Posted By: Nack Ballard In Response To: Trad to Nactation conversion algorithm (Matt Ryder)
Date: Wednesday, 9 February 2011, at 8:31 a.m.
It's the logic of making an "educated guess" that will be tricky.
If a given move is impossible (such as S when the remaining back checker is pinned behind a prime), it seems to me a plethora of other plays might be possible instead. I could run through all the legal characters in that case and suggest the user pick one (perhaps suggesting legal members of the same family as probable substitutions?)
That's way more complicated than necessary. The "educated guess" is simply from a set of instructions that you give it (or more likely that I suggest that you give it). For example: If V is illegal, try U. If B is illegal, try S. If e is illegal, try E. If A or W is illegal, do nothing. Most mistakes are fundamental aberrations of that type and even predictable.
The main point is that whatever is added, however basic and naive, is an improvement on no phase 2 algorithm at all.
But it would be very dangerous to automatically substitute a letter and hope that the "sequence will right itself". What if (say) the next couple of characters in the sequence can be played legally after either an e or E? The sequence might only "stop" or deviate after quite a few missteps downstream from the actual error (especially if the computer is allowed to make several such "educated guesses" before stopping).
That's possible, but there is still no danger. As I stated, the original mistake is earmarked, so if you choose you can always go directly to that character and ignore everything else after that, in which case you break even. Of course, you can do better because seeing the possible integration after the mis-character (which might already be fixed by the straight substitution) provides you with additional information, again: if you choose to look at it (which I'm convinced you should).
If you prefer, you can wait on phase 2 until you've discerned for yourself enough of a pattern in the errors that you are convinced a few simple substitutions will be of obvious value.
I vaguely remember that < and > have something to do with fanning with half a number?
< means enter one checker. > means enter both/all checkers.
Is # a variant of @?
Yes. # means "alt anchor." It's one of the least used characters.
I was labouring under the delusion that Q could occur in any of the quadrants. Upon rechecking your original tutorial, I see you've specifically defined it as "a play made entirely within the opponent’s inner board area (including her bar point)" But isn't this restriction somewhat artificially limiting? A character like this available across four quadrants, plus an extended family of variants, might assist in answering Timothy Chow's complaint that Nactation cannot describe this move.
Nactation can describe any legal move. Few people would want to learn how or to bother to nactate some of the worst plays imaginable in that position, but fortunately such a position would never arise in practice, and if it did, people would still be hitting if they can (presumably), which would make the cream (the capitals) rise to the top.
[Btw, the position that Tim originally posted, testing the limits of Nactation, is different from Zare's most-legal-play position posted by Ian later (which you just linked); not that it matters.]
A quadruple split (Q) rarely arises (even though its definition now includes the entire far side), and when it does some member of the U/V or R family will suffice. Hence, I added a doublet-only definition for Q, which covers certain multiple-quadrant possibilities. GYM and the underlined families cover the rest.
Nack
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.