Open yarcat opened 12 years ago
This kind of analysis looks a bit too complex to me. The easy way I see is to collect all partial productions and see what next elements could be there by composing all prefixes with all terminals and choose 3 most "popular" terminals.
The alternative could be a bit easier. It is to choose final formulas, collect their terminals, shuffle them randomly and put to a queue of next elements. It requires to extend registry, to let formulas known their terminals. It is less smart.
I'll try first version with the reduced amount of atoms. Maybe it will be easy enough. The analysis that I've mentioned is very simple! I'm not going to find all the compounds which are almost complete – just check what runes are on the screen, what probably could be built with them. The complexity is kinda linear from amount of possible compounds. It isn't complex at all. I'll create an experimental branch. We'll see what is it possible to get out of it.
I didn't mean to look for almost complete, just for most "popular" 3 continuations. But still it requires to find all existing productions plus to look through all terminals to get continuations.
Looking for "most popular continuations" take only prefixes standing next to a free cell.
Combine with formula shuffle.
Background
Currently we drop elements randomly, but even if we find good weights [#13], it might still be hard to play.
Proposal
Analyze elements on the field and drop new termins that most likely would generate compounds.
Questions