[r-t] Method Generator

[Mark Davies]

Stephen Beckingham writes,

> Good question. A few months ago I ran a search using smc32,
 > on approx 3.5 million methods, to find the "most musical"
 > for a length of 1280. I'm looking to expand this to include
 > more methods

Ah yes, I remember this! Yes, the trouble is every time you relax a 
restriction, even by allowing just one new place notation (such as 78), 
the number of resulting methods increases exponentially.

However, if you are looking for musical compositions, you are likely to 
get the best results from musical methods with little falseness. You 
have to be a bit careful, because sometimes unexpected method designs 
can give amazing results, but in general if you are going to get lots of 
run-based music then I think you need a method which generates lots of 
the same type of runs in the plain course, in a way such that certain 
coursing orders are favoured. That however is no good if the favoured 
coursing orders are false against each other and so cannot be included 
in a composition, hence the internal falseness shouldn't be too high either.

It should be possible to use this kind of heuristic to limit the method 
search to a reasonable number of candidate results, even with the place 
notation restrictions completely relaxed. However it would not be 
completely straightforward, give the interaction between the music and 
the falseness and the need to prune section-by-section as the methods 
are constructed.

MBD