[r-t] New plan
holroyd at math.ubc.ca
Tue Dec 29 00:32:12 UTC 2009
I wholehartedly agree with everything Mark says here. Two little things
- re 2, to my taste at least, _completely_ mixing up all the types of
music might also be a bad thing. It's nice when you have a block that
concentrates on one type of music for a while (perhaps with pthers thrown
in) - the band notices it more, and get to repeatedly practice striking
that type of run/whatever. As in any type of music, or art for that
matter, you want the right mix of predictability and unpredictability.
One thing I was wondering was: in some domain where a complete search is
possible, formulate some precise score function based on 1 & 2, and just
try maximising that...
On Sun, 27 Dec 2009, Mark Davies wrote:
> Ander writes,
>> Is it time to consider establishing some standard metrics for
>> music distribution?
>> Has anyone tried computer searches for the best composition
>> according to "distributional" criteria?
> This is a good idea. Most of my recent compositional effort has been focused
> on exactly this: getting the music nicely distributed throughout the peal. It
> seems to me there are three main things you want:
> 1. To avoid long "bad patches". Even if your peal contains a lot of music, it
> can be spoiled if this is concentrated in one part, with swathes of rubbish
> elsewhere. Music should be distributed throughout the composition. The "no
> duffers" concept is the beginnings of a way to measure this, but as Ander
> implies is not sufficient.
> 2. To mix different music types. Finishing with three Homes is dull, because
> the music is too samey; similarly, I don't much like cyclic compositions
> where the same cyclic runs appear in each part. Much better to have different
> types of music continually mixed up throughout the composition. This is
> harder to measure by machine.
> 3. The right music at the right time. I'm a firm believer that peals should
> be paced nicely, with music placed where it will be appreciated most.
> Stabilising courses at the beginning are good - surely only a computer would
> put the one tenors-split course in a peal at the start. Conversely, I
> sometimes like to put the plain course in the middle of a peal. It's usually
> the most musical course - put it somewhere where the ringing is likely to be
> at its best. Structural concerns come in here, too; for instance, I think it
> is lovely when the first part of a peal presages some music, which will
> return in a later part in a different variation, or extended to greater
> heights. That sort of thing is, to me, the hallmark of a truly exceptional
> There is no metric which can encapsulate the third rule, which is why I think
> the eye of a human composer is so important. But that's not to say we
> oughtn't to put some effort in to measure rules 1 and maybe 2.
> The way I currently work, to try and produce compositions which follow these
> rules, is something like this:
> a) I'll start on paper, looking for short blocks which pack in the music I
> like. I might use the computer to help generate these blocks, too, but it's
> very much a manual assessment process: have I got the beginnings of the music
> I want, threaded together in an efficient way.
> b) Next I'll try to link the blocks up, directly into a peal length if I can;
> into large blocks if a more complex, staged approach is required.
> Machine-linking of blocks is good, and will highlight the best ways of
> linking them in terms of absolute music counts.
> c) But, absolutely music counts are no use on their own. Many of the
> machine-generated linkages will be sub-optimal with respect to the Rules
> above. So lots of hand-sifting goes on here, to identify the compositions
> that have the best mix and positioning of musical sections.
> d) Often a given set of blocks cannot be linked nicely, so at any point I
> might need to head back to the drawing board (stage a) to look for other
> e) Finally, when publishing the finished composition(s), I will try and
> describe in words what I feel I have achieved. There aren't currently any
> good ways of measuring rules 2 and 3, so a description of how closely I've
> matched them is the best I can do.
> ringing-theory mailing list
> ringing-theory at bellringers.net
More information about the ringing-theory