[r-t] Blue Line Difficulty

[Graham John]

Pip, Don and Mark make some good points. The discussion on this topic
in 2005 made the points about people finding different things hard and
it being based upon what you know already, so I accept all that.
However, I think there can still be value in having a measure of blue
line complexity (perhaps a better term than difficulty). There are
over 20,000 named methods. If you are trying to search for a new
method to ring, wouldn't it be useful to be able to search for methods
with a similar complexity to Lessness, say? It would filter out all
the methods that are far too hard, or too easy.

Mark says:

> I think looking at the line as a whole, rather than
> trying to factor out specific aspects such as
> hunting, is more useful.

and

> Measuring the compressed size of the line is
> therefore the obvious place to start

I don't believe that this is that different than what I am proposing,
as a good compression algorithm would compress the hunting too.
Separately identifying individual pieces of work has other benefits,
as it would be possible to create a master dictionary of the pieces of
work and name them (dodge, point, fishtail, stedman, bedspring,
Yorkshire places, Cambridge frontwork etc). It would then be possible
to search methods for specific pieces of work, or identify methods
that have work that is unique across the whole collection of named
methods.

I can also easily eliminate the reversals and inversions, as Pip
suggests. This would reduce the number of unique pieces of work,
especially for double methods, although from a difficulty point of
view, some ringers find reversal/inversion easier to apply than
others.

I have been looking at eliminating repetition (a form of compression
that makes sense from a learning perspective) for scoring individual
pieces of work. For example to differentiate Derwent's 11 dodges at
either end of its frontwork from a similar frontwork with randomised
places.

Graham