I missed the original message from Robbie Rhodes in AMD 95.11.01 quoted
by dick@via.at (Claus Kucher), mentioning Artis Wodehouse's interest in
"randomizing" :

> the timing of a heavily-edited piano roll performance in order to
> make it sound more like a live performance.

The only respectable ambition here must be to replicate from profound
knowledge of Jelly Roll Morton's entire recording resource appropriate
attack characteristics to be imposed on a modern performance based on
over-edited piano rolls.

Randomization strikes me as a quite inappropriate approach.  A person
who has already been doing this in his own replications of Jelly Roll
Morton recordings for over 15 years is pianist/roll arranger John
Farrell who is entirely capable of playing a whole JRM piece end to end
note for note from memory.  He started in Jazzmaster days cutting his
chords square, but things are different now on Hot Piano Classics.

Pianists develop attack styles as an integral part of their art and it
is almost insulting to suggest that "naturalization" by random note
attack variation would be an adequate representation of what they do.
I have been slowly making rolls of lesser-known classical music, using
records for artistic guidance, and the order in which notes are put
down in a chord is crucial.  In my case I have to gerrymander things to
enable the Themodist "snakebite" system to bring out the note I want,
but since I am trying to produce a generalized, optimum result I have
to average between more than one performance.  In practice, because I am
a dowser (water-diviner who finds, and finds out, other things using
the same technique) I usually arrive at an final attack pattern using
dowsing - and after a few false starts, I haven't had to change a
single one.  The rolls sound like a pianist playing and not a machine.
I'll predict that randomization will merely sound like a beginner.

Dan Wilson