Piano Rolls.. (recording of Debussy playing Debussy)

[violinist]

How reliable are these piano rolls in reproducing the original way the masters played?

I'm listening to a piano roll of Arabesque #1 (Debussy) played by Debussy.  It really sounds a bit disjointed....

I got it to try to study how this piece should be played, but I must say it wasn't what I had in mind, although some of it I like a lot and other parts, just plain weird.

[m]

To understand how piano rolls work look here:

https://www.pianostreet.com/smf/index.php/topic,13794.0.html

[Nordlys]

How reliable are these piano rolls in reproducing the original way the masters played?

I'm listening to a piano roll of Arabesque #1 (Debussy) played by Debussy. It really sounds a bit disjointed....

I got it to try to study how this piece should be played, but I must say it wasn't what I had in mind, although some of it I like a lot and other parts, just plain weird.

There must be a mistake here. Debussy didn't record the arabesque. Which CD do you have?

He did record several pieces on piano roll though, among them five preludes, and childrens corner. I have the Pierian CD, and it is very interesting to listen to. Like you said, not what one have in mind. But I find them far from unmusical. Sometimes they sound strange, sometimes a bit clumsily, but often very inventive ideas, and very alive, and also elegant. I can imagine Debussy did play like this. He was after all not really a concert pianist, and probably didn't have a polished technique.

Whether one really can trust these piano  rolls or not, is an important question. It is not an acoustic recording. At the very least, they were recorded on a different piano than the piano they are replayed on, and one can never be sure if Debussy really meant it to sound like this.
  But the reproducing system of Welte, which was the company that Debussy (and many other famous pianists) recorded for, was quite advanced. Generally the piano roll records the time distance between the notes, by plotting them on the roll, so the rhythm should be correct.
  Dynamics was more difficult, but Welte invented an impressive system. The dynamics of each note was actually recorded by a graph, but then the player piano had to reproduce it. The reproducing of the dynamics was done only approximately by a mecanical system, programmed by a technician. One limitation is that you get only the same loudness for all the notes in a chord (notes that are played together). You can't voice a chord. But crescendos and accents are reproduced.
  When it comes to rhythm, which should seemingly be easy to reproduce, it can apparently not be completely trusted. There was some uncertainties, especially after the process of making copies of the master roll. The holes in the roll resembles the way a computer works, but the accuracy was not like in a computer.
  Also, the tempo can in theory be set to any tempo, but Welte did have a standard tempo for their rolls, so it is probably correct.