The problem seem to be very short files where naturally, the percentage of deviation is biggest in respect to the total length.
But do other programs do it better?
I took the value display from MP3tag, Foobar from the list, Foobar raw from the properties sheet and Windows Explorer.
Here a selection
| File | Mp3tag | Foobar | Foobar raw | Explorer |
|---|---|---|---|---|
| 1 | 00:02 | 00:02 | 0:02.250 (99 207 samples) | 00:02 |
| 2 | 00:04 | 00:04 | 0:04.213 (185 795 samples) | 00:04 |
| 3 | 00:03 | 00:03 | 0:02.913 (128 467 samples) | 00:03 |
| 4 | 00:03 | 00:03 | 0:02.740 (120 844 samples) | 00:02 |
In total the calculation is 116 thousandth of a second off the real value.
It looks as though MP3tag does not always round to the last whole second.
Actually, the MP3tag data matches fairly accurately that of the other 2 programs.
Except ... that the WE shows 1 second less for the last file.
So I am pretty sure that the more files you have, the closer the result will be to the actual time as it will level out on average.
So the statement
is not quite right as apparently other programs do exactly the same thing.