Measuring the resolution of TORUS
How many shape characters does TORUS need to identify tunes? Currently it uses an average of 13.0 characters, although the number needed varies between 7 and 26.
Looking in more detail, suppose we pick tunes in the database uniformly at random and give TORUS the first k characters of their shape strings, for some fixed k. Then we can define two possible measures:
- the uniqueness probability = the probability that a tune is identified uniquely;
- the resolution = the reciprocal of the expected number of matches found.
We can express each of these measures as a number between 0 and 1: the higher the value, the better TORUS is at identifying tunes by using k shape characters. We can then plot these measures against k to get an idea of the number of shape characters we might need to use:

The graph above suggests some puzzles: What shape should we expect the curves to be? Do they approach each other as the number of tunes becomes large?