Harry Truman wrote:
Is there any way to factor in the length of the set lists? I think the predictability is higher as the set lists have gotten shorter, but.... that is due in no small part to the typical encores and the shorter set lists.
I'm not sure if PI values are influenced by the length of the set. In my list of the least predictable concerts many had relatively short sets - Supper Club, Toads Place.
It depends on how short and long sets are put together. There are two views:
(i) Bob first makes a core set from the tried and tested staples and then racks his brain for a few rarely played numbers which he adds to this core set. The longer the set, the more rarities.
(ii) He starts with the rare songs and then adds the staples to bump it up to whatever length suits the occasion. The longer the set, the more regulars.
If he works by (i) then short sets will have proportinately more staples and will accordingly have higher PI values. If he goes by method (ii), the opposite is true: it's the long set that will have the preponderance of staples and it will have the higher PI value.
My instinct is that he doesn't use either method consistently. Sometimes he'll base a short set from the rare songs, sometimes from the regulars.
Well there's one way to check my instinct: measure the correlation between set length and PI. If it's more or less zero, my instinct is right. I'll do this later today and get back to you with the result.
Harry Truman wrote:
1) Can we do this without the encores?
.
Yes. Encores mostly - though not always - consist of familiar songs. So if we ignore the encore, the PI value will be lower. I could calculate three values for each concert: one for the main set, one for the encore and one for the set overall. However I'm not that keen on this. For one thing I'd have to do a fair bit of work to get all these calculations done. Also it complicates the picture even more. I'm trying to keep it as simple as possible. Part of the inspiration for doing the project was to try to put some general framework on the NET which, with 2397 concerts to date, can be overwhelming to contemplate.
Harry Truman wrote:
2) Can we factor in the length of the set list?
As I've stated above I'm inclined to think set length does not affect the PI value. But if, when I do the correlation test, it comes out that there is a significant effect, then I'll give the matter some more thought. I'm reluctant to complicate what is already complicated.
Thanks for your interest.