Taking the concept from chess to chess960 might be easier to discuss by way of an example. I used the dice-rolling method described in A Database of Chess960 Start Positions, and selected a random start position. It is shown in the following diagram (SP201: QNRBBKNR).
Start Position 201
I then looked at the first moves played by the CCRL chess960 engines (see the link in the sidebar and my previous post Chess960 Opening Theory) in this position. Since the CCRL insists on playing chess960 with 'Book learning: Off for all engines', the engines play openings according to how various positional principles -- the center, open lines, piece activity, etc. -- are weighted by the software. This guarantees variety throughout the games.
A glance at the diagram is enough to see that in SP201 the first moves conforming to positional principles are similar to the choices in traditional chess (SP518: RNBQKBNR):
- One of the four central Pawns can advance two squares (1.c4, 1.d4, 1.e4, & 1.f4), or
- the Knights can jump to the same natural development squares as in SP518 (1.Nc3 & 1.Nf3).
24 x 1.Nf3
16 x 1.d4
14 x 1.c4
7 x 1.f4
6 x 1.d3
3 x 1.Nc3
2 x 1.h4
Five of the six moves predicted above are in the list, most notably the first four. Only 1.e4 is missing. There might be a tactical reason for not choosing 1.e4 (attacks by ...g6, ...g5, or ...Nf6?), but I don't see anything that is not easily parried.
Also in the list are two moves not predicted above: 1.d3 and 1.h4. Both moves could therefore be classified as extravagant openings. The move 1.d3 has the same drawback as 1.d3 or 1.e3 in traditional chess; it fails to contest the center. The move 1.h4, one of the worst first moves in SP518, looks just as bad in SP201; it fails to do anything useful except liberate the Rook on a1, the same Rook that White needs to castle O-O.
Also worth noting is that the six games starting 1.d3 resulted in a +3-2=1 edge for White, while 1.h4 resulted in +1-0=1. Just as in life, extravagance in chess isn't necessarily a doomed cause. We would all be a lot poorer if it were.