Horizontal loop is my #1 choice. Even a low one (20' or so) will do well on most bands with, in my experience, less tuning effort than any dipole. At one wavelength it will be a bit of an NVIS - high angle - antenna, but on higher bands (more wavelengths of wire) the angles will come down nicely. Personally, I hope to go for 2 wavelengths on 160m on my next one. My last one actually tuned well all the way up to 2m (though it probably benefited from increased system loss up there), and even worked a bit on 80m even though it's fundamental resonance was around 60m.
Second choice (which I have now) is a fan dipole. It works great even though the ends of the 80m wire each hang straight down more than 20' due to space limitations. No tuner needed, as I tuned the wires carefully as I added them, starting with the 80m and working my way up. It now has 20, 40, and 80m elements, is up about 40', and is my most efficient antenna.
I also have an HF9v ground mounted with a dozen 30-60' radials. It works fine, but is best on 40m (and up) where it is nearly a full quarter wave in height. It's about $300, and tuning is finicky and time consuming - definitely requires a good antenna analyzer - but it works.
Another option, which I haven't tried, is a wire dipole for the lowest band of interest, but with stub-wires suspended vertically from it at points that help it resonate on other bands. I haven't tried it, though, and will look for other people's write-ups on it ... www.cebik.com?
I haven't seen much modeling info on this type yet, but am interested ...
Good rx & 73 - Tim, KT8K