I'd also point out that more advanced autopilots will have a yaw sensor, and can tell the rate at which the boat is turning, in addition to having a compass. This can make it far more usable.
As for having a windvane... most of the boats out there don't. Some don't due to issues with having a windvane work properly on them. This is often the case with multihulls and high-performance monohulls.
To be sure, there are many variables. I've found that looking the vane when the AP is strugling (my boat or others) will often reveal the problem; if the vane is swinging all over due to motion at the mast head, it will strugle. There is damping in the circutry, but only with limits.
How does mine perform?
* Flat water. Perfect, any point of sail, compass or wind. Sometimes I think it steers better than I can.
* Swell. Good up wind, OK off the wind but higher power consumption (hand steering is more work too). I can point better hand steering. I can fly the chute a bit better hand steering. Set sails more conservatily/over trim a bit if hard on the weather or running chute. Other courses, set normally.
It would be nice to have a simple button to increase the gain in swells, to help it anticipate the way a helmsman can.
But I have never found conditions, other than wing-and-wing in a blow, where I couldn't trust it for hours.
That said, my last boat was a performance cat (Stiletto 27). Would I trust the AP under sail? Nope. The boat was just too tender and the AP couldn't keep up unless the water was flat and the wind very steady. Of course, lashing the tiller would have been pittiful also. It was VERY dependable under power, though, in any conditions.