First, modern, properly installed autopilots are very rugged. Many, many boats used them exclusively with no problems operating them many years. They steer unstable
racing machines around the southern ocean non-stop for months without problems. Those boats actively seek out the worst
weather.
Fishing boats in the Bering Straights and North Sea use some of the same systems available to cruising boats day in and day out for years without issue. It does matter which type of AP you
purchase, though. Some are deservingly bad, and any attempt to use full-time a tiller pilot or
wheel pilot will fail quickly.
Additionally, many boats traveling long distances simply cannot mount or use windvanes. Ever hear of a
catamaran sailing the world? How do they do it? How about power boats? Large monohulls?
Installation is crucial. Most autopilot failures are
installation failures because the owner or installer did not understand the forces involved and the physical tolerances necessary for operation. Improper installation breaks mechanical
parts and causes electronic components to
work beyond designed performance.
Installation is equally as crucial for windvanes, with the same results for poor installations (minus the
electronics part). The difference is that windvanes LOOK like they could tear your transom off and their installation is treated with the appropriate respect. In addition, a bad installation is immediately noticed in poor (or non) operation of a windvane and is corrected at once, where a poorly installed AP may operate for some time without noticeable problems before everything goes sproink.
If you are concerned about redundancy, an autopilot costs the same as a windvane. So an option is to have a spare autopilot in case the mounted one breaks. This, of course, assumes that you have the power to run an AP as a primary self-steering device and would be happy using one as a primary. If not, a windvane is a better choice for you.
The remaining concern would then be a complete and unrecoverable loss of your
electrical system. This is rare, the probability can be mitigated down close to non-existent with some planning, and if it did occur with many 100's of miles from nowhere, you will be hurting regardless of your self-steering choice.
As for Tony Gooch's article, it doesn't take much logic to see where his data are not comparable and his conclusions incorrect. For example, he claims that the better performance of windvane over his AP was due to his miles made good each day were better for the windvane because the windvane adjusted itself to windshifts while he was sleeping and not adjusting
sails. However, he either did not understand how to set his AP to
wind mode (it was available on the model he used), or chose not to do so for some reason. Either way, his conclusions are erroneous. Many of the rest of his conclusions are equally as easily dismissed.
Mark