I wonder how we could empirically test this. We could see who makes more accurate predictions, but people without beliefs about something won't make predictions at all. That should probably count as a victory for wrong people, so long as they do better than chance.

We could also test how quickly people learn the correct theory. In both cases, I expect you'd see some truly deep errors which are worse than ignorance, but that on the whole people in error will do quite a lot better. Bad theories still often make good predictions, and it seems like it would be very hard, if not impossible, to explain a correct theory of physics to someone who has literally no beliefs about physics.

I'd put my money on people in error over the ignorant.