Interesting question!
There is of course a mathematical approach versus unknown (GTO). But since we can fairly safely assume villain is a fish, I think it is simpler to try to play exploitatively. What I mean is that one should keep the eyes open, and (perhaps more than usual) rely on reads based on relatively few hands. A seen showdown is worth a lot more than usual in case someone is off main track.
One approach would be to play very tight without any read. You loose a little EV here, but gain in terms of small variance. Once the read comes, adjust.
A different approach would be to be fairly loose (based on your observations in a post above about smartphones). Probably +EV, but high variance. Again, once the read comes, ...
These are just my thoughts. GTO is probably too hard to even try to approximate. Hell, even open sizes aren't nearly figured out. Yet still, the way to do it (approximate, that is) is probably to experiment with ranges on a trial and error basis. Takes time.