Cyed is completely wrong and his reasoning makes no sense at all. They can get lucky wins, but they will still demonstrate their poor understanding of the game when you consider the moves. It is very possible that somebody a couple of hundred points higher rated than yourself (probably the ideal amount) would be able to somewhat accurately assess your level of chess understanding.
I would go on to argue that correspondence chess is the exception to this being the same as assessing their likely rating. In these games, the amount of time and aid used has too great of a link to rating to accurately estimate it. Your chess understanding could be reasonably gauged though, possibly even from a single game.
Ratings systems can take a long time to stabilise on your 'true' rating since they purely look at each game in terms of win/loss/draw.
But a human (or engine) could probably watch you and immediately spot that you are a terrible/OK/good/expert player based on things like whether you make any blunders, if you miss obvious tactics, etc.
Clearly it would be unofficial but out of interest how accurate do you think a good chess tutor could be, based on playing you 1-3 games or looking at a handful of your games, at estimating your current rating?
I'm curious because my rating continues to trend upwards and while I think I improved a little, mainly this is just because it hasn't levelled out yet and I have no idea if I'm actually a 1400 player, or a 1600 player, or what.
My rating is pretty unimportant, but the scientist/geek in me is interested.