I.
The efficient market hypothesis says that you can't pick out which stocks are undervalued versus which are overvalued. Likewise, I claim that you can't pick out which restaurants are underpriced versus which restaurants are overpriced.

Think you've found a great company, so that their stock will outperform on a risk-adjusted basis? Nope, someone else has already incorporated that information into the stock price and pushed the price up.

Think you've found a great restaurant which offers meals at a decent price? Nope, they've already raised their prices to the point where the extra cost just equals the extra utility you get from their extra delicious cuisine.

II.
A. But, first of all, we need to emphasize that this is on a risk-adjusted basis. A portfolio of stocks might have higher expected returns – but only if it's riskier.

This applies to restaurants as well to stocks – trying a new exotic cuisine could be eye-opening and awesome, or awful. Admittedly, this is quantitatively much less important for restaurants.

(This is the essence of modern asset pricing theory.)

B. Similarly to stocks, fund managers will not consistently deliver alpha to their investors: if any manager can consistently deliver alpha, that manager will simply raise their fees to capture it for themselves. (This is the essence of the "rational model of active management" model of Berk and Green 2004.)

III.
Moreover, second of all, cheap restaurants and cheap managers might exist, but they can have very high search costs.

Truly great cheap restaurants might exist, but you have to pay a lot in time, money, and energy spent searching and reading reviews to pinpoint them. These search costs, this time wasted digging around on Yelp, are real costs: they take time and money that you could otherwise have spent on better food or anything else which gives you utility.

This is likewise true of asset managers. Cheap asset managers that provide alpha might truly exist, but you have to spend so much time and money searching and evaluating potential such managers that these search costs will eat up that alpha. Otherwise, other investors would have already found the manager and grabbed that alpha.

(This is the essence of Garleanu and Pedersen's "Efficiently Inefficient" model.)

IV.
Third and finally: the utility of eating out at a restaurant is not just a result of tastiness and search costs. It incorporates every stream of services provided by the restaurant: convenience of location most of all, but also quality of service, ambience, and the social aspect of the other patrons. If a given restaurant achieves higher on these marks – e.g. a restaurant full of beautiful fashion models – then it should be expected that the quality of the food is less.

Similarly, to a lesser extent, with assets or with asset managers. Assets provide more than just a stream of returns: they provide the service of liquidity, or a "convenience yield". We can think of people enjoying the comfort provided by liquid assets, much like they enjoy the ambience of a nice restaurant. And just as a restaurant full of fashion models will – all else equal – have lower quality food, an asset or manager that offers higher liquidity should be expected to provide a lower pecuniary return.

(The idea of a convenience yield has been discussed by Cochrane, Koning, and others. This is also the entirety of the value behind cryptocurrencies.)

[Personal aside: This area is a core component of my own research agenda, as I currently envision it.]

V.
Conclusion: in equilibrium, assets or asset managers should not be undervalued or overvalued, on a risk-adjusted, fee-adjusted, search cost-adjusted, liquidity-adjusted basis. Likewise, in equilibrium, restaurants should not be underpriced or overpriced, once one takes into account their riskiness; the time spent searching for them on Yelp and reading reviews; and the ambience and other "convenience yield" services provided by the restaurant.