In the experimental scenario several agents repeatedly invest in n (n ≥ 2) state-specific assets. The evolutionarily stable and equilibrium (Blume and Easley, 1992) portfolio for this situation requires to distribute funds according to the constant probabilities of the various states. The different treatments endow none, one, three, or all subjects in groups of eight investors each with probability information. Will investments follow the theoretical benchmark or the 1/n-heuristic of equal investments in all assets? Further, will agents with probability information be asked and paid for advice on how to invest? Although investment does not converge as predicted, portfolios of informed agents reflect the probabilities of states, and even uninformed agents do not invest according to the 1/n-heuristic. Advice is demanded and readily paid for. Surprisingly, clients do not always follow the recommendation. Competition among advisors reduces their fees as expected.