The buy-side of the business (pension plans, mutual funds, hedge funds etc.) seems to still work with these kinds of algorithms. I would go out on a limb and suggest that the sell-side (investment banks etc) does not in any way use these algorithms in trading. The prevailing paradigm used on the sell side since the early 1970’s has been the no-arbitrage paradigm, although it has taken quite a knock in the last decade or so due to the acknowledgement of counterparty credit risk.
The great practical difficulty in using portfolio analysis algorithms is estimating the vector of means returns. Frequent Wilmott contributor Alan has made this point before. You get that wrong and everything else is wrong.
The Black Litterman extension of the 1950’s Markowitz framework introduced the idea of incorporating one’s views into the mean-variance framework (or at least I think it did, I have never worked on the buy-side and have not given it much if any attention). You might as well do this given the difficulty in estimating mean returns. That would probably be considered the state of the art, for what it is worth. There used to be a neat website named
www.blacklitterman.com that had related research papers and even rudimentary Excel models but it appears to have disappeared.