There are two kinds of randomness, one that is harmless and one that can hurt.
Knowing which one your investments contain is important.
For example, an active portfolio manager who invests in one hundred companies out of S&P500 and weights them proportional to their market capitalizations, might outperform the benchmark as he intends to, or might fail to do so. Assuming the portfolio manager does not take any systematic factor exposures, a failing strategy will be look like random noise. So an average return of a failing / random strategy will be benchmark minus some trading costs and fees. In such a case, randomness produces a predictable result that is close to the benchmark. This type of randomness is disappointing to investors, but is rather harmless.
In asset allocation, however, randomness can be much more harmful. A growing number of sub-asset classes are entering typical investment portfolios. Unlike the stock picking example above, where the investible universe is specified by a third party index provider, the sub-asset class universes are constructed by the portfolio managers themselves and can vary widely across firms. For example, some might select from individual MSCI countries while others just use MSCI World for all the equity exposure. Some might have half a dozen of fixed income sub-styles, while others just use the Barclays Agg index. Some might select various real sub-asset classes and styles of alternatives, while others might ignore those all together.
A random selection from widely different sets of sub-asset classes will produce portfolios that can potentially vary greatly from a simple 60/40 allocation. Making things even worse, is the fact that many of these alternative static asset allocations are backtested over a couple of decades of data, producing a false sense of safety. Static asset allocation backtests with short histories ignore the long term risks and returns of sub-asset classes and also make an unrealistic assumption that investors would have constructed such portfolios in the past. One can always find a time when adding MLPs, TIPs, Commodities, REITs, MBSs, Levered Loans, Liquid Alternatives, CTA’s, and Hedge Funds would have produced a better short term backtest. However, most of the time, going forward, the actual results contain little of the simulated asset allocation alpha and are much closer to randomness.
Unfortunately, randomness in asset allocations has the potential to be much more harmful than in stock selection. Random asset allocation portfolios (i/e the ones without asset allocation alpha), depend heavily on the original investment universe of the sub-asset classes from which the portfolios are constructed in the first place. The more diverse the universe, the more likelihood that randomness can take the portfolio on a statistically different path than a simple 60/40.
Investors should consider the impact of allowing advisors to freely select asset allocation universes for their portfolios. Of course, if an advisor has an edge / skill / alpha, then selecting asset classes from a large number of potential choices can contain additional benefits. However, as data clearly suggests, most active investing is much closer to throwing darts on the board. And throwing darts at dozens of sub-asset classes is not a great idea because in this case randomness might work against you.
In sum, if I had to make this choice, I would much rather let my 2-year old pick stocks from an S&P500 index, than select a random set of sub-asset classes from a long and growing list that is used in typical wealth management portfolios today.