No, I'm not a masochist. And I don't think I was being punished for wrongdoings in a past life (though you can never be sure about such things). I was happy to go, and not just to get a nifty T-shirt with a graph showing how $1 invested in stocks, bonds or T-bills would have grown since 1926. I went because I wanted to see what one of the country's most respected investment research firms had to say about a new breed of asset-allocation programs called "optimizers."

Optimizers are all the rage today (Ibbotson markets one itself). Brokers and financial planners rely on them, as do a growing number of Web-based services (including Financial Engines, which offers a version of its product at money.com), to create those ubiquitous multi-color pie charts showing how to divvy up your assets among a variety of asset classes. You can't beat the precision: You'll be told exactly how much of your portfolio should be allocated to large stocks, small stocks, foreign stocks, etc.

But many people, including myself, have serious doubts about whether this highly quantitative process produces portfolios that generate superior performance. And some critics (again include yours truly) believe you can sometimes be worse off with an "optimized" portfolio.

The theory behind optimizers is undeniably sound, based as it is on Nobel Laureaute Harry Markowitz's groundbreaking 1952 paper "Portfolio Selection." Markowitz showed that you can pinpoint a combination of assets that will give you the highest return for whatever level of risk you're willing to take. Markowitz also offered another intriguing--and counter-intuitive--insight, namely, that adding risky assets could actually lower the volatility of your portfolio, as long as the new assets were only loosely correlated with those you already owned (that is, they zigged while your other assets zagged).

Translating this theory to the real world is another matter. To pinpoint optimal combinations of assets, Markowitz needed to know the expected return for a group of assets, the volatility of each asset (as measured by standard deviation) and how much their returns moved in synch with one another (their coefficients of correlation). And that is exactly what planners and all these Web services need to make their optimizers work. Now, they know the value of those variables in the past, but the question is, Can they accurately forecast the future? That distinction is critical because the "optimal" allocations change drastically if they change those variables just a bit.

To illustrate that point, I asked one of the speakers at the Ibbotson seminar, Gerald Buetow, Jr., founder of BFRC Services, a Charlottsville, Va. financial consulting firm, to help me put together an "optimal" portfolio for a moderately aggressive investor. To keep things simple, we looked at just three assets--large-company U.S. stocks, foreign stocks and intermediate-term bonds. And just for the purposes of this example, we assumed that the future performance of these three assets would be the same as over the past 20 years. That gave us projected annual returns of 17.7% for large-stocks, 15.4% for foreign shares and 9.8% for bonds. (For the nitty gritty on such things as standard deviations and correlations, click here.

On the basis of those assumptions, the optimizer recommended "Optimal" Portfolio #1: 57.4% in large-company stocks; 30.2% in intermediate-term bonds; and 12.4% in foreign stocks.

Seems reasonable enough. But what if we changed our assumptions just a bit, lowering the return for large-cap stocks by a percentage point to 16.7% and boosting the return for foreign shares by one percentage point to 16.4%. Hardly a dramatic change (nor implausible given the huge run of large domestic stocks and the relative under-performance of foreign markets). But look what happens. The optimizer now recommends "Optimal" Portfolio #2, which allocates 25.8% of your money to foreign shares, more than twice as much as before.

This kind of instability is one of the major pitfalls of optimizers. The asset allocation solution you get is tailored to a very specific scenario of how assets will behave. If they behave exactly as projected, your portfolio should do fine. If one or more of the assets behaves somewhat differently (gee, given the track record for return projections on Wall Street is that a possibility?), you could have a portfolio doomed to under-perform.

There are a host of other problems. You've got to find the right securities to fill the slices of the pie. Sticking a small-cap growth fund into the small-cap growth slot might not work too well, for example, if the fund manager has a tendency to poach into the mid- or large-cap market when small stocks are in the doldrums. There's also the question of how often you should re-optimize your portfolio to reflect changing market conditions. Do it too often and taxes and trading costs could zap your returns. Wait too long and your optimal portfolio might be based on outdated information.

I'm happy to report that Buetow and the other experts at the Ibbotson seminars didn't try to gloss over optimizers' shortcomings. In fact, almost half a day was devoted to detailing these and other pitfalls and going over ways to deal with the inherent limitations of optimizers.

Clearly, asset allocation is a crucial component of any serious investing strategy, and optimizers can play a role in building a portfolio. But before I accepted the allocations recommended by any broker, planner or website, I'd want to know something about the assumptions that led to the allocations. What returns are being forecasted and why? How often does the forecast change? What assumptions are being made about volatility and correlation between assets?

Most important, you should also consider whether any portfolio you're considering passes the sniff test. Optimizers love putting together mixes of high-return, high-risk assets that aren't highly correlated. But is it reasonable to put 30% your money in emerging markets or 40% in small growth stocks? Asset allocation software can help you see how various combinations of portfolios performed under very specific conditions in the past--and might perform in the future. But don't dismiss common sense. Otherwise, in my opinion, you're doing nothing more than making your investment decisions based on pies in the sky.