(MONEY Magazine) – It's around two o'clock on a Friday afternoon in January, and things are looking chaotic on the floor of the New York Stock Exchange. The Dow is already down about 100 points and looks as if it could slide a lot more. But just two blocks away in the lower Manhattan offices of Chase Manhattan, the investment world appears a lot more manageable. Sue Feitelberg, one of the bank's financial consultants, sits with 40-year-old attorney Joel Moser and his accountant, Michael Abler. They huddle around Feitelberg's computer, building a portfolio of mutual funds to pay for college for Moser's two daughters, now 10 and 14. And they're doing it scientifically. Between sips from his can of Dr. Brown's cream soda, Moser answers questions about his tolerance for risk and runs down his finances in minute detail: annual income and investments, the value of his interest in his law firm, the size of his home mortgage and so on. Feitelberg plugs this information into her PC, which is equipped with an asset-allocation software program known as an optimizer, so named because it creates portfolios that aim for maximum return with the lowest possible volatility.

Two hours later, as the Dow is finishing up with a 288-point loss for the week, the mighty optimizer struts its stuff. The software crunches the data, the printer whirs, and voila! Within seconds Moser is looking at a pie chart with seven asset slices, each a different color, showing exactly how he can get the biggest gain with the smallest downside. Moser scans the rainbow of allocations, unsure for a moment whether he buys into the mix. "It's not something I would have come up with on my own," he says, especially the amounts in small and international stocks (15% each). "I probably would have done a lot more fixed-income and large-cap stock." Pause. "But this is good, this is exciting."

Exciting, indeed. In a market where tech stocks are imploding one day, emerging markets collapsing the next, most investors would kill for a road map to the land of maximum gain-minimum pain. That understandable yearning for certainty in an uncertain world is one reason the popularity of asset-allocation software has soared the past few years. It's almost impossible to do business with a major investment firm these days without being bombarded by literature showing that the key to superior returns isn't the specific securities you choose but the way you divvy up your money among a broad range of assets. But with so many investments to choose from, how do you come up with the right combination? Easy. The broker or planner or insurance agent revs up one of the optimizer programs developed by such respected investment research firms as Chicago's Ibbotson Associates, San Diego's Frontier Analytics or Wilson Associates International of Encino, Calif. Suddenly, treacherous markets seem more tame.

But for all this high-tech wizardry, do the portfolios created by such software live up to the precision implied by their glitzy graphics? There's no doubt that optimizers occupy the theoretical high ground. After all, the software is based on the ground-breaking mathematical formulas of Harry Markowitz, who showed that by combining highly volatile but loosely correlated assets--that is, ones that don't zig and zag in unison--you could create a portfolio that's less risky than the sum of its individual parts. So, in theory at least, optimizers offer the investing equivalent of a free lunch: appetizing gains with minimum heartburn.

And when it comes to actually assembling a portfolio, optimizers offer flexibility that's nothing short of astounding. Fire up Wilson Associates' Power Optimizer, for example, and you can mix and match as many as 300 different asset classes, including Russian stocks, self-storage reits and food-chain retailers, and then carve your portfolio into slices sized down to two decimal points: 11.13% in large-cap value stocks, 14.82% in foreign equities, 11.67% in world bonds. In some cases, optimizers even project maximum and minimum returns for the portfolios they recommend.

But there's one thing these can't provide: proof that they work in the real world. That's right. The software is great for showing you how different combinations of assets performed in the past. But little, if any, convincing evidence exists to show that actual portfolios created for individual investors by optimizers have outperformed portfolios built without fancy software. Markowitz stands firmly behind his theory (for which he won the 1990 Nobel Prize in economics) and continues to apply it in consulting work for institutional clients. But he readily agrees there's no quantitative way to prove these programs deliver superior results to individuals. "You're asking if, when real brokers do things for real clients, are they going to make the clients better off?" says Markowitz. "I don't know."

At least one asset-allocation expert is convinced that the opposite is true--that optimizer-generated portfolios are doomed to underperform in the real world where, as Long-Term Capital Management's Nobel laureates discovered last year, investments don't always behave according to theory. "The illusion of certainty that comes out of these programs is an extremely dangerous one," says Richard Michaud, author of Efficient Asset Management, a new book that criticizes optimizers as they're used today. "Dangerous because you think you have the right answer when it may be a lousy one, and other answers could be more appropriate." What's more, skeptics contend that optimizers are brandished more as a sales come-on than an investment tool. "Some advisers are saying, 'Hey, we use Harry Markowitz's Nobel-prizewinning theory,'" says Venice, Fla. financial planner Gerard Stellwagen. "If it's on the computer and refers to all sorts of academic research, it's a good way of getting somebody to buy."

Given the prevalence of this software--more than 25,000 copies have been sold to date--chances are you may be "optimized" at some point. For that reason alone, it pays to take a closer look at how these programs actually work and their potential pitfalls. Even if you're a staunch do-it-yourselfer, understanding optimizers' benefits and limitations can provide insights into how best to structure and tend to your own portfolio of investments.

PRECISION OR GUESSWORK?

Virtually all optimizers rely on historical data to arrive at the three key variables they use to create portfolios: an investment's expected return, its correlation (how much it moves in sync with other assets) and its standard deviation, which measures the short-term volatility of those returns. The goal is to create the ideal asset mix, offering the greatest potential gain with the least possible risk. Considering how crucial these factors are in arriving at portfolio nirvana, you'd figure there would be universal agreement on which historical statistics ought to be used. But you'd be wrong.

For its Portfolio Strategist optimizer, for example, Ibbotson draws on performance stats for U.S. stocks stretching back to 1926. Wilson Associates goes back only to 1962 for its Power Optimizer and makes 1968 the cutoff date for another of its optimizers, ramcap. Optimizers also use different periods for different asset classes. Ibbotson excludes Treasury bond data before 1970 because "deregulation of the bond markets in the '70s" has changed the relationship between Treasuries and other assets, says account manager Vito Schiro. Emerging markets' performance data typically run only from the mid- to late '80s, although Ibbotson's number crunchers simulate relationships between U.S. stocks and emerging markets back to 1926. (Never mind that some emerging market countries didn't even have stock markets as we know them then.)

Clearly, behind the facade of a highly scientific quantitative process, there's a whole lot of shaking and baking going on. Some critics say this process of simulating data and mixing and matching time periods makes optimizers more like Cuisinarts than computer programs. "These guys are making sausage," claims William Jahnke, a Larkspur, Calif. investment adviser. "They pick this period and pick that period, and you're left wondering how much confidence you should have with what they're doing. In my opinion, the answer is not a whole lot."

The ambiguity doesn't stop there. After all this data has been processed and blended, there's still the issue of which numbers actually get plugged into the optimizer. Most advisers likely just go with historical data. Ayco, the Albany, N.Y. financial planning firm that does asset allocation for Dreyfus LION account shareholders, uses historical returns provided by Ibbotson. "We don't feel we're market prognosticators or have a link into a crystal ball," says Michael Iacolucci, chief training and investment officer for Ayco's financial services department. Others combine the past and future. Chase Manhattan, for example, takes past performance figures from Standard & Poor's Micropal but factors in the investment outlook from the bank's economists. "The large majority is rear view, but we're also forward looking," says Peter Wall, who oversees the optimizer model for Chase Investment Services group. "I'd say we're 80-20."

The consensus among most experts, including Markowitz himself, is that unless you believe the past will repeat itself, projected returns are the only way to go. That makes sense, except predicting future returns isn't exactly a scientific exercise. One way or another, most optimizer companies extrapolate the past into the future to come up with their forecasts. And different optimizers may contain startlingly different projected returns.

Last year, for example, Wilson's Power Optimizer was projecting a 9.08% expected return for foreign stocks and 8.08% for small growth stocks. Frontier Analytic's Allocation Master, meanwhile, was forecasting a more bullish 11% for foreign shares and 12.25% for small-caps. In short, depending on which optimizer is used, who's running it and whether they're using historical data or forecasts, crucial numbers like expected returns and volatility can vary dramatically.

MISTAKE MAXIMIZERS

None of this would matter much if optimizers weren't sensitive to such variations. But optimizer advocates and critics agree that even a small change in variables such as the expected return or risk level can trigger huge shifts in the supposedly "ideal" mix. As the portfolios below illustrate, a single percentage-point change in just one variable--in this case, the expected return for asset classes--shifts the recommended allocation dramatically. "It's not just a question of garbage in, garbage out," says Michaud. "It's more like, a molehill of garbage in, a mountain of garbage out."

These wide swings result because optimizers are programmed to zero in on extremes. They look for assets that have big returns and high volatility but that don't move in lockstep--such as emerging markets and U.S. small stocks--and then load up the portfolio with them. Such ultravolatile combos work fine in theory. But things can get hairy in the real world, as owners of emerging markets funds discovered when their shares plummeted 39% from April to October last year at the same time that U.S. small-caps dropped roughly 24%. "The optimizer will always tend to favor the asset that has the highest return and lowest correlation," says Scott Leonard, a Santa Monica, Calif. planner. "But if that investment doesn't perform as predicted, then the program has compounded that mistake by overallocating you to that asset." Which is why Leonard and other advisers derisively refer to optimizers as mistake maximizers.

Optimizer companies are well aware of this tendency for the software to generate what they call "unstable" portfolios. But they offer an easy way around the problem: Just "constrain" the optimizer by setting a limit of, say, 10% for emerging markets or 30% for foreign stocks and small-caps. Sounds reasonable, except that constraining is really little more than a euphemism for imposing your own preferences. And once you do that, you're essentially subverting the idea of creating an "optimum" portfolio. "You end up designing a portfolio that intuitively makes sense to you, and then use the optimizer to prove it mathematically," says Leonard. "It's backward."

Once the optimizer has delivered its tidy little pie chart, it's simply a matter of translating the slices into specific investments--large-cap stocks or funds, intermediate bonds, foreign stocks and so forth. Sounds easy, but this key step can undermine the entire process. That's because the portfolio allocations reflect how asset classes--typically indexes or other benchmarks --interact with each other. But the specific investments you choose will likely behave differently than the index, which could throw off the symmetry of the portfolio. "There are all these shades of gray," says William Sharpe, who shared the 1990 Nobel economics award with Markowitz and another professor for work on risk theory. "You're fooling yourself if you think you're getting a perfect match."

To get an idea of just how difficult matchmaking can be, consider some of the fund recommendations made for the Prudential portfolio on page 134. For the international stocks portion, a Pru broker suggested Janus Worldwide, a top fund run by renowned manager Helen Young Hayes. No problem there, except that the fund doesn't limit itself to foreign stocks. At the end of December, for example, the fund had 28.6% of its assets in U.S. shares, which means the Pru portfolio would actually have less foreign and more U.S. exposure than the optimizer recommended. Meanwhile, even though Pru's allocation makes no mention of junk bonds, they may be in there too. How? The recommended fund for the long-term bond section of the portfolio--Invesco Select Income--can invest up to half of its assets in high-yield bonds, and had a 42% junk stake at the end of September. A Prudential spokesperson contended that such deviations should not have a material impact on the portfolio's long-term performance.

ALLOCATION IN THE REAL WORLD

Aware of optimizers' limitations, some firms, including Fidelity and Vanguard, have decided to create model portfolios that aren't built around optimizers. As part of its Personal Investments Review, for example, Fidelity takes clients through a risk-tolerance questionnaire that guides them to one of four target mixes of stocks, bonds and cash that vary from as little as 20% in stocks for conservative investors to 85% for aggressive investors. These allocations are based on Fidelity's analysis of how various asset mixes have performed in past bull and bear markets--without looking for an "ideal" risk-reward formula.

The fact is, there are no definitive answers; asset allocation is still as much art as science. With that caveat in mind, here are some guidelines for building your own portfolio with or without an optimizer.

--DON'T OBSESS OVER PRECISE PERCENTAGES. No software program and no person has enough insight into the future to say exactly what percentage of large- or small-caps or foreign stocks or bonds will offer the best blend of risk and return. "In this business, you're misleading yourself if you think you can come up with an answer down to the decimal point," says Van Harlow, head of Fidelity's investment advisory division. "If you're plus or minus 5% in any asset class, you're still going to be pretty close to on target."

--AVOID THE URGE TO FIDDLE WITH YOUR PORTFOLIO. Stick with your allocations once you've chosen them, unless your financial circumstances or the investing outlook changes drastically. That's especially true with optimizers. "Little wiggles in expected returns can lead to big wiggles in the portfolio," says Harry Markowitz, "and the transaction costs will eat you alive."

--KEEP IT SIMPLE. The more asset classes you add, the greater your chance of "operator error." For most of us, five to eight asset classes in three broad groups--U.S. stocks, foreign stocks and bonds--should provide adequate diversification without turning the portfolio into an unmanageable mess. Remember, you're not creating a portfolio to prove that you can juggle 20 mutual funds while balancing a copy of Markowitz's 384-page Portfolio Selection on your nose. Your aim is to build a mix of assets that will give you a high enough return to achieve your financial goals, yet provide protection during market setbacks. "I don't think people need computer programs to do that," says American Association of Individual Investors president John Markese. "I'd rather bet on a simple allocation."