There are a variety of styles to PM.
The quantitative style says that if you have accurate statistical analysis of how different assets are correlated and what they return, you can create a portfolio that gives you the maximum expected return for the level of risk you are willing to take. Of course, coming up with accurate statistical descriptions and prognoses is tricky, and figuring out how to combine known factors appropriately is really just an art, disguised by mathematics to look more scientific than it really is (I admire the math, but don't get fooled into thinking that math makes it scientific).
The qualitative style suggests that you can identify trends in business and the economy by putting together information about companies and countries and industries and other sources. Now this can be rigorous too (don't think that qualitative work is necessarily unrigorous just because there isn't a formula attached), but the balancing is done in ways that are harder to replicate precisely. Often this means that portfolios are closer to equal weighted.
Interestingly, Fabozzi points out in his book on equity modeling that optimized portfolios do not perform substantially better than equally weighted portfolios in most cases, and points out that the problem is that estimation errors in quantitative models tend to degrade their theoretically superior allocations. It boils down to the problem that although you can't predict what the next period's return will be, you still need to have an exactly accurate number for the *average* return (and standard deviation and correlations). In reality, you almost never get exact values, but estimates, and this is why equally weighted portfolios are often about as good, despite being theoretically inferior.