Modern Portfolio Theory for Beginners: The Diversification Math

Modern Portfolio Theory (MPT) is one idea: combining assets that do not move in lockstep reduces risk without proportionally reducing return. The math is simple. The discipline is holding two assets when one of them is losing.

The two-asset intuition

Suppose asset A returns 10% per year with 20% volatility, and asset B returns 10% with 20% volatility. Hold 50/50:

• If correlation is +1: portfolio volatility is 20%. Diversification does nothing.
• If correlation is 0: portfolio volatility is 14.1%. Same return, 30% less risk.
• If correlation is -1: portfolio volatility is 0%. Risk eliminated.

This is the entire engine of MPT. Correlation, not the number of assets, is what creates diversification. Ten tech stocks correlated at 0.9 diversify almost nothing; two assets correlated at 0.2 diversify a lot.

Why return does not fall with risk

When you add a volatile asset to a portfolio, the portfolio's return is the weighted average of the two returns. But volatility is not the weighted average — it is reduced by any correlation below 1. The free lunch is that you keep the average return while the risk drops.

The beginner's practical takeaway

Three rules capture 90% of MPT for a working trader:

1. Mix low-correlation return streams. A trend strategy and a mean-reversion strategy, run together, have lower combined volatility than either alone — if their returns are not correlated.
2. Add assets that feel uncomfortable. The asset dragging your portfolio today is probably the one providing diversification. Selling it locks in the correlation at the worst moment.
3. Weight by risk, not by capital. A 20%-vol strategy and a 5%-vol strategy at equal capital mean the loud one dominates the portfolio. Equalize risk, not dollars.

Where the intuition breaks

Correlation is not constant. The 0.2 correlation you measured in calm markets becomes 0.8 in a crash — exactly when you need diversification. This is why MPT is a compass, not a recipe: it tells you the direction (diversify across uncorrelated streams) but not the exact weights, because the inputs themselves move.

The bottom line for beginners

You do not need the optimizer to use MPT. You need to (a) hold multiple return streams whose edges come from different sources, (b) check that their returns are not highly correlated, and (c) resist the urge to cut the stream that is currently underperforming. The math says diversification works; the hard part is behaving as if it does when one position is bleeding.