Two stocks with expected returns of 10% each, held in a 50/50 portfolio, produce a portfolio with an expected return of 10%. But if those stocks are imperfectly correlated (correlation < 1.0), the portfolio's volatility is lower than the average of the two stocks' individual volatilities. This is the core magic of diversification: risk reduction without sacrificing expected return. The lower the correlation between assets, the more risk reduction diversification achieves.

The mathematical reality: portfolio variance = w₁²σ₁² + w₂²σ₂² + 2w₁w₂σ₁σ₂ρ₁₂, where ρ is the correlation coefficient. When ρ = 1.0 (perfect correlation), the portfolio variance equals the weighted average variance — no diversification benefit. When ρ = 0 (zero correlation), portfolio variance is reduced by the cross-term. When ρ = -1.0 (perfect negative correlation), a theoretically perfectly hedged portfolio is possible. Real assets have correlations between 0 and 0.8, making diversification consistently valuable though never perfectly efficient.

Risk decomposes into two categories. Idiosyncratic (specific) risk is company-specific: a failed product launch, a CEO resignation, an accounting fraud, a patent loss. This risk is eliminated through diversification because company-specific events affect individual stocks but average out across a large portfolio. The research shows that 80-90% of individual stock idiosyncratic risk is eliminated with 20-25 randomly selected stocks across different industries.

Systematic (market) risk is the portion of risk that cannot be diversified away because it affects all securities simultaneously: recessions, interest rate changes, geopolitical crises, pandemics. Beta measures a stock's exposure to systematic risk. Because systematic risk cannot be eliminated, the capital markets offer a risk premium for bearing it — which is the theoretical justification for why equities earn higher expected returns than risk-free assets over time.

Diversification has diminishing returns. Adding the 6th stock to a 5-stock portfolio reduces risk substantially; adding the 101st stock to a 100-stock portfolio reduces risk negligibly. More importantly, over-diversification — owning so many positions that the portfolio converges to index performance — eliminates the possibility of outperforming the market from security selection. If you believe you have genuine investment insight about certain stocks, concentrating enough to let that insight matter is a feature, not a bug.

Sector and factor concentration can undermine apparent diversification. A 20-stock portfolio that is 100% large-cap US growth technology stocks is not diversified even though it has many positions — all positions share the same underlying risk factors. True diversification requires spreading across sectors, geographies, market capitalizations, and potentially investment styles (growth vs. value, momentum vs. mean-reversion). AIQ's portfolio analysis tools evaluate factor exposures across positions to identify hidden concentration that superficial count-based diversification metrics miss.