Why quants never trust diversification ? The formula in the video breaks portfolio variance into two terms. One shrinks as you add assets. The other doesn't. That second term is your correlation floor and no amount of diversification removes it. But the problem is that this floor moves, Longin and Solnik showed in 2001 that correlations between equity markets spike during downturns. The matrix you calibrated in calm times stops representing reality the moment you need it. Markowitz works until it doesn't. What I'm not covering in this video: Pearson correlation only sees average co-movement. Two assets can look independent for years and still crash together. That's tail dependence, and it's invisible to standard correlation. Copula models handle this but almost nobody uses them in basic portfolio construction. There are strategies designed around this. Risk parity, trend following, tail-risk hedging. They don't diversify across assets within a regime, they diversify across regimes. Different game entirely. I'll probably make a video about it! Also worth noting the formula shown uses equal weights. Real portfolios optimize weights, which can quietly concentrate risk even further. And the Longin-Solnik result is specifically on international equities. Bonds, commodities, alternatives each have their own regime dynamics. -> Markowitz (1952), Journal of Finance -> Longin & Solnik (2001), Journal of Finance. -> Ang & Chen (2002), Journal of Financial Economics. #finance #quant #trading #algotrading #stocks - @vince.quant