Week 24 Day 3: Simple Beats Complex, Every Time

Warren Buffett bet $1 million that a simple S&P 500 index fund would beat a portfolio of hedge funds over 10 years (2008-2017). He won easily: the index returned 125% versus 36% for the hedge funds. The most complex, expensive investment strategies in the world lost to the simplest, cheapest one.

Why complexity loses: (1) Fees: hedge funds charge '2 and 20' (2% management fee + 20% of profits). If the fund returns 10%, you keep 6.4% after fees. The index fund at 0.03% lets you keep 9.97%. (2) Overconfidence: complex models give the illusion of control. LTCM (Long-Term Capital Management) had two Nobel laureate economists and blew up spectacularly in 1998, nearly collapsing the financial system. (3) Data mining: complex strategies often find patterns in historical data that do not repeat. This is 'overfitting' -- the model memorizes the past rather than understanding it. (4) Turnover and taxes: active strategies trade frequently, generating short-term capital gains taxed at ordinary income rates. (5) Behavioral complexity: complicated strategies are harder to stick with during drawdowns. The data: After fees, approximately 90% of active strategies underperform their benchmark over 15+ years (SPIVA). This is true for mutual funds, hedge funds, tactical allocation strategies, and options strategies. The few that outperform cannot be identified in advance. The simpler your strategy, the less can go wrong.

The underperformance of complexity has theoretical support in information theory and portfolio theory. The curse of dimensionality (Bellman, 1961) states that as the number of variables in an optimization problem increases, the amount of data required to estimate parameters accurately increases exponentially. Complex active strategies with many inputs (factor exposures, macro variables, sentiment indicators) require far more data than is available in financial markets (which offer approximately 100 years of monthly returns -- just 1,200 data points). The result: complex models overfit to noise. DeMiguel, Garlappi, and Uppal (2009) showed this formally: the simple 1/N equal-weight portfolio outperformed 14 different sophisticated optimization methods (including Markowitz mean-variance, Black-Litterman, and minimum variance) in out-of-sample tests using 25 years of data across 7 data sets. The estimation error in expected returns overwhelms any benefit of optimization. Markov and Lee (2019) extended this finding to factor-based strategies: simpler one-factor or two-factor models outperform multi-factor models in out-of-sample tests because additional factors add noise faster than signal. The conclusion aligns with Occam's Razor: the simplest model consistent with the evidence is most likely to be correct -- and in investing, that model is: own the market, keep costs low, stay the course.