Personally I am not a fan if diversification. I am afraid this is a contrarian view. Most people and institutions will tell you never put all your eggs in the same basket. As if putting the eggs in six different baskets will protect them against a 7.0 Earthquake. It will not. However, that is not the point. The point is diversification is a good strategy and a safe strategy when executed correctly. However, before I get into more detail, let us discuss definition and pros and cons first.
Diversification in investing is a strategy where investors spread their money across a variety of investments, including different asset classes, industries, and geographies, to reduce the risk of losing a significant portion of their portfolio if one investment performs poorly.
Diversification have the following benefits:
Risk Reduction: Diversification lowers portfolio volatility. Modern Portfolio Theory (MPT) shows that combining assets with low correlation reduces unsystematic risk without sacrificing expected returns.
Consistency: Diversified portfolios tend to perform more stably across market cycles, avoiding catastrophic losses from concentrated bets.
Tailored Diversification: Within stocks, diversifying across sectors (e.g., tech, healthcare) or styles (growth vs. value) can capture upside while mitigating sector-specific downturns.
Arguments for Mediocrity:
Dilution of High Performers: By spreading investments across many assets, you may reduce the impact of top-performing stocks or sectors. For example, holding 50 stocks means a single stock’s 100% gain has less impact on your portfolio than if you held only 5 stocks.
Market-Like Returns: Broad diversification, such as through index funds, often mimics market performance (e.g., S&P 500). This can feel "mediocre" compared to concentrated bets that outperform during bull markets.
Over-Diversification: Studies suggest that beyond 20-30 stocks, additional diversification yields diminishing risk reduction. Holding too many assets may lead to average returns with unnecessary complexity.
The chart above is relatively simple to read. It shows two lines, risk and return.
X-Axis: Number of stocks in portfolio (e.g., 1, 5, 10, 20, 50).
Y-Axis: Portfolio risk (standard deviation of returns) or return (%).
The key thing to note from this plot that risk decreases rapidly from 1 to 20 stocks, then flattens. Returns stabilize near market average.
However, the next point is even more critical. Once you have about 20 stocks in a portfolio, increasing it to 50 has no effect.
We can increase the number of data point, and variable ranges, but the outcome will remain the same.
This plot is the main concept that led me to think that:
Diversification guarantees only one thing: mediocrity
Empirical Evidence
A 1986 study by Meir Statman found that 20-30 stocks provide most diversification benefits in terms of risk reduction.
The S&P 500 (highly diversified) has historically delivered ~10% annualized returns (1928-2024), competitive with many active strategies.
Concentrated portfolios (e.g., Warren Buffett’s Berkshire Hathaway) can outperform but carry higher volatility. For instance, Berkshire’s annualized return (1965-2024) is ~20%, but with significant drawdowns (e.g., -44% in 1998-2000).
These are all facts but still my hypothesis is not easy to test.
Seeking Alpha!
We arrived at Alpha. The elusive alpha, everyone is after it, but few can get it. It is also said that alpha is luck, and no one (or very few people) can consistently do it! So what is it?
Alpha in financial markets represents the excess return of an investment or portfolio relative to a benchmark, typically a market index like the S&P 500, adjusted for risk. It measures the value an investor or portfolio manager adds through active management, such as stock selection or market timing, beyond what would be expected from passive exposure to the market.
The chart above is just like the previous two charts, but with another green line added. This is the return of a portfolio that generated a higher return because of a positive alpha.
Yellow Line (Risk): Shows risk (standard deviation) decreasing from 40% (1 stock) to 15% (50 stocks) as diversification reduces unsystematic risk.
Red Line (Actual Return): Shows the portfolio’s hypothetical return, starting at 12% (1 stock) and converging to 10% (50 stocks), reflecting market-like returns with high diversification.
Green Line (Expected Return): Calculated using CAPM, it starts at 14% (beta = 1.5 for 1 stock) and converges to 10% (beta = 1.0 for 50 stocks). This represents the return expected given the portfolio’s risk exposure.
Alpha: The vertical gap between the red and green lines.
Notice, that 20 stocks, it is nearly impossible to generate alpha. So by definitely, if you seek alpha, you must run a concentrated non-diversified portfolio.
Real Data :)
What if I had a 5 stock portfolio over the last 10 years? I did. All these names are well known.
Yellow Line (Risk): Starts at 142.05% (NVDA’s risk for 1 stock), drops to 84.49% for the five-stock portfolio, and converges to 17.07% (SPY’s risk) at 500 stocks, showing diversification’s risk reduction.
Red Line (Actual Return): Starts at 111.64% (NVDA’s return), hits 78.62% for the five-stock portfolio, and converges to 14.00% (SPY’s return), reflecting the portfolio’s outperformance.
Green Line (Expected Return): Starts at 21.20% (beta = 1.6), hits 18.32% for the five-stock portfolio (beta = 1.36), and converges to 14.00% (beta = 1.0), based on CAPM.
Alpha: The gap between red and green lines.
At 5 stocks:
Actual Return: 78.62%
Expected Return: 18.32%
Alpha: 60.30% (significant outperformance, driven by NVDA’s high returns)
The concentrated portfolio (5 stocks) generates substantial alpha (60.30%) but with high risk (84.49%). As diversification increases, alpha approaches 0, and returns align with the market, confirming my earlier concern about diversification potentially leading to “mediocrity” (market-like returns).
Basically, you need one winner, that's all. Often the winner is well known, and it is no secret and in plain sight.
PS. I used numpy and matplotlib to generate these plots. Trying to have fun with python.
