This study compares the performance of a small portfolio to a benchmark portfolio that is assumed to be diversified. We develop a new performance index that is calculated, for any given test portfolio, as the ratio of a Sharpe-like measure of the test portfolio to a Sharpe-like measure of the benchmark portfolio. Full diversification can be said to have occurred when the value of the index reaches one. This new performance index accounts for risk, portfolio size, return, initial wealth invested, transaction costs, expense ratio and investment horizon length. We use repeated sampling with replacement, for each simulation, from the set of all S&P 500 firms' returns to create test portfolios. Portfolios are formed first using equal weights and then using market weights. Results show that the number of stocks that must be included in a diversified portfolio is usually much larger than the 10-40 stocks cited in previous literature. We also investigate how optimal portfolio size varies with initial wealth, transaction costs, expense ratio and investment horizon length.