How We Construct and Backtest our 50 Stock Portfolios

We construct portfolios beginning with a population of the top 2,000 U.S. stocks by market cap. After calculating expected returns on these, we rank the stocks from lowest to highest expected return.

Next we filter out stocks with a price under $5 per share or a market cap under $300M. In the tests shown here, we imposed no constraints for sector weighting.

From that remaining population, we select the best 50 stocks for each strategy (the highest expected return stocks for the long only strategies; highest and lowest ERs for a market neutral strategy).

After a three month period, on a rotating basis, we rebalance each portfolio. Based on newly generated expected returns, we replace up to 50 stocks (100%).

All backtests use date-stamped data to eliminate the possibility of look-ahead bias. We include a benchmark in each chart for comparison purposes.

Starting with 1996, at the beginning of each month we successively form 12 equally weighted portfolios (tranches). One begins in January, the 2nd in February, etc. The monthly returns for each tranche are linked and (for 1996) annualized. Those 12 annualized returns are then averaged, simulating a strategy whereby one invests equally in each tranche.

To see portfolio returns for each strategy, choose from the menu to the left.

For more details, click on the terms at the bottom of each chart.