The article does not describe a fully defined trading method and I have to make some changes to create a method. For example, I see no mention of the risk level used when calculating position size, or if losses are stopped, or the magnitude of the loss when taken, or the initial account equity. or allowance for slippage and commissions. No mention of growth rates or draw downs. The authors provide observations and a summary of statistics. If you have any ideas about the design of the trading method then I am interested in hearing your ideas and possibly incorporating them into the software. I find test results change significantly depending on the number of consecutive lower closings used to generate a buy signal and the time constant of the moving average used to calculate the exit (sell) signal.
The article addresses equities. Not indexes. You can spend all the time you want dealing with SPY data, and it has nada to do with the point of the article. 14 trades over a period of 15 years is meaningless.
i did no overlapping and 100 shares at a time. if you have access to a F test you can check for mean reversion.
These are the results of trading simulations of General Motors symbol GM daily stock prices with a mean reversion system. Simulation is over a 44.55 year period. The following result is typical. I am not able to identify a combination of parameters that shows a profit. Moving Ave. Days 10 Days Closing Lower 5 Number Trades 139 Net Profit -43636 % Growth Rate -0.98 % Greatest Draw Down 35.08
Bank Of America hourly stock price data tested with the mean reversion system. None of these tests show profitable results.
Federal Express daily stock price mean reversion system test results. One test shows a small profit (1.26 % growth) but greatest draw down is 15.89 per cent. Method is tested with 26.31 years of data.
Bank Of America symbol BAC daily stock price mean reversion system test results. All tests show losses. Method is tested with 20.66 years of data.