P123 screen performance

When I list the P123 screens using the performance tab I get performance numbers that don’t at all match the performance I get when I backtest the screens. How is the performance measured? For example, the CANSLIM screen’s 5-year performance is listed as 189.8%, but if I run a 5-year backtest on it I get a total return of 17.85%.

When you do a backtest of CANSLIM with dates 09/20/2006 - 09/28/2016 you will get that 5 Y performance. But notice the huge jump on 2/20/16 due entirely on one stock which jumped due to acquisition.

This screen is very sensitive to starting dates because the number of stocks that pass is very volatile: sometimes 20 stocks pass, other times 1 or none. Combine that with a 4 week rebalance and you end up with wildly volatile results simply by changing the start date. It’s just the way William O’Neil defined it, with specific rules like “EPS%ChgPQ > 25”, that cause this volatility.

Ah, that explains it nicely. Thanks a million.