I have made this point several times on other ET threads but see no harm in repeating here. Most people totally fail to grasp it. If a strategy works well over a few months of data from one asset class it means absolutely nothing. All you're doing statistically is fixing your data. Try backtesting over 30 years, using as many liquid markets as you can. Can you achieve a good expected return with low standard deviation?
I downloaded Yahoo EOD data of approximately 6 years and tested it with the parameters that he provided in Excel. It was not easy to do as he used all the US stocks. The results were not even close.. Don't think I was doing anything wrong as his rules were not complicated and I checked it a lot of times..
I agree, 30 years do not make a lot of sense as the volume and vola is 10 times more nowadays. We use 10 years. Sharp should be good and the strategy should work in 9 out of 10 stocks.
A system should be able to adjust dynamically for vol/vola and thus work equally well, long term, on data from any period.