I received this private message:

Your robots performance is too good to be true for me. Even it works half as good as the past performance, it is already excellent. Did you just back test from 2007-2011? Why don't you test for 10 or more years performance?

The main difficulty is that before July 2007, shorting was only accepted on an up-tick. This added a positive bias to the EV method and hence, the 20DMF was also showing a much more positive MF, simply because short players were refrained from pushing the price down.

Thanks to some friends from this board, I collected the minute data down to the decimamilization period (April 2001 I think). I could develop an "new" 20DMF market timing model (both for the upper and the lower panel), adapted to the market conditions at that time, but would a back-test using this adapted indicator be reliable in today's condition?

Instead of going further into that direction and spend time to produce something whose results would probably be useless, I preferred to start the development of the robots, which at least are practical in today's environment.



Pascal