@Dionysus

Even you use past tick results from 2011 or back from 2006 there is no guarantee for walking forward similar results.
This is easy to explain as the market behaviour over these years was very different.
Only using the mass of past results and number crunching them tends to go curve fitting which is not the base
for a good automated system.
Better to know the traded instrument inside out and squeeze for some points in the detected good levels...

Highly possible that only 3 out of 100 auto systems do well enough bringing in the money in the long term.

GFIs1

GFIs1 

@Dionysus Even you use past tick results from 2011 or back from 2006 there is no guarantee for walking forward similar results. This is easy to explain as the market behaviour over these years was very different. Only using the mass of past results and number crunching them tends to go curve fitting which is not the base for a good automated system. Better to know the traded instrument inside out and squeeze for some points in the detected good levels... Highly possible that only 3 out of 100 auto systems do well enough bringing in the money in the long term. GFIs1

I couldn't possibly agree more, GFIs. . . a backtest is only a starting point, and one you should never place much faith and trust in, in and of itself.

Well said.

What i often see missing also, is that most automated systems function in a certain market
being able to detect that market is a key starting point

I don't believe there are (many) systems that work equally well in any type of market...

To start with, there is a console, that will trigger a system on/off
that console could be very well be a manual switch board