Something is bothering me with this day breakout test system - the smaller I put the stop loss, the better the results.
Best results with 5 pip stop loss - this is not realistic, 5 pips stop loss will hit many many times.

Any thoughts?

Thanks a lot. It was possible to stay with the bug with years.

[edit]
Let check again

Of course 1.004

[edit2]

There is an error in your exe . I'll post the update after 10min.

EDIT:
Update?...

Hahaha  - mising brackets. Seems a new update after 30 minutes.

[edited]
now is OK

            for (int iBar = nMA; iBar < Bars; iBar++)
            {
                afUpBand[iBar] = afMA[iBar] * (1 + fDeviation / 100);
                afDnBand[iBar] = afMA[iBar] * (1 - fDeviation / 100);
            }

expect it in the net:

Where is the fault?

EDIT:
Should be 1.004 and 0.996 I believe?

Yes, it is obvious that FSB is faster than others.

About the results when we have two signals in both side of the opening price. I develop this method and it is tested and optimized using shorter periods. It is almost 90% reliable.
You understand  I can't discuss it here. It is the core of the application.

I do agree, that in more elaborate systems, you only allow longs or shorts based on other indications, but when we deal with strategies that hold position for a bar or two, and only capitalizes on the "noise" then the direction is sometime of no importance.

Wouldnt you agree?

EDIT:
Sounds good Popov (didnt mean to rush you) - I am becoming hooked! my entire day was coffee -> forum postings -> system testing and back to 1....

Also, if FSB will someday support automatic usage of the higher resolution data, and you have up to 5 minute or 15 minute bars, than this will be perfect.

You refer to it as "use previous bar value" - I am simply always buying at next bar (calculate on this bar, and buy @ limit at Bar+1)

So we got at least one thing clear; the entry logic is:
if not BuyAtStop( Bar+1 ) then ShortAtStop( Bar+1 )

This, btw, may be problematic - since it favors bull markets.
If this logic is reversed, we may get different results - but I guess that the only way to truely solve it is to use data from higher resolution.

Now, you touched the point nicely, and I would like to discuss more:
Talking about SL/PT - what would be the logic (in pseudo code) for that?

You mention that you interpolate the bar to determine what hits first - judging by the charts, it looks like FSB is doing something like this (for long positions):

( Bar is any bar from the positoin entry bar onward, including the entry bar itself )

if( Bar is Rising ) then {
  if not SellAtLimit( Bar )
  then SellAtStop( Bar+1 )
}
else {  // Bar is falling
  if not SellAtStop( Bar )
  then SellAtLimit( Bar+1 )
}

Did I capture this somewhat correctly?
Also, how did you extract the 87% accuracy level?

It will be absolutely fantastic if FSB will use data from higher resolution to understand what hits first - this will take it to a whole new level.