Or maybe there is a chance to implement dynamic spread? Had conversation by email with mr footon and he suggested that it can be done. We would need to import bid/ask data from broker and calculate bar spread.

After this backtest quality of 99.9% would have real meaning.

I tried to avoid big spread time, works partially well. But problem is with fixed take profit, it can not be filtered if used, and this strategy requires small one. Also the problem is that strategy results curves around that unrealistic price action caused by bigger spread not the real market.

Example of the problem and impact for large trade sample:

And about data, more the better. Planning to use at least 3-4 years for optimization and one year for unseen check. Multiply it with other pairs data and you will have long period ... I always try to have at least 50 trades per variable for meaningful backtest, also I always check with unseen data period to see if it pass.

Another thing what does not go away from my worried mind is how much impact this overnight spread has in all portfolio level. I have lots of strategies based on day opening. Will need to check it someday same way like in example above.

Some update for future searchers, who will wonder how to make mt4/5 backtest faster with FSB EAs.
My strategy trades only 3 hours in a day, and all other candles was not necessary except when needed to use exit strategy. For different strategies probably there is no way. I added in OnTick() before fsb code more checks, to check my trade-able hour and if there is already open trade, then execute all FSB EA code. If not, then skip it. So it made me about 7-8x faster.

void OnTick()
  {
   int x = TimeHour(TimeCurrent());
   if( x<MyHour && PositionsTotal()==0 ) {  return; } else 
   {
        ....   FSB code...
   }

Also there is comparison about my spoken overnight spread with mt5 and FSB on GBPAUD. Same data, same spread, same strategy, same settings. With lower spread pairs impact is lower.