This should be configurable to apply on another data period. Generate on 2015- 2018 but retest on 2003 to 2018 on the same instrument. This speeds up the generation of new strategies I would say you could go thru 5 times more strategies during the same time period. I am traveling and are on a laptop testing SQ4 it has 4 cores an core i5. I get up to 220000 strategies per hour generating on 2015-2018 but only 34000 2003 to 2018 and the ones that would make it to retest with higher precision are much more likely to pass this if already pre-sorted on a shorter generation period on selected time frame. It seems like only 10-15 % make it to the higher precision TEST.
If unclear I meant the automatic cross checks should be configurable in regards to data history length used. Now it uses the period chosen for generation.