I found out that backtesting for day trading to be extremely unreliable as a way to analyze a strategy.
I prefer to run my strategies in a representative week under Market Replay (500x speed) and do my analysis, coding, adjustments, settings, etc... then I increase the data set for 1 month, etc... and do the same.... of course this is all very consuming, but at least it will be 'closer' to what you would get in a live environment (still not the same of course...). Once the strategy looks good then you can run it for a period against live data in SIM, and then finally live.
Of course if someone found a way to make the Market Replay results very similar to NT backtesting I would love to see an example here on how you achieve that (as well as result of both - MR and backtest summary). In my experience the results were so different that were (at least for my complex strategies) considered close to garbage... That is the reason I much prefer to use NT to develop strategies after working with Tradestation for a while... I could not do that on Tradestation (TS does not have market replay capability unfortunately).