Hello boza,
In my opinion, there is nothing wrong with curve fitting a system perfectly using in sample data. Just make sure once you have it curve fitted how you want, you test the system again using out of sample data. Split the sample date 50%. 50% data for in sample and another 50% data...