Is there anyway to model second level options data? We have been running a live algo for some time now that has incredible results on the back test, over 2000% percent during less than a 5yr period. That backtest though is using minute level data for options. When running live, we are using second level data for options, and we seem to be getting stopped out and algo is getting erroneous entries. Is there anyway to model second level data for options to backtest it? We're also open to running it live using the minute level data for options; but our results were worse than they are now that we're using second level data. How accurate are the backtests? Thank you. 

Maurice