In NT7 due to lack of higher resolution backtesting we had to resort to adding additional bar series (eg. 1 tick or 1 range bar) and amend the strategy accordingly - this added unnecessary overhead on our development time and caused additional limitations. Personally I very frequently use MRO and LRO for my conditions but these 2 methods have a limitation - they don't work with multi time frame / multi instrument strategies. This forced me to delegate all my condition testing to custom indicator. Other issues with additional series include lack of some simple overloads with only BarsInProgress as parameter, eg. ExitLong(int barsInProgressIndex), etc.
So the problem with the assumption that we don't need higher resolution than 1 minute is that majority of people requesting higher resolution backtesting in the past are quite aware of the additional performance overhead since we already use additional series for execution in NT7 - only in a very cumbersome way - we were hoping to avoid with NT8. As it stands I'll have to keep doing it the same way as in NT7, so there will be no performance benefit for people like me anyway.
The only time I backtested on the orignal bars is to check if the strategy is functioning properly. For backtesting/optimization statistics I always used 1tick/1range for executions and like most people in this situation I am aware of the performance overhead.
So the artificial limitation, ie. deciding we won't need more than 1 minute is very disconcerting for those of us who looked forward for the higher resolution backtesting in NT8.
Sincerely
Greg
Comment