Topic: Ambiguous Bars

I assume that the Acceptance Criteria for "Max Ambiguous Bars" should be changed according to a percentage of the total number of bars you are testing over?

What is an appropriate percentage of total bars to set this value at? Does the answer change when you are using only IS versus using both IS and OOS?

For example, if I generate over 100,000 bars, what would a good value be?


Re: Ambiguous Bars

Generally we want zero Ambiguous Bars because, depending on the strategy, sometimes even a smaller number of Ambiguous Bars leads to a big difference between the different interpolation methods.

The backtesting algorithm is designed produce lower output profit (provided that you use the Pessimistic method) and in most of the cases if a strategy shows a good backtest, it is reliable enough to be traded. However, it is possible to find a strategy that exploits the particular interpolation method in order to show an overestimated result. We detect such cases with the help of the Comparator.