I understand that AmiBroker runs so quickly because the indicator is calculated on a one time pass through the entire data set. Is it possible in AmiBroker to calculate an adaptive indicator, such as an adaptive RSI, using a dominant cycle calculation or volatility variable where the parameter within the data set is unknown in advance? An example would be John Ehler's MESA using 1/2 the dominant cycle length.
Ok, would like to ask one more clarifying question. Your answer implies that in the 2nd phase of portfolio simulation it is possible to create/liquidate positions on bars that have no entry/exit signals form the 1st phase. In extreme case all entry/exit signals can be overwritten by the decisions based on the portfolio analysis in the second phase. Am I correct ? A simple example would be that the VaR constraint and correlation analysis tells that the best decision is to add the instrument XX into a portfolio, despite that there are no buy signals at all for this instrument based on pure price/vol analysis. Or we understand in a different way trading a list of instruments as a portfolio with a goal to control portfolio risk, not only risk of a single trade. By the way, could you post previous link to the description of the objects available for the decision making at the portfolio level. Somehow it has disappeared.
I would like to get started in backtesting. Some of the things I would like to do are very simple, analyzing what if scenarios such as: what happens when the market goes up/down for 5 days in a row ? others I am not sure if they can be done easily ex:: what is the winning percentage of buying on Thursday's close and selling on Monday's close? Does the software know which day of the week or month each bar is ? How easy is it to do these things with Amibroker? I hear the programming isn't easy. to me, the problem of systems is they only look at prices to determine signals, there are other criteria that may be used for signals such as economic events and seasonality . Integrating those in your computerized system must be an almost impossible task for the average user. Also in my limited experience with basic scanners such as what is now Ameritrade Advanced Analyzer for ex. scanning for patterns is not very efficient because the program returns tons of results most of them I discard. Patterns are not well identified by computers it seems, what 's more a lot of the more effective patterns would be difficult to define even in plain English , unlike a basic H&S or double tops which don't really work anyway.