Hi, I am a systematic mechanical trader (almost exclusively forex). For the last couple of years I have been using MT4 to code and backtest my systems, but I am at the point where I need a lot more than what the MT4 backtester can offer. After looking at different options it appears that Amibroker is just what I am looking for, especially because of the execution speed, the possibility of executing walk-forward tests (a must) and the integrated 3D graphics (not a must, but nice to have).
I have downloaded and installed the evaluation version and have spent a few hours familiarizing myself with the program interface and -especially- looking into AFL. Unfortunately, I am having difficulties understanding a number of things about the language, or more accurately, about the whole philosophy behind the execution/flow of the backtests (I am a pretty decent C programmer, not a professional but I have been using it for nearly 20 years now, and from time to time my job also requires that I program algorithms in MATLAB, so I am familiar enough with programming languages, all sorts of arrays and matrices... the syntax of AFL, or the language in itself are not a problem). The first and most important question is this: how does the backtester operate to go through the historical data? How is the flow of data controlled? As important as this is, I don't seem to be able to find this information in the manual (I am sure it is there, but I haven't been able to find any section that addresses this specifically; maybe I need to look better, but after a while I was just getting frustrated). Let me be more specific: for example in MT4 the whole thing is event-driven, a tick (real or generated) arrives with a price update and triggers the execution of your code. When the program reaches the end, it just waits for the next tick. Next tick arrives, triggers next execution... and so on and so forth. Now, the AFL manual makes it very clear that in Amibroker the data is structured in arrays and it looks like everything is bar-based, but I haven't seen much about the specifics. So, lets imagine that my whole set of historical data is just a long vector of bar data (ok, several vectors, with the open, close, high, low, timestamps... data). What happens during a backtest? Does the program assign static indexes to the bars and then just runs through them, executing the whole code for each bar? If that is the case, is the first -oldest- bar always the 0th one? Or does the numbering of the array change so that the 0th bar is always the bar being currently executed? Assuming some of what I said above makes sense... when are the buy/sell/stoploss/takeprofit conditions checked? At the close of the bar? This is of course important for my systems, many of them are intraday and I need to know the level of granularity I will require from my data in order to obtain an accurate simulation. I have a zillion more questions, but this is a good starting point. Well, one more: is there a tutorial that I have missed and that addresses this kind of general questions about program execution and structure? Thanks in advance for your help!
