Re: the ZigZag part of this discussion - AB's implementation of backtesting is different than many other programs.
In AmiBroker, the backtest itself is not a process that proceeds on a bar-by-bar basis. At the detail level, the criteria necessary for a valid "bar-by-bar" backtest and the criteria necessary for a valid "all-at-once" backtest (AmiBroker) are not identical. What I'm saying in this context is that it is not only about "how functions work", but about how the backtesting process itself works. No one should be saying that a measured retracement from a pivot point can never have a statistically tradable probability of further extension. If anyone is saying that, they are incorrect, as a single counterexample will prove. However, I don't think that is what was said (though what was said could easily and predictably be interpreted as such, IMO). How a measured retracement from a pivot is detected in AmiBroker, for both validly backtesting and realtime trading, is the real issue at hand. Code has been presented for this. This code provides insight into the "AB way". No doubt, there are multiple ways in AFL to accomplish the same thing. --- In amibroker@yahoogroups.com, "jhnlmn" <jhn...@...> wrote: > > What does matter is how functions work. > Every trader must know exactly whether TroughBars[bar] does > look into the future after bar or not. In WL it does not. > Therefore, I can use TroughBars[bar] in WL for backtesting.
