Hi Robert -- In response to your questions --
---------------------- Equity Curve Filtering. This is the process of monitoring the equity curve that would have resulted from taking all trades signaled by a basic system (call this stage one). When used as part of an equity curve filter system, this equity curve is called the "shadow equity". In stage two, apply what is essentially another trading system to the shadow equity to determine whether the stage one signals should be taken or passed. To gate just means to allow to pass or to block passage. -------------------- A simple example of an equity curve filter. Apply a moving average to the daily shadow equity. If the shadow equity is above its moving average, allow trades -- that is, generate the buy and sell signals. If the daily equity is below its moving average, block trades. ----------------- The system begins by allowing the first trade, or some number of trades. Once the equity curve filter becomes active, there are two equity curves. One is the shadow equity from stage one, the other the final equity from stage two. The shadow equity is always being updated, even when the trades are blocked. When the shadow equity improves, trades are allowed to pass again. ------------------- Note -- there are many other ways of monitoring and controlling the ongoing condition of a trading system. Monitoring the shadow equity is just one example. ------------------ You wrote: "<...> multistage approach assumes that the best individual A, B, and C create the best portfolio P. <...> I'm hard pressed to see where the assumption comes from. I certainly disagree with it strongly." I do not understand what you are questioning or disagreeing with. -------------------- You wrote: "WFA in these groups is often mentioned as a validation process but often seems to be used as a dynamic optimisation process. To my way of thinking these are two separate beasts." I agree with your statement. These are two very different beasts. There is nothing special about optimization. It is simply an organized method of generating a large number of alternative versions of a trading system. The key point relative to an optimization is the method by which the alternatives are ranked. The ranking is accomplished by applying the objective function to each alternative, then sorting on that score. The objective function is the important component. There is something very special about walk forward analysis. Every walk forward step gives you, as the system developer, a look at what is likely to happen as your system moves from development to trading. That is, from in-sample to out-of-sample. If we think about trading being a participation sport, walk forward analysis is the practice court. Live trading with real money is the tournament. ----------------------- There is a more thorough discussion of these concepts in my book, "Quantitative Trading Systems," and in some of the previous posts in this forum. There is also an extensive discussion of system development in several of the threads of Aussie Stock Forums. Search using my name "howardbandy" http://www.aussiestockforums.com/ ------------------------ Thanks for listening, Howard
