Topic: Consistency and Volatility

I'm still pretty new to this program but I came up with a few very simple Daily strategies for Eur/USD, GBP/USD and USD/JPY that appear to be very consistently profitable based on the data that came with the program from 1989 to Sept 2010. Before I potentially waste my time testing these systems in my practice account, maybe someone on this forum can point out something I'm missing or misunderstanding.

First of all I want to point out that all of the strategies are specific to their respective pairs, they do not appear profitable in any other pair. They are all simple strategies based on MACD crosses, RSI, and or Moving Average crosses.

My logic(please correct it): IF I'm using the Builder correctly, my Eur/USD strategy, for example, trading a consistent number of lots just 1 lot at a time, and NOT a percentage of the balance so that the balance should increase exponentially, shows that with the account starting at $10,000 in 1989, the balance rises very consistently in a tight channel from 1989 to Sep. 2010 with an ending balance of $134,000. The minimum balance was $10,000, never dipped below the starting balance. SOOO, I'm assuming that if I execute this simple system accurately in my practice or live account over years I should almost definitely achieve similar results, and of course if I were to trade more and more based on the rising balance the gain would be exponential instead of linear. I mean if the system was working so consistently with data for over a decade, I am assuming that doing the same for the next decade would produce similar results, maybe the slope or the rate of balance increase could change, but money-making nonetheless.

What am I missing here? Seems to make sense, but at the same time I get the feeling that one minor difference between this software's execution of the strategy and an actual execution could change everything. Obviously, testing it is the best way to find out.

When trading a consistent number of lots, what would you all consider good or bad? What should I look for in the balance? Is 10,000 to 134,000 over a decade good with a consistently traded 1 lot at a time? Should the system do better to build in kind of a buffer zone in case there are mistakes? I was thinking that I was really looking for consistency, something that really creates a tight, increasing balance, instead of something volatile that spikes up for 2 years then down for 2 years, not good for someone who doesn't have infinite funds to work with. The balance looks very similar to the sample photo on the front page of forexsb.com, maybe even tighter.

Am I on the right track here? What am I missing?

Please excuse my ignorance with this program and with creating Forex strategies in general as I'm just a beginner. Any help is greatly appreciated.

Re: Consistency and Volatility

Hey,

All in all you do make sense. I'm convinced that only you can answer whether something is good or bad, if you see that backtesting results are up to your standards, then what more do you want?

The relevant thing here is to not believe backtesting results blindly, if there is a probability (which is based on the backtesting results) that your strats have the edge, you have to confirm it, whether in a demo or live trading. At first keep the risks low, of course. If I were you, I'd look up the most minimal time period, in which the strat stays profitable, for example if you know that at the end of a three month cycle your strat is always on the plus side, demo or live trade your strat for 3 months. Then you know more about the validity of the edge.

Hope this helps,

F

Re: Consistency and Volatility

If you haven't already, I'd recommend reading a few articles on curve fitting and overoptimization, and walk forward testing. I've had a lot of strategies where they look really good over the data set I used to develop, but then find the strategy doesn't work going forward or if tested on another set of data. They were probably random paths through the data that happened to be profitable, with no real or true "edge". I've found that techniques to avoid curve fitting help a little in recognizing these types of false starters.