Say you develop a strategy and want to check that it is not over-optimized and how it will handle itself in out-of-sample data. The best way to do that is to trade it on a demo account, going forward in real-time. Pretend it is a live strategy and that your demo dollars are real.
Walk-forward testing is, by far, one of the best tools you can possibly use in strategy development. You obviously have to use it properly and make sure that you understand what you are doing. Having a walk-forward perform well doesn’t necessarily mean it’s funded forex good. Having a walk-forward perform exactly how it is supposed to in the environment of the walk-forward is what you’re looking for.
Say I develop a ranging strategy, and my walk-forward data happens to consist of a breakout environment. If my strategy does not do well over that period, then I’m happy. Why? Because it did exactly what it was supposed to do. At that stage, I’m still not ready to run it because I’d like to see that it actually performs well when it is supposed to, but I’m not going to throw it away. If my ranging strategy performed well over the break-out walk-forward data, then I’d be a lot more worried!
There is a lot to be said about splitting your historical data in two sections. Optimizing on the first half, and treating the second half as walk-forward data. This is an amazing way to really speed up the whole walk-forward testing process. However, there is a lot to be said for subconcious bias. Because the second half of the data is not exactly 100% unknown, you may be selecting strategies that perform well in it, without actually knowing it.
The best way to guard against this is to use data that you know you have definitely never seen before and never could have predicted – i.e. future data. It is hard to slow down your excitement when you have a strategy that is almost ready and watch it trading on a demo account for 3 months while your real money is elsewhere doing other things, but the value you obtain from doing an excercise like this makes an enormous difference in your understanding and robustness of your strategy.
The Bounce Is Your Friend
Okay, so the subject isn’t as catchy as, “the trend is your friend” but it’s just as valid (more below).
With all the JPY crosses crumbling on daily charts like it was going out of style and the majors experiencing similarly strong moves, I hope you were all able to catch a piece of the action.
I would hope no one is making the mistake of fading these kinds of gigantic moves. I’m sure some traders may disagree with me on this and unless you’re extremely experienced I wouldn’t recommend it for the average person. With such broad sweeping moves this past week it becomes even more important to be patient and pick your spots carefully. On daily charts I’d prefer to wait for an actual retracement/bounce before considering getting into some of the pairs.
For example, it’s possible that there could be more downside follow through for USDJPY next week but, it’s best to wait for a bounce first and gauge how likely continued selling will be based on the extent of the bounce. Friday’s Legendafx is fairly indecisive so if a bounce were to happen then this bar sets up the potential for it early next week. If and when the bounce happens we’d need to see it develop to figure out if it’s worth a shot to short again. By waiting for a healthy bounce we’d be able to get a better price while increasing the chances of catching a wave of selling pressure as bears attempt to retest the most recent lows and beyond.
It’ll be interesting to see how the markets shape up next week relative to this past week’s action. Buckle up, it could be a bumpy ride.