Tuesday, April 8, 2014

How To Interpret Backtesting Results In Metatrader 4



Video Script: 
Before we get too far into backtesting, we need to cover what backtesting is not.
The results of your backtest are not something that you should automatically extrapolate into future performance. The most common rookie mistake forex traders make is finding an expert advisor that makes money historically, and then ending their analysis there. They see that it made money in the past, so they expect it will do the same in the future.
This is the wrong approach.  The primary purpose of backtesting is not to prove that a strategy has made money historically.  It’s to try and figure out whether or not it will make money in the future.
So, how do you use backtesting to figure out if a strategy might work well in the future?  You try to break the strategy.  If you shoot a bunch of holes in the strategy’s historical returns and they hold up, then that is likely a good sign about its chances for future profitability.
Let’s walk through the process of how we go about that. If you haven’t watched our last video on how to perform a backtest in Metatrader 4, pause this video and come back after you have.
If you are watching this video then you have likely purchased or found an expert advisor and are now looking to backtest it.  One of the reasons why you purchased it was likely the graph of historical results that were advertised along with the EA.
If you bought an EA, then the first thing you want to do with your backtest is to make sure that you have received the expert advisor that you paid for to receive. Does the performance actually live up to the hype
You’re going to do that the same way that you run any backtest. The EA vendor should supply some recommend settings. Your job is to use those settings over the period of the claimed performance. For example, if the vendor claims a 20% return in 2013 on EURUSD H1 charts, you want to test the EA over that exact same time period.
The results will vary somewhat between brokers, but the return should at least be in the ballpark of the expected results. If you find a substantial variation, that’s a major cause for concern.
Once you have verified that the EA backtests in line with the claimed historical performance, the next step is to look at how the EA performs when changes are made to its parameters.
Based on my experience, the best strategies tend to make money without extreme sensitivity to the settings used. It indicates something inherent within the strategy that captures market inefficiencies, without requiring the trader to calibrate settings like a nuclear physicist.
You’ll ask yourself questions like:
  • Is there any combination of settings that cause my EA to overtrade?
  • What values for the stop loss and take profit definitely kill the performance?
  • Say that I’m trading an expert advisor that uses moving average crosses. How sensitive is the historical return to changing the MA’s period? For example, does changing the fast MA from 25 to 30 make a small or large difference in the final outcome?
Let’s use a moving average cross strategy as an example. If changing the fast MA from 25 to 30 makes a substantial difference in the final outcome, that is an extremely troubling sign. What if the optimal period in the future is actually a 40 period fast moving average? If your inputs are super sensitive, that might be the difference between positive and negative performance. If your inputs are not sensitive, at least you’ll land on the same side of the coin.
You test the parameters of a strategy by doing the work manually. Pull up your MT4 backtester’s inputs tab. Try different combinations of settings. If you’re using moving average crosses, try “walking” the fast period from a super small value up to a value near the slow moving average cross. Take notes of the differences in the final performance.
When you do the work manually, you’ll notice hot spots appear in the returns. One combination will outperform all the others. Is that best combination similar to the nearby settings that you used? It’s a really good sign if the results are similar.
The period that you use for backtesting is also very important. Most traders feel comfortable when they see five years of historical performance. They assume that means the strategy will continue working in the future.
My answer to backtesting is this question: do you care how much money an EA made 5 years ago or how much it will make tomorrow?
EAs and strategies decay over time. One of those most legendary trading strategies in history are the Turtle traders. These guys traded commodity futures and averaged over 80% returns in the late 1980s.
People still Google them because their strategy is open source. All of the buy and sell rules are given away. There’s one catch, though…. The rules don’t work anymore. At least not in the forex market.
The market evolves and changes with time. Just because a strategy worked 20 years ago, or 5 years ago, or even yesterday doesn’t mean it will work tomorrow. The best tool in your arsenal is to find an EA with encouraging performance on recent historical data. Check the parameters to make sure that they’re not overly sensitive. When you see something that you like, then and only then have you found an expert advisor that you might use on your live account.

No comments:

Post a Comment