I recently stumbled upon a tool for backtesting and validating EA strategies in MT4 that is quite interesting actually. I have manually been doing the same a few times, but it is a lot of work. The testing strategy is called a Walk Forward Analysis.
Check out the Walk Forward Analyzer. There is a 14 day free trail and cost (only) $30.
Just a quick note that Birt has updated his excellent collection of tools for tickdata backtests in MT4 using Dukascopy tickdata (some time ago, but I’ve missed it until a few days ago). The big news is the new compact FXT format (deletes duplicate ticks) which really saves on file size allowing you to test for longer periods and a extension of the 2 GiB limit to 4 GiB on Vista and Windows 7 systems. I’m still on Windows XP SP3, so I haven’t been able to test the last feature, but with the new scripts I still can test 2+ years on any currency without hitting the 2 GiB barrier. The tests also run faster on the same time period because it contains fewer ticks, which is always very welcome.
If you are new to tickdata backtests or run older versions you have to check it out.
I just asked myself that question as its been a few days since I started backtesting with tickdata from Dukascopy. Even short runs take a lot of time… Optimizations of a few parameters can take weeks. I just aborted a optimization run that was supposed to take 1800 hours (that is 11 weeks roughly).
So, naturally I started thinking that maybe it is time that my 3 year old Intel Q6600 is replaced by a faster part. I must say that it have performed absolutely excellent, especially since I’ve overclocked it to 3GHz all the time I’ve had it. It is a quad core CPU, but sadly MT4 is a single threaded application, so only one of them is used at a time. I usually solve this by running four instances of the same test with one of the parameters range cut in four, but it is a hassle. MT5 solves all this as it not only is able to use all available cores in the host system it is able to out-source processing to agents running on other computers. But reality right now is that MT4 still is the main platform, and it probably will remain so for at least another 6-12 months.
I’ve googled for benchmarks, but I can’t find any. Sure it is a niche, but I really expected at least some results for MT4. I thought surely someone had bought a new computer sometime and tested it by running the same backtest on both. I know I would, but it seems I’m alone in this. Can’t find any benchmarks for either MT4 or MT5.
I found a great (old) article about getting optimal data for backtests in MT4. I personally usually use history data from JadeFX because that’s the broker I, mainly, use scalping strategies at, but it is full of errors and rarely gives even 90% quality.
Check it out: http://eareview.net/tick-data
I’m gonna test this as soon as I get a hour or two over… IE hopefully on Saturday.