Metatrader's strategy tester: Limited functionality?

[Highfield]

Do you find the strategy tester in Metatrader usful at all? I find it lacks a great deal in terms of the data the results generate and feeling that I posted an open discussion on a forum whereby I outlined what would be a valueable contribution to the results generated by the strategy tester. The response was almost ideal and the programmer went a good way to seeing this through before giving up; I can only suppose it took up too much of his time but I was extremely greatful to him for his splendid effects all the same.

The idea for what is effectively an indicator of the strategy tester results was too find the best average profit level for an EA rather than relying solely on Metatrader's implementation of maximum take profit level. For example, (dealing only in the profit level for the moment): Within a trading system there may be a profit level that occurs at a high frequency; an optimum level to take profit. If it were possible to figure out where this level lies in relation to the EA's current variable settings then that would go some way to honing an EA's settings. The same could also be said of stop loss level settings whereby levels which are rarely met are then the most likely best place to site a stop loss providing the drawdown isn't excessive and does not exceed the profit levels. Does that make sense?

I have attached a screenshot of the the indicator I'm refering to, a visual snap-shot of the strategy tester results. I could upload the MQL files for you to look at if you want these but I would first need to ask for permission from the programmer to do this.

[Highfield]

It may be that a picture speaks a thousand words but I realise that this screenshot may need some explanation.

The chart in the top half of the picture is displaying a moving average crossover system driven by an EA. The lower half of the picture shows a histogram of each trade measured in pips (pre fractional pip Metatrader makeover). Long positions are drawn coloured blue and short positions are drawn coloured orange. Centered on a zero line (no profit/loss), the histogram plots the maximum pip level reached by the position for each complete bar. When a position is in profit, the histogram plots the above the zero line; below when not in profit. Ergo, we can see that anything below the histograms zero line is -ve.

This is perhaps a useful tool in being able to determine the freqency of pip levels that the EA is achieving and where stop losses could be sited out of reach. However, it's not quite complete in that the indicator does not calculate the frequency itself and so backtesting over longer periods is difficult and most likely not exactly accurate.

[Jim]

Hi Highfield,

Do you find the strategy tester in Metatrader usful at all? I find it lacks a great deal in terms of the data the results generated

The MetaTrader strategy tester is useful in so far as it is free of charge, and allows people to get a feel for what's involved in backtesting and optimising (but not over optimising!) a system without parting with any cash. However it suffers from a number of limitations, particularly if you're testing a "scalping" style strategy. I've covered some of these on the blog.

I could upload the MQL files for you to look at if you want these but I would first need to ask for permission from the programmer to do this.

If the programmer in question is happy to place his source code in the public domain for the greater good of other MetaTrader users then by all means upload it. I'm afraid I can't promise to dive in and finish the job, but possibly someone else might come along who is willing and able to do that?

[Highfield]

As I've not been able to successfully make contact with the author of the EA and indicators that I've referred to and the fact that the original thread which they appeared on now seems to have been moved to some sort of elitist section of it's web site, I've taken the decision to upload them here in the knowledge that the files are already in the public domain having been previously openly submitted to a public forum.

The contents are:-
- Hull MA Expert;
- Hull MA indicator;
- Tester Report.

It's the tester report which I find the most intriguing as it's a good visual aid to how the EA performs given a set of data and rules. What it doesn't currently do is work out the frequency to which trades are profitable or where the ideal place for a stop loss could be sited.

[Highfield]

The Tester report is a visual aid to understanding how the EA performs given a set of data which I outlined in post #1. In order to use the Tester Report one needs to perform a simple operation post EA optimisation (if required). Once the Hull Expert EA has been selected as the strategy to be tested, it's expert properties will need to be chosen making sure that, under the section HMA Parameters, 'Mode' is set to equal 2. Now the EA needs to be run in visual mode and this can be achieved by selecting the 'Visual mode' checkbox in the Strategy Tester window. Next, the EA is set to 'pause' by clicking on the pause button also in the Strategy Tester window. Next, the Tester Report is dragged onto the chart where it's properties dialogue box will open automatically. The settings for the Tester Report are:- Mode = 1, Shift = 0; vMode = 1 which are set as the default values. Once these operations are completed the final stage is to press the play button in the Strategy Tester window.

[Highfield]

It would be very nice if the Tester Report were able to produce the same output from any EA which, from what limited understanding I have of the code, is controlled by two global variables set with the MQl function GlobalVariableGet(). Ultimately it would be ideal if the Tester Report was able to identify the frequency to which trades made n pips and the frequency to which trades lost n pips. Given this data we would able able to fine tune an EA into something profiable.

For example:-
Taking 20 trades the EA makes, a tester report ought to be able to understand that were it the case that each trade made a minimum of say 20 pips and a maximum of 120 pips yet the EA made only 1 trade at 120 pips then it's clear that the EA ought to be exiting a trade after 20 pips (20x20 pips = 400 pips versus 1x120 pips). Ditto the stop loss.