Forum

So you're a swing trader? That was what first got me interested in forex actually, but it seems like not enough profit can be made trading only once per week. So I assume you use MACD and RSI, then the trend is your friend until some trailing stop?...

What are your take on EA's that do indeed use randomness to win, like the "Simplest Hedging EA Ever" http://codebase.mql4.com/6601

This one makes money, yet it never predicts anything, really a 50/50 fools betting EA, but seems to work.. and with nothing to go wrong, it will probably forward test better than complex indicator based EAs........ what do you think about these 50/50 types?

Aren't all EA's just "on the surface" - they only work for a while, then you need to tune them again".. I believe it was Boris who said that... There isn't really any EA that works without optimization..

I use random entry/exit strategy EA's to capture a lot of information regarding the market dynamics. It speaks to entitlement, and the data itself are useful when applying statistical analysis of the market.

In my case I use "time" as my indicator. A position is taken at a random time, held open for a random amount of time, and then closed at a random time. The resultant P/L as well as MAE/MFE are tallied. Rinse and repeat on years of historical data. It's all MonteCarlo driven, then I do it a few thousands of times, pool the results, apply statistics.

With those results in hand I then have a benchmark by which I can assess the performance of any EA strategy I might consider entertaining. If it can't outperform the random strategy (and P/L is not the metric to optimize for in backtesting) then its worthless.

Oh and one lesson learned in the creation of my random entry/exit EA is that one person's idea of "random" is another person's definition of "biased". You should have some expectation of what a truly random strategy ought to produce in terms of results.

In my case the results set me back because of their true simplicity:


Here we see that the resultant profit/loss from a random entry/exit strategy conforms ridiculously well to the statistical expecations of a generalized normal distribution (and not that of a Gaussian function as one might expect) and the average P/L turns out to be exactly the spread for the instrument (in favor of the broker, meaning the trader over-all experiences an average loss per trade exactly equal to that of the spread...which is the right answer!).

The general form of this graph is true no matter the financial instrument I apply it to. The mean P/L value of the population is always the spread in favor of the broker. This is what we'd expect from a truly random entry/exit strategy.

Anything you do that results in something contrary to this is to be interpretted as you actually implementing something that is NOT a random strategy...somewhere in your strategy you have built-in a bias and the results of that strategy are really speaking to the bias (favorably or unfavorably) and are not the result of your effort to have a random entry/exit strategy.

1005, Wow! That is a truely perfect bell and sigma curve, exactly what I would expect from a random strategy.

You've given a lot of food for thought.. So what I conclude from it is, a "random" strategy is actually a "really-really simple" strategy, because I do bias the trailing stops (cut the loosing, and let the winning run free) - which in the long run as I tested last night, will result in a nice upward slope do to the mathamatics.

But what is to be said then, that a "random-Entry" trailing stop EA really does make profit, I guess so, that is just such a simple answer to the weighty question of creating a profitable EA - it really is -->that<-- simple? As randomly entering trades, and setting trailing stops?.........