How to choose a good EA - a practical way
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
People just think that a good backtest, excelent winning rate and many other indicators of Tester are enough.
However, the most important thing here (I mean in the entire MQL5) is not even discussed or worried about. Which is strongly related to the most common question in this Forum: why most of EA fails in live trades? The answer is quite simple, but ignored: most of them doen't take into consideration a proper statistical analysis.
You may think the backtest report when is good is enough, but it is not good at all!! It doen't provide all answer you need. And here, I also complain to MQL5 cause a lack of data is actually worse than NO data: no data leads you to study more, worse data may confuse you.
Trully, there are infinite equations that fit well to a cloud of points or, in other words, show profitable results in backtest. Some of them are really good on pictures, others go up and down consistently and others just go down. The developers, when decide to sale an EA, just choose the good looking one and ignore the others... and the buyer simply believe, cause he runs the same "cloud of points" in Tester and get same result.
---
Think of climbing a hill: there are many ways to do it: with or without tools, with or without good shoes, running, jumping, crowling. You can be fat, skinny, bodybuilder, child or adult... The difference between all of them is how much time and energy it takes from you. And how is the chance of failure.
Earning money with EA is just the same: you have a mountain of profits to climb and specific conditions to be on top. How will you do this? With great tools? (indicators), with good shoes, climbling rope and hat(protection), running? (frequency)? Some EAs go straight to the top like a falcon, others go up all around the mountain with very low inclination and most part falls into abysm in between...
OK, let's focus on the "best one", those ones that looks good, the falcons: 500 trades, no losses, exponential profit growth. Fantastic. Purchase price >= USD 1.000.
Have you ever got noticed of someone that climbed Hollywood hills in 10 minutes running fast? Probably yes. Do you believe he's prepared to climb Everest? If someone who climbed Hollywoods and tries just after Everest, will likely dies on camp base... You can make same comparison between Mont Blanc and Everest. One doesnt prepare you to another.
If you consider all professional climbers that succeeded in Mont Blanc and tried Everest just after, I guess 10% succeeded again. These are the EAs that MAYBE work in real time. I mean MAYBE because part of them might dies on the top of Everest.
There are infinite paths, climbers, amateurs, tools, indicators (and its combinations) that fit perfectly well onto a mountain of profits (backtest) but fail in the next step (live trading). Only statistics can identify which one will leads to real profit with a high probability. MQL5 doens't help you to choose, cause as many time you try to find it, as much time you waste/$$ and as much money MQL5 wins.
How to find a good EA so? Is it possible to do it through backtest? Yes it is!
---
1. First thing: keep in mind that YOU HAVE TO USE EXCEL to help you to do it;
2. You need to run Tester considering all parameters all together, at same time, with a WIDE RANGE, from the bottom to the top level. For instance: if you have a parameter that vary from 0 to 100, consider it. If you have 100 parameters with same range, consider it too all together in same Test. It will take a lot of time, but certainly less than trying in real time for months loosing money (it is important to mention here that as better is the EA as less parameters it needs to be profitable).
3. Export all gererated data to Excel;
4. Create a gaussian normal distribution or a histogram. If it is smooth (https://en.wikipedia.org/wiki/Normal_distribution), than you got a good EA and the median offers you a good combination of parameters. If not, maybe cracked, certainly it is a fake EA. Also, 90% of entire distribution must be with positive results, it meas you have 90% of probability to earn money with this EA. Finally, the median must be high, cause you want to earn a lot of money, right?
5. Suposing you found a smooth distribution, get the combination of parameters and come back to MQL5. Run backtest and see how it looks like. How many trades per week you found? What is the Profit Factor, Positive Hit Rate, Drawndown etc. Must also be good.
6. Come back to the first step and start over again, but now with a forward test with 30% of all history. Did you find quite similar parameters? Yes? Than you have a good EA with a good probability of success.
7. Try in real account, the shape of profit curve and behavior should be alike to the backtest.
---
This is a practical sequence of tests, there are many technical concepts behind. There are other ways to test, but this one will reduce the chance of failure.
Hope it helps you (and sorry all guys for my bad english).
Alexandre