Announcement

Collapse
No announcement yet.

Forex EA's - The Lowdown from Nick's Perspective

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Forex EA's - The Lowdown from Nick's Perspective

    Hey guys,

    Over the past 3-4 years I've spent a massive amount of time trading and researching heaps of EA's and I thought I'd share some of my experiences with you.

    Before you consider trading an EA on a live account I recommend you run through the following checklist:

    1. Does it have a verified forward test?
    Most vendors offer this now, but unless it's got a completely verified and open 3rd party analytics (Myfxbook or MT4i) then don't waste your time. A screen shot doesn't mean sh#t, I want to see a full statement.

    2. How does it trade?
    Get an understanding of what mechanisms the EA uses to open and close trades. Then if it starts to perform poorly you can make the decision on whether it's a temporary change in the market or time to pull the pin completely.

    3. Find other happy customers.
    If people are making money on a system you will find plenty of reviews and comments across industry forums. Dig hard enough and you will be amazed at what you can uncover.

    4. Pay attention to the website.
    If it looks like the website was put together in an afternoon by a 12hr old kid then steer clear. It's very easy to put together a basic EA, stick up a website, get a month or two's positive performance (through luck normally) and start marketing through the forums and review sites. It's a typical "pump and dump" scenario. I've seen it a few times with vendors signing up to be signal providers on our copier business SimpleTrader.net. A professional company will invest in a professional website.

    Feel free to help me add to the list and share any experiences you have.
    Click here to check out the most popular forex channel on YouTube

  • #2
    Although a vendor may provide a live forward test, sometimes sitting out for 2-6 months can save you a ton of time and effort. It is very easy for some inexperienced to make an EA, curve fit the backtest to make it look highly profitable with low drawdowns, and then market it. Sitting on the sidelines will enable you to A) See if the system actually works, and B) see how the system works in relation to your emotions. It is a lot easier said than done to watch your account go into a 30% DD. Many people think they can handle it, but most cannot.

    Contrary to what Nick said, I feel that you should not pull out an EA just because the market may have changed. The market has changed many times, and if the system has been backtested properly and you had a time machine, people were probably saying the market changed back then too. Instead set a plan for yourself to know exactly how much you are willing to lose. This should be based off of your backtests and risk analysis. This will give you a clear-cut number of when you will pull the plug as to not miss out on profits due to, as Nick said, temperary changes in the market which you thought were permenant.

    David

    Comment


    • #3
      What a great debate! I thought I'd add my irrelevant opinion in on this.

      I agree on point 2 that us humans will always have the advantage over any EA for understanding market conditions (assuming you have watched markets for long enough to know). So if you have a clear understanding of what each EAs logic is and you can apply it to the right conditions, I think you have a better chance of long term success. I don't think the answer is just to leave it to do its thing IMO.

      This idea that EAs are curve fitted on back tests always intrigues me when I hear it. Are they all not? Unless AI has been realised in EA design I'm pretty sure they are all designed on what has happened. Is that not curve fitting?

      Phillip aka Mr Tech

      Comment


      • #4
        Hey Phil,

        Yes, the idea behind curve fitting is always an arguement. They are based off of prior data. Indeed. However, according to traditional EA building thought processes you should only optimize your EA to a small portion of your entire backtest period. With a span from 2000-2013, again, according to traditional thoughts you should be optimizing 2000-2002(optimize for 2 or 3 years) and then backtest for the entire 2000-2013 period to see how it performances outside of the optimization period.

        From what I have read (I will try and find the link), now that we have more than enough data (10+ years) an EA should actually be optimized for the entire period. Doing this takes into account the multitude of times the markets have 'changed'.

        Cheers,
        David

        Comment


        • #5
          Originally posted by Tempestshade View Post
          Hey Phil,

          Yes, the idea behind curve fitting is always an arguement. They are based off of prior data. Indeed. However, according to traditional EA building thought processes you should only optimize your EA to a small portion of your entire backtest period. With a span from 2000-2013, again, according to traditional thoughts you should be optimizing 2000-2002(optimize for 2 or 3 years) and then backtest for the entire 2000-2013 period to see how it performances outside of the optimization period.

          From what I have read (I will try and find the link), now that we have more than enough data (10+ years) an EA should actually be optimized for the entire period. Doing this takes into account the multitude of times the markets have 'changed'.

          Cheers,
          David
          The idea of testing an EA through 10 years of data doesn't sit well with me personally. The market is so different from way back in 2003.

          I feel it results is a system that doesn't quite fit any markets, because it tries to fit all markets.
          Click here to check out the most popular forex channel on YouTube

          Comment


          • #6
            I am in no way shape or form trying to promote one way of optimizing versus the other because both have their valid arguements.

            The arguements to support optimizing an EA through an extended period of time (the max possible) is that, as you have said, markets do indeed change. So what is the point in testing a system for a 'sample period' in say, 2000, when we are currently in 2013? Optimizing for the entire period gives you a strategy, that in theory, will work in all of the market changes that have occured. And a continuous optimization as time continues will only strengthen the results, supposedly.

            Now, I am only paraphrasing. I haven't been able to find the article that discussed this topic in great detail. I too am very skeptical with this methods but I monitor several private EAs that use this method of optimizing and I must say the results are pretty extraodinary. One is based off of RobinVOL, FGB, and other volatiliy breakout EAs which are all taking major hits. This particular EA is killing the market right now, and has been for the last 4 months (during the current DD period of the other EAs). Obviously this is a rough example as no two EAs have exactly the same logic, however, it is interesting to watch.

            David

            Comment


            • #7
              Despite having looked for the past hour I have been unable to find the link to the article describing the study performed on back testing.

              It would be interesting to see an EA that is optimized using traditional methods vs using an extended period. If anyone has a EA that fits the traditional method, I would be highly interested in seeing the two different methods put on (demo) accounts and run side-by-side to see which performs better.

              Cheers,
              David

              Comment


              • #8
                forex

                looks very promising thank you

                Comment

                Working...
                X