Welcome, Guest. Please login or register.

September 28, 2020, 03:09:13 AM

Login with username, password and session length

Shoutbox

Recent

Members
  • Total Members: 95
  • Latest: prasono
Stats
  • Total Posts: 748
  • Total Topics: 70
  • Online Today: 9
  • Online Ever: 144
  • (August 14, 2019, 12:20:12 PM)
Users Online
Users: 1
Guests: 8
Total: 9

8 Guests, 1 User

Author Topic: Backtest related questions  (Read 602 times)

Attractor

  • Administrator
  • Sr. Member
  • *****
  • Posts: 299
    • View Profile
    • SmartForexExpert
Backtest related questions
« on: May 29, 2020, 11:17:46 PM »
hi Joel, can I ask for your forex EAs, which data set do you use to backtest and optimize on? And do you find difference in performance in backtest in different brokers? As I know IC Markets doesn't have historical data and backtesting on recent ICM data gives different results from other data such as Alpari or Dukascopy 99%

I don't want to reveal my 'secrets' but I never mind much the quality of the data, because If a systems works with data A and don't work with data B, we have a problem.... this mean the edge of the system is tiny.

The proof of that the quality of the data is overvalued, is that I arrived until here using always 'open prices only' in mt4. As you can see, I have my own rules and I don't follow the industry standars :))

For me the main problem of the historical data is if is consistent or not in relation to the sessions. I mean, some EAs like NS, works well only in a range of 1-2 hours. And I observed that some historical data his GMT is not consistent along the years, then is difficult backtest these kind of EAs. Maybe 2 years the session ends at 22, then the next months ends at 24 and so, this is a problem, but last time I cheched the data of ICM the closes were consistent along the years.

Usually, at least in my EAs, when I have found important differences in the results beetween brokers, has been related with this time issue.

But I'm in other screen now, because work with MT5 has been a great improvement for me. Why? because I don't know the quality of the data, but what is true is that the backtest modeling and the received results are more realistic than in MT4. In the historical data, the spread is stored, is an average spread of the life of the bar, but better this than nothing. If you system have a lot of pips/trade of expectation, you will not note the difference, but in systems with less expectation, the downgrade in the performance when you backtest in MT5 is incredible.

I think this is one of the reason why a lot of developers don't want to move to MT5, because keep the backtest 'sexy' is a lot more hard there.

But if you are a developer that really are trying to do something that works, and not only trying to sell to the newbies based on all the known tricks, then, the MT5 backtest EA helps a lot because give you a more realistic perception of what you can expect of the system.

So, more important that the quality of the data, is the quality of the backtest internal mechanincs, and in this, MT5 overcome MT4.

And not mention that I able to backtest all the symbols at the same time instead of compound individual backtest, for this reason, my speed of development has improved a lot last year, nad I can explore more ideas with the time I have than when I worked in MT4.

Sorry for this long explanation, but I want that people understand why I never care much the quality of data, and why I'm focused right now only in MT5 as develop tool.