Thread: Back Testing Limitations and Suggestions

I have been doing back testing on many ea's, along side with forward testing with demo's,

I have found that when using back test, it doesn't recognize hidden TP's and SL's like in Firebird, have since been testing my back test based on what would possibly be used in forward, IE. Firebird True SL is set to 300. however I have found in forward it hit's around a SL of 80.


Optimization seems too work great. super computer recommended however if there is something set that is unreliable in backtesting, could have dreadful results.

does anyone have any other experiences with back testing limitations or suggestions?

hmm limitations of back testing... let me see...

Probably if you're working with an EA that is being used on multiple currencies at the same time and the lot size of each trade is taken into consideration in relation to your current account balance as well as the maximum open positions allocated to each currency pair being used.

With this kind of setup, you can probably only test it fully with forward testing. Or test each currency pair individually and see what kind of results you get. XD

hi back testing works different data.i have EAS that will make millions on Metaquotes data but fail in forward test. If you can, get Alpari data from their site.Some of the problem is,that when you go to history centre to download data,it asks you to download the Metaquotes Data which is so smoothed that even a poor EA will look good.
Johnno

Originally Posted by johno

hi back testing works different data.i have EAS that will make millions on Metaquotes data but fail in forward test. If you can, get Alpari data from their site.Some of the problem is,that when you go to history centre to download data,it asks you to download the Metaquotes Data which is so smoothed that even a poor EA will look good.
Johnno

If that's the case, then EA programmers should incorporate data smoothing into the ea code. Perhaps EAs based on raw price action have most back/forward test inconsistencies

MetaTrader doesn't simulate margin call. This is the reason for some backtesting failure.

I use backtest to see if the idea i had is good in many market condtions, for example, you can test 1999 until 2009 and see if your EA makes money, or you can test the crashes at 2008 etc. If an EA fail on backtest you save time because it will sure fail on forward demo and live tests, that's why i use backtest, i don't use backtest as a final answer for the EA, after i backtest and the EA end up with an good profit and low DD i start testing it on demo for about 3 months before go live.