Backtesting FX`

[Option1]

I have been forward testing a breakout strategy (using MT4) for the past 3 weeks, and it seems to be a valid idea

However I need to do a backtest, just for comfort sakes. I do not want to use MT4 programming, is there something else ?

I am considering purchasing TS 8

Any ideas ?

 

[Doctor Leo]

Tradestation is a nice tool for backtesting, although WealthLab provides much more powerful and flexible programming environment. Briefly, of you have no programming experience you might want to try TS, if you're aquainted with common programming languages, I'd better go for WealthLab if I were you (to be precise, WealthLab uses Pascal as a scripting language).

 

[PKFFW]

I recently bought Forex Tester from forextester.com It's a relatively cheap simulation programme that also has an API thingy so you can program in the stuff and have it backtest for you. Pretty much all the standard indicators, time frames and drawing tools are included with more stuff being added all the time.

It is certainly not as powerful as something like Wealthlab or TS though. However, I like it because you can get lots of practice actually identifying the setups on the charts and trading them in a very short period of time.

My two cents anyway.

Cheers,
PKFFW

 

[Trader333]

The problem with back testing is that all trades tested are at the last traded price and this is totally unrealistic. Normally you would enter a trade at the Ask price (if going Long) and exit at the Bid price. You can also experience slippage so that compared to the last traded price in a back test, you may be several points or cents or pips away from the back test price depending on the instrument being traded. This can have a massive impact on the theoretical profit of that given by back testing on the last traded price. I once developed a trading simulator that would only buy or go long at the Ask price and sell or go short at the Bid price. The difference in profit compared to back testing was substantial but it did more accurately reflect how real trading is. In my view it is of great importance to consider this when developing trading methodology using back testing.


Paul

 

[Option1]

Thanks for that

Paul: I do agree with you and hence just started forward testing only - using the bid/ask prices and to check out the validity of my ideas

How to move forward then ?

 

[Doctor Leo]

Easy: you can count for slippage and spread setting up corresponding parameters in TS. When I had used it (before I switched to my own software) I used to set up slippage 2 pips and spread (which is able to be conted for in TS as brokerage commission) 6 pips. In reality I almost never have a slippage and spread is 2 to 4 pips, so if my strategy shows positive results on history, then I might be confident that in real trading I'd receive results at least no worse than on backtesting. So, it's not a problem.

 

[FX_jake]

Of course, by using Forextester (and manually trading through the past), you also are not trading the "last traded price." For those who do not use EAs (or other automated trading systems) to trade, this is probably a good way to test.

 

[Alibubble]

You can use backtesting to try out a forex trading strategy to see if it might work in real conditions, but even if it does well that doesn't mean that it will work in the future.

What I mean is that something that worked in the past won't always work forever. The market conditions could have changed and all of a sudden the strategy that looked like it would make a lot of money is now losing a lot

 

[rhw]

You can use backtesting to try out a forex trading strategy to see if it might work in real conditions, but even if it does well that doesn't mean that it will work in the future.

What I mean is that something that worked in the past won't always work forever. The market conditions could have changed and all of a sudden the strategy that looked like it would make a lot of money is now losing a lot

Yes - recently I tested a system that can be optimised such that it works on all market conditions, but with different optimisations for each. I have yet to find an effective 'market condition' detector... I 'detected' the unhelpful market conditions, because those were the time-periods when it was getting 'stuck' (not increasing equity). How you do that programmatically, and not post-facto, I dunno.

It's working on the current conditions in EURUSD, but will probably do less well when a long trend starts again. I periodically retest and re-optimise.

This is also one of the issues I have with the dictum that one must optimise on a range of past data, test on a separate range. You can never be sure that your training and test data embody the same market conditions - a perfectly good system may seem not to work simply because you're implicitly expecting it to work under all market conditions, which is kind of unreasonable unless you've explicitly built that in (which, of course, if you know how to, you should).

 

[NVP]

Yes - recently I tested a system that can be optimised such that it works on all market conditions, but with different optimisations for each. I have yet to find an effective 'market condition' detector... I 'detected' the unhelpful market conditions, because those were the time-periods when it was getting 'stuck' (not increasing equity). How you do that programmatically, and not post-facto, I dunno.

It's working on the current conditions in EURUSD, but will probably do less well when a long trend starts again. I periodically retest and re-optimise.

This is also one of the issues I have with the dictum that one must optimise on a range of past data, test on a separate range. You can never be sure that your training and test data embody the same market conditions - a perfectly good system may seem not to work simply because you're implicitly expecting it to work under all market conditions, which is kind of unreasonable unless you've explicitly built that in (which, of course, if you know how to, you should).

I presume you mean "not work" when the market is not trending or whipsawing.....

thats the crux of the matter....getting a system to stay out in non profitable conditions and not bleed you dry...........any EA can make money off of a half decent Trending market....

N

 

[oscarthecat79]

There are a heap of ways to backtest. None are perfect though.
There is a video here looking at how Bloomberg do it on the terminal using tech analysis

Backtesting Trading Strategies: Technical Analysis Part 2 - Fintute

 

[adrid]

I believe that each trader has his own vision , and we should respect it but there is one question that I heard it during my last 10 years of trading , should I follow fundamental or technical analysis .

The answer is simple , if a positive news for euro will come and your indicator will tell you to sell , you will loose . so even I m technical analyst I have my personal indicators but each trader should check the economic calendar before taking any position .


I m new in this Forum , I ll do my best to bring any positive info for traders .