To use chat, please login.
Back to contacts

can some expert analyze my result? please help

Dec 18 2011 at 07:46
3 posts
Attached is my EA backtest result.

I built my history data from dukascopy data and it shows 99% quality.

I have delveloped this EA myself with complex money,lot management techniques and also i used only pure mathematical decisions. (no indicators used)

The results I see is too good to be true.

So i need an expert eye to give me some comments on it.

Every trade has SL and TP. Risk % per trade is 1% of equity.

Test period - 01/01/2011 to 17/06/2011 (there is some gaps in other period so i didnt use it. Even if i use it is not a problem as I have proper SL and TP on trades)

Attached is my results page.

Thanks in advance
Dec 18 2011 at 08:32
202 posts
6 months is not enough testing period. You need to be thinking in terms of years.

With your current MM settings, your loss trades (as few as they are) = 2x profit. It should be the other way around. But that's me being picky, a ~90% win rate would cancel that out....if indeed you could maintain that high probability in real trading......
Steve Walker
Dec 20 2011 at 19:45
1341 posts
seems good. but as Splexin said the BT should be at least 3 years.
you can also post the full report and we can discuss more.
Jan 03 2012 at 07:18
63 posts
run this in forward live account and then see what the results are that is always the best way imho
You will never go broke by taking profits
Jan 03 2012 at 15:01
244 posts
What is the spread on your test and does it comply with the typical live spreads at the time your EA will be running live?

An EA that yields 50% per annum can yield 1,000% over the same period if the backtest spread is reduced to 0.1 pips.

It's a common trick for EA marketers, or for anyone out to deceive.
Consistency above all.
Jan 03 2012 at 15:50
1916 posts
it does not tell much,nowadays we can not trust backtest anymore ,the only trusting path is (LIVE) other than that has no value to a serious trader
Archer Managed Forex (archertrading)
Jan 03 2012 at 19:57
7 posts
@ jpbiznes - as a few people have said, you've got to run it for years. The more the merrier. A curve fit model is easier to work on a shorter period. Try to see the results in as many market conditions and cycles as possible.

@ Raiden - That is common and for your backtests we normally employ something like this: and set it to a value a tad higher than the statistical highs of the spread. Performance is reduced naturally, but you get a more realistic picture as slippage does occur in live trading.

@rob559 - every serious trader and quant I know backtests. To say that it has no value is beyond ridiculous. It shows you that the model is working as you want it to and if it's a viable model to even consider wasting time and resources testing. Almost every system (and not just MT4/5 here) I've EVER seen usually goes:

development > backtesting > adjustments > backtesting > demo testing > live testing > live

Backtesting, while not a necessarily accurate measure of future performance, plays a key role in designing systems. I think a lot of people got wrapped up in fact that the inherent design flaws in the MT3 and 4 back testers skewed results so, that they've disregarded it. That's just not the case in the world outside of Metatrader and while MT4/5 backtests aren't the best, they're OK if the data you feed them is good. Dukascopy tick data is about the best solution for MT4 backtests. Any backtest is only as good as the data you feed it.

It's only money...
Jan 03 2012 at 23:53
244 posts
Just in case it is construed as such, I did not mean to imply the TC, jpbiznes, is out to deceive anyone, I meant it in the context of not deceiving ourselves when we backtest.

@ McLean Can Cleve - Thanks for the link, it is a much cleaner and hassle-free method than editing the symbols.sel file. And I completely agree that backtesting is a critical step in designing a system. A backtest report that I don't do myself, however, I can't trust at face value. Simply because it is too easy to fabricate excellent results. rob559 might be saying that.
Consistency above all.
Archer Managed Forex (archertrading)
Jan 04 2012 at 05:58
7 posts
Sure, you can curve fit the hell out of a dataset, but it becomes increasingly hard over time and in varying market conditions. Hence a longer backtest. This is precisely why you get these EA sellers that do a backtests on a year by year basis. The same inputs wouldn't have worked throughout all of the years in the test and while it's not a bad thing to re-optimize your strategies for current conditions, we don't know what they are in advance and these guys naturally know what the market conditions were like in 2006 and they engineer the backtests to do the best for that timeframe.

So yeah.. I'm totally agreeing with you, just expanding a bit.
It's only money...
Jan 15 2012 at 20:16
7 posts
Be engaged better in manual trade
Jan 17 2012 at 21:18
1 posts
Backtesting is the first step in creating a good ea but as Splexin said you need to test it on a larger timeframe. Once that is done you will want to forward test it on a demo account. If that works then move to a live account to test.
Please login to comment .