If I can now just figure out what happened near the end of 2006 till now? - page 2
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I attached the source file to the other post.. it's not commented though. Take a look at it, and then let me know if you have questions.
I found only the compiled code CochiseEAb22a.ex4 from here
'Misleading results in back tests'
I can have a look at your code if you send it to gpwr9k95@yahoo.com. From your results, it appears to be a scalper. Which currency pair and timeframe does it use? Does it have restrictions on trading times? I know a few scalpers that stopped working in 2006. It had something to do with the currency prices. As an example, check these EURGBP 1M prices from the MQ server before and after 09-29-06 14:20. Notice the High-Low range before and after.
I am using EURUSD, 15M timeframe. It trades between 7 & 20 GMT. That's interesting about the high/low range. I wonder if other data sources have the same drop? Is there any reliable way to backtest? I am completely aware that a backtest is not a true test but from others who have let me seen their profitable robots, their forward tests show about a 75% correct curve to the backtest one. So it is one way of success measurement, as long as you factor in some sort of margin for error.
You can probably tune the input parameters of your scalper to make it work again, depending what indicators it uses.
I will try to tweak parameters and see what happens. I just don't want to over-optimize. This system doesn't use any indicators, just price movement.
From technical analysis, there are bull traps and bear traps. In this case, I call it newbie, or novice, trap.
Just looking to your results, I see that modeling quality is 90%, which might be considered very good. But it isn't. It just means that you are using a scalp strategy on bigger timeframe. At older dates, the tester doesn't have from the minute timeframe, so it has to guess a lot (read about how prices are obtained when history is not available). The result is that the tester opens and closes trades at price levels that didn't really happened - on the prices it had to generate. To assert this, after testing try observe the profitable trades on the minute chart.
As long as testing time advances, the TESTER finds more precise prices to use from 1min, 5 min... so, it has to guess less prices. This way trades are made at more real prices, and therefore, you start to see more umprofitable ones.
All in all, it's a matter of time resolution. You are trying to model a behaviour that happens at the minute frame, but using a base scale of hour or alike.
Don't blame the TESTER for this issue. It's just a matter of available history adequate for your pip trading EA...
So,
if you want develop such kind of EAs, you should always test using the minute Period. If your code needs to use indicators from higher time frames, you must use function based, instead of array based indicators:
Close[] --> iClose(...)
High[] --> iHigh(...)
....
Using the minute period for testing, you will get modeling quality of 25% (which is the best you can obtain). That is, 1 minute and four available points (H,L,O,C) yields a quality = 1/4 = 25%.
Additionally, if you only use ticks, then you cannot rely on the TESTER, as it uses the least timeframe only, which is 1 minute period and its four points. Your only option is to test the EA live on a demo account and as you observe its behavior, you can change or improve it.
Joao
Thanks for the input. That makes sense. I know the tester is not a good true result...what I said above:
I am completely aware that a backtest is not a true test but from others who have let me seen their profitable robots, their forward tests show about a 75% correct curve to the backtest one. So it is one way of success measurement, as long as you factor in some sort of margin for error.
Is there any way, or place, to get an accurate back test using reliable historical data?
Here is a retest with the same parameters using iClose and all th "i" functions:
M1 all tics, 2% equity risk, trading for 15M timeframe
M15 all tics, 2% equity risk, trading for 15M timeframe
I don't really mind uploading the MQ4 source file. It would be nice for the system to work and I would love everyone to benefit from it and contribute it. I figure a collective mind is better than 1. This might naive. If anybody thinks it's foolish to give the source please let me know. Of course, I won't be surprised if the system never reaches success. This wouldn't be the 1st EA to show inflated profit, and then fall on it's ear. To start, attached is the EX4 file:
I don't really mind uploading the MQ4 source file. It would be nice for the system to work and I would love everyone to benefit from it and contribute it. I figure a collective mind is better than 1. This might naive. If anybody thinks it's foolish to give the source please let me know. Of course, I won't be surprised if the system never reaches success. This wouldn't be the 1st EA to show inflated profit, and then fall on it's ear. To start, attached is the EX4 file:
This is the same as CurrencyProfits. It exploits a bug in all historical data prior to 2006.09.29