EA Performance | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

FX Dealer Performance 92 replies

Performance statistics 4 replies

Question: evaluating performance 1 reply

Performance, Fortune,& the U.S. President 0 replies

FXWINTRADES managed fund performance 3 replies

Attachments: EA Performance
Exit Attachments

EA Performance

  • #1
  • First Post: Jan 14, 2007 8:04am Jan 14, 2007 8:04am
If you have an EA that opens/closes positions on a open candle. How reliable can its back test be with 90% modellig quality?
Thanks
  • #2
  • Jan 14, 2007 2:04pm Jan 14, 2007 2:04pm
Quoting saaedb
If you have an EA that opens/closes positions on a open candle. How reliable can its back test be with 90% modellig quality?
Thanks
As long you have the M1 data for your EA, it will be as accurate as you can get if you're using MT4 to test your EA.

Doesn't matter when you open the trade.

GF
  • #3
  • Jan 14, 2007 2:51pm Jan 14, 2007 2:51pm
It isn't very reliable. The tick data is simulated so backtesting results don't seem to match live trading. If I need to test a strategy that requires intra-interval trading, i let it run a while on a demo account.

Backtesting strategies that use completed intervals is very reliable though.
  • #4
  • Jan 14, 2007 2:53pm Jan 14, 2007 2:53pm
You may want to keep up with this site. I stumbled upon it a few weeks ago and it'll be interesting to see how it turns out...

http://www.fxtradeblog.com
  • #5
  • Jan 14, 2007 4:19pm Jan 14, 2007 4:19pm
Quoting Palmer
You may want to keep up with this site. I stumbled upon it a few weeks ago and it'll be interesting to see how it turns out...

http://www.fxtradeblog.com
Thats an interesting link. I'll keep an eye on it and if I have the time I might do the same for my EA.
  • #6
  • Jan 14, 2007 4:20pm Jan 14, 2007 4:20pm
Quoting GirlFlyer
As long you have the M1 data for your EA, it will be as accurate as you can get if you're using MT4 to test your EA.

Doesn't matter when you open the trade.

GF

GirlFlyer thanks for the reply, I know your EA uses M1 Data, and I wanted to ask you if there is a way to get better then 25% modelling quality using the M1?
  • #7
  • Jan 14, 2007 4:28pm Jan 14, 2007 4:28pm
for 2004,2005 and 2006 respectively.

All the trades were a fixed 1 lot during these back tests. I'm running a demo account and so far so good.
Attached Images
  • #8
  • Jan 14, 2007 10:03pm Jan 14, 2007 10:03pm
Quoting saaedb
GirlFlyer thanks for the reply, I know your EA uses M1 Data, and I wanted to ask you if there is a way to get better then 25% modelling quality using the M1?
No, if you do your backtest using M1 setting, 25% is as high as it goes. It's just the way the formula is calculated. I don't know why they made their formula that way, but it is. While you may think 25% is bad, on that setting you have to ignore it. M1 data is as good as it gets (but it still doesn't mean the backtest will be perfectly accurate). I generally find that when I backtest my EA for the same week that I forward test something, it comes pretty close to the same results, so its not a bad deal.

Cheers

GF
  • #9
  • Jan 14, 2007 10:33pm Jan 14, 2007 10:33pm
I disagree. If your strategy uses M1 data and trades within an interval, the back test results will be the least reliable. The bigger the time compression you use, the less dependent it is on the "simulated" tick data being accurate. This is why you get a modeling quality of only 25% when using the minute charts. You will get much more realistic results if you shift 1 period (trade on completed intervals). In this case, you can ignore the 25% modelling quality.
  • #10
  • Jan 14, 2007 11:59pm Jan 14, 2007 11:59pm
Quoting GirlFlyer
No, if you do your backtest using M1 setting, 25% is as high as it goes. It's just the way the formula is calculated. I don't know why they made their formula that way, but it is. While you may think 25% is bad, on that setting you have to ignore it. M1 data is as good as it gets (but it still doesn't mean the backtest will be perfectly accurate). I generally find that when I backtest my EA for the same week that I forward test something, it comes pretty close to the same results, so its not a bad deal.

Cheers

GF
Thanks for your reply
  • #11
  • Jan 15, 2007 12:02am Jan 15, 2007 12:02am
Quoting fourex
I disagree. If your strategy uses M1 data and trades within an interval, the back test results will be the least reliable. The bigger the time compression you use, the less dependent it is on the "simulated" tick data being accurate. This is why you get a modeling quality of only 25% when using the minute charts. You will get much more realistic results if you shift 1 period (trade on completed intervals). In this case, you can ignore the 25% modelling quality.
Hey fourex, thanks for the reply.

I've done a visual representation of my trades and the trades are firing at the correct time using the M1 with 25% modelling quality.

Now, since I'm trading on a open candle, if I attach my EA to a hourly chart on the strategy tester, I get a modelling quality of 90% with the exact results as I get when attaching my EA to the M1 with 25% modelling quality.
So I guess GirlFlyer is right about M1 and 25% modelling quality (max)
  • #12
  • Jan 15, 2007 12:55am Jan 15, 2007 12:55am
I wasn't disagreeing on that point. I have no idea how MT4 calculates the modelling quality and it does seem that 25% is as good as it gets on the M1 data. Keep in mind though, if you are trading on tick data in any time compression it is probably using the same "simulated" tick data that it generates for the M1 data. So I'd expect the same trades at the same time. The question is, which modeling quality is more accurate, 25% or 90%. In my experience, no matter which time compression is used, you just can't rely on the backtest results if you are trading per tick. As GirlFlyer suggested, run it on a demo for abit. A good strategy will work both ways.
  • #13
  • Last Post: Jan 15, 2007 12:58am Jan 15, 2007 12:58am
Quoting fourex
I wasn't disagreeing on that point. I have no idea how MT4 calculates the modelling quality and it does seem that 25% is as good as it gets on the M1 data. Keep in mind though, if you are trading on tick data in any time compression it is probably using the same "simulated" tick data that it generates for the M1 data. So I'd expect the same trades at the same time. The question is, which modeling quality is more accurate, 25% or 90%. In my experience, no matter which time compression is used, you just can't rely on the backtest results if you are trading per tick. As GirlFlyer suggested, run it on a demo for abit. A good strategy will work both ways.
Yup, thats what I'm doing. Running it on a Demo and will gauge it after 2-3 months.
0 traders viewing now