Is this EA good based on backtest?
Hi,
Sounds good. Can we test it?
The way to improve it is to forward test it.
moneyline
Hi DeadEye96
You should watch the modelling quality. If this is lower than 99%, then the backtest is irrelevant.
Have you imported historical data and convert the periods?
Here is a guide on how to improve the quality of backtesting:
http://www.metatrader.info/node/66
Why don't you attach the report here? I could tell you more if I see the report.
Anyway...the forward test is the best. Don't use an EA on live account unless you have tested it for a few months on a demo account!
Bye.
Here's a report....
Moneyline - The model uses a partner's algorithm. I will ask him if I can post the execution file. Thanks for your input!
Cucurucu - Modelling Quality
Cucurucu, why do you say the test is irrelevant if the modelling quality is less than 100%? That is the one report statistic that I really don't understand. Can you tell me what it represents in plain language? I have done many backtests and have never seen anything close to 99%. Thanks, DeadEye96
I don't know how to get 99% but it's possible to get a lot higher than 44%.
I can't tell you how the modelling quality is calculated but i think 44% is too low to be really significant. But the curve looks great.
Thanks Eric79. I would have thought that based on the number of bars, number of ticks modeled and the number of trades executed over a two year period that it would indicate significant statistical reliability. If you guys still have concerns then is it really the quality of the tick data and the strategy tester? Does anyone keep more/better quality data even at a price?
A lot of people think the MetaTrader Strategy Tester is not worth much anyway. But i think if you get a high modelling quality number then it gives at least some sort of idea how good the strategy is.
If you have tick data for that period then i'm very surprised why you would get only 40 something %
Eric79, a member in another forum (HarryHid) says that the modelling quality number represents the overall quality of the tick data used - not the quality of the model. There is a description on the Metaquote site forum that describes the tester, but the description of modelling quality is hard to decipher. Essentially, it says % of data which is why EA's in faster time frames have typically higher percentages (more time slices). Any other thoughts?
I may be wrong but i don't think the timeframes play a role there. If you test the EA let's say on the M30 or M5 timeframes, do you get a better modelling quality?
Edit: By the way i have tested EAs in higher timeframes than hourly and gotten a quality of around 90%
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Have run MT4 EA on EuroUsd H1 chart over past 25 months with following results: 169 trades, 2.13 profit factor, 8.5% max drawdown, 65.09% profitable trades equally split between shorts and longs with constant lot size. The model has had consecutive losses and rebounds with equal consecutive wins. In your opinion, is this result an indication of probable success with live trades? What else would you do? Any suggestions would be welcome! Thanks for your help!