Struggling with EA Development – Am I Missing Something?
In your backtest, it worked on 1 symbol only.
The profit factor is good enough in my opinion, but surely not the amount of trades. 35 trades in more than 5 years of backtesting is surely a too low number to make statistic consideration (in my personal opinion). I don't know if you worked with high timeframes (like H4 or above), or if the amount of trades is just the consequence of a lot of filters you put to try increase the profitability.
If I'm in your position, I will consider to disable one by one filters you added to the strategy (if any), and try to see which is the one who restrict too much compared to the improovement of the profitability.
For example. instead of a strategy with profit factor 1.8 that take 35 trades, i will prefer a strategy taking 200 trades with a profit factor of 1.5.
You may also consider to test on multiple symbols, if you can keep that profit factor among at least 8-10 pairs, the statistical edge should be more solid.
lots of eas and strategies are best for specific symbols. I wouldnt be concerned about your strategy not working on other pairs.
8 months is a common point where frustration kicks in. A few thoughts that might help:
The real issue is usually not the code, but the strategy validation process.
Most developers (myself included, early on) fall into this trap:
1. Find parameters that look profitable in backtest
2. Go live
3. Strategy fails
4. Tweak parameters
5. Repeat
The problem is overfitting. Your "optimal" parameters are optimal for PAST data, not future data.
What actually works:
1. Reduce parameters drastically. If your EA has 10+ inputs, you're almost certainly curve-fitting. The best EAs I've built have 3-5 parameters maximum.
2. Out-of-sample testing is mandatory. Split your data. Never touch the OOS portion until final validation. If you peek, the test is contaminated.
3. Test on multiple symbols. A strategy that only works on EURUSD H1 is suspicious. If it's capturing a real market inefficiency, it should work (with minor adjustments) on similar pairs.
4. Accept lower returns for higher robustness. A strategy with PF 1.4 that's stable across 10 years and multiple symbols is far better than PF 2.5 that worked great in 2022.
5. Minimum 200+ trades in your backtest. Below that, statistical significance is questionable.
The mental shift: Stop looking for the "perfect" strategy. Look for a "robust" strategy. They're very different things.
What's your current strategy based on? Sometimes the issue is the underlying edge (or lack thereof), not the implementation.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Hey everyone,
I’ve spent the last 8 months trying to build a profitable Expert Advisor (EA), and honestly, I’m starting to lose hope. I’ve developed multiple EAs during this time - some showed great results on a single pair, but as soon as I tested them on other pairs (even after optimization), they failed miserably.
My most recent EA uses Order Blocks, and while it’s been stable across different pairs, the returns are… underwhelming. Over 5.5 years of back testing, it only made around $1000 (10% total). Running it on multiple pairs might help, but even then, the profit seems too low to justify going live.
My questions for you:
Is this normal for Order Block-based strategies, or am I doing something wrong?
How do you improve an EA’s profitability without sacrificing stability?
Should I keep refining this approach, or is it time to explore a completely different strategy?
Any advice, critiques, or personal experiences would be hugely appreciated. Thanks in advance!