Independent Forex EA research. 15+ years in algo trading. MT4/MT5 backtests with tick data, real spreads, forward monitoring & critical analysis
Joined June 2023
- Tweets182
- Following17
- Followers51
- Likes203
Pinned Tweet
I've tested 1,000+ Forex EAs over 15 years. What would you disagree with here?
After 15 years of Forex algorithmic trading, after testing over 1,000 advisors, after losing 20+ accounts, after earning over 100,000% in profit, I can say the following:
1. The simpler the strategy with fewer parameters, the better it works.
2. Be sure to conduct high-quality testing on real spreads and quotes with a quality of at least 99%.
3. Be sure to compare backtest results with real trading, and monitor the degradation of the trading system.
4. Use brokers with the best execution conditions, low spreads, and low commissions.
5. Be sure to use rebate services. Don't underestimate them; they can significantly reduce costs.
Years of trading research, real EA testing, and honest analysis — barely any likes.
Maybe it's time to switch to the trading guru playbook: Lambos, luxury, and the millionaire lifestyle.
I can do that too. Let's see how this performs. 😁
#ForexTrading #TradingHumor
ALT Forex Trading Humor
The formula for profitable algorithmic trading:
High-Quality Testing × Good Broker Conditions × Luck = Profitable
You can spend months developing and testing a trading strategy, find a broker with excellent execution, and still lose money.
Why? Because markets are unpredictable, and even a statistically sound strategy can fail.
We can control the quality of our testing and trading conditions.
But we can't control the market.
That's why luck will always be part of the equation.
ALT Profitable formula in Algo Trading
An interesting case study: impressive backtest results, but a losing forward test on a demo account.
Some time ago, we backtested a Gold (XAUUSD) trading EA from the MQL5 Market. The results were promising and gave us reason to believe the strategy might also perform well on a live account.
One of our community members decided to run the EA on a demo account and later shared a Myfxbook monitoring link showing three months of trading results.
We then ran our own backtest using real tick data over exactly the same period and compared the trades with the demo account's forward-testing results.
Here's what we found:
- 99% of trades matched between the backtest and demo account, excluding commissions.
- The developer has since removed both the EA and its trading signal from MQL5.
- We suspect the EA was most likely deliberately overoptimized on a long period of historical data to produce exceptionally strong backtest results.
So, what's the takeaway?
Even the most impressive backtest results don't guarantee that a trading strategy will be profitable in the future.
This is exactly why independent forward testing is so important.
A beautiful equity curve is not proof of a profitable trading strategy.
ALT Comparison of Sharkyra Gold EA backtest and demo trading balances from August to October 2026. Both results show losses, despite the EA's impressive long-term historical performance.
ALT Historical backtest equity curve of Sharkyra Gold EA on XAUUSD, showing steady long-term growth with relatively small drawdowns.
ALT Myfxbook monitoring of Sharkyra Gold EA 1.1 on a demo account, showing a 23.86% loss and a maximum drawdown of 28.96%.
ALT Sharkyra Gold EA historical backtest showing a smooth upward equity curve and periods of relatively limited drawdown.
The recent Binance hack is another reminder of why diversification matters.
And it's not just about spreading your money across different assets or trading strategies.
Don't keep all your capital with a single broker or exchange, no matter how trustworthy it may seem.
Real risk management starts long before you place your first trade.
#Trading #RiskManagement #Diversification
One thing I've noticed after years of algorithmic trading:
Most beginners want a profitable EA.
Very few want to spend months testing, optimizing, analyzing drawdowns, and rejecting strategies that don't work.
Everyone wants the results. Few are interested in the research behind them.
ALT Test Quant Analyzer
As an experienced systematic Forex trader, I want to highlight a serious problem that is emerging right before our eyes.
With the rapid development of AI, building analytical websites and SaaS platforms has become easier than ever. And while this creates exciting opportunities, it also opens another door for scammers looking to deceive inexperienced traders.
The problem is simple: anyone can now build a professional-looking trading performance monitoring website in a matter of minutes and present it as independent verification of a live trading account.
For years, platforms like Myfxbook and FX Blue have effectively dominated this space. Whatever their limitations, they established a level of credibility and a widely recognized standard for verifying real trading performance.
Now, with hundreds of similar services emerging, that standard risks becoming increasingly diluted.
I'm not saying competition or innovation is a bad thing. Quite the opposite.
But a professional-looking dashboard is not proof of real trading performance. And a website claiming to verify results is not necessarily an independent, trustworthy source.
This is a new challenge for our industry.
And I believe those of us with real experience in systematic trading have a responsibility to help newcomers understand the difference between verified trading results and convincing-looking numbers on a screen.
Because in trading, trust should be earned through verifiable data, not polished design.
#ForexTrading #AlgoTrading #AI
ALT The Myfxbook logo surrounded by emerging Forex trading analytics and monitoring platforms. The image highlights a growing challenge for the trading industry: as AI makes financial dashboards easier to build, distinguishing independently verified trading performance from unverified claims becomes increasingly important.
Another reminder that demo trading ≠ live trading.
On October 27, MetaQuotes-Demo received ticks dated 1970, XAUUSD briefly showed $2.04, and GBPUSD positions were closed at absurd prices.
This happened on a demo server, not live accounts.
Demo trading is only an approximation of reality.
ALT Screenshot of the MetaTrader 5 Journal showing abnormal tick data on the MetaQuotes-Demo server on October 27, 2026. Multiple “old tick” warnings for GBPUSD and AUDUSD contain timestamps dating back to January 1, 1970, highlighting a serious demo price feed malfunction.
One of MQL5’s most popular EAs, Waka Waka: 438 weeks of history, 73.9% winning trades and a 99.6% drawdown.
During an earlier crisis, the account was topped up by $3,000.
Grid/martingale lesson: years of smooth profits can vanish in one move. Watch equity, not just balance.
ALT Waka Waka EA MT5
I found an interesting AI trading experiment on MQL5.
A developer ported Kronos from PyTorch to native MQL5, allowing the model to generate future candlesticks directly inside MetaTrader 5.
The implementation worked. The predictions looked convincing.
Then came the actual test.
On EURUSD H1, across 947 forecasts:
- Directional accuracy: 51.83%
- No statistically convincing directional edge
- Worse price predictions than simply assuming the price wouldn't move — across all 16 forecast horizons
Here's the interesting part.
A trading strategy built around those predictions still finished in profit: 133 trades, Profit Factor 1.05.
But EURUSD had risen roughly 3.7% during the test period. The long trades made money, while the shorts lost.
The profitable backtest wasn't convincing evidence of predictive skill.
To be fair, this was a limited two-month test, using synthetic volume on a market and timeframe different from the model's main training domain.
My takeaway as a systematic trader: making AI generate price forecasts is impressive engineering. Demonstrating a statistically reliable trading edge is an entirely different challenge.
Before trusting any AI trading model, compare it against the simplest possible baseline.
Sometimes doing nothing is surprisingly hard to beat.
#AlgoTrading #AITrading
Found a neat MQL5 experiment AI: 20 Grok agents trade EURUSD M15 using identical market inputs but different prompts. Every 20 trades, the weakest 25% are replaced by mutated winners.
First test: 22 trades, PF 1.04.
I'd benchmark evolved vs original prompts on unseen data.
ALT Illustration of AI-powered algorithmic trading in MetaTrader 5. Grok AI connects to multiple trading agents that analyze market data and compete through evolutionary optimization. Each agent uses different trading strategies, with weaker performers replaced by modified versions of stronger agents. Based on an MQL5 experiment involving 20 AI traders.
Do you also come across similar scammy posts that have no value or practical purpose?
But most importantly, why do they always get thousands of views and hundreds of likes? 🤔 X What the hell? 😅
Friends, I'd love to share my experience. I spend a lot of time in the online investor community. And I'm truly surprised by the fact that when some people start telling potential investors about their profitable robots, the investors simply believe how convincing the person is. They don't even ask for trading results or even a benchmark.
Therefore, no matter how convincing and confident someone sounds, always ask for proof of the profitability of their trading systems. This could be Myfxbook or a trading report from their broker. Think critically.
ALT An example of a highly profitable trading system on Myfxbook
If I could give one piece of advice to new traders, it would be this:
⚡️Learn how to backtest properly.
It won't make you rich overnight, but it could save you from blowing your account.
First, learn how to stop losing money. Making money comes next.
#algotrading #forex
ALT MT5 Expert Advisor backtesting results using 100% real tick data. Strategy Tester report showing 5,915 trades, $4,434.95 net profit, 1.25 Profit Factor, and 2.86 Recovery Factor. Forex algorithmic trading strategy performance analysis.
This is HUGE for algo traders!
MT5 now supports AI integration through MCP. AI agents can help backtest, optimize EAs, analyze results, and even execute trades.
After years of testing trading robots, I'm genuinely excited about this.
#MetaTrader5 #AlgoTrading #AI
ALT MetaTrader 5 AI integration through MCP in MetaEditor. An AI assistant generates and compiles an MQL5 trading script for automated grid limit orders, demonstrating AI-powered Expert Advisor development, Forex trading automation, and MQL5 coding.
Today I came across an interesting 2016 study by Chris Davison from Nottingham Trent University.
According to the research, only around 20% of retail Forex traders end up with profitable accounts.
And damn, it feels good to be on the profitable side of that statistic.
The chart below shows the performance of my own trading account darwin UAE at #Darwinex
Years of hard work, thousands of backtests, failed strategies, blown accounts, frustration, and starting over.
Looking at this equity curve, I realize those years of trial and error weren't wasted after all.
Still learning. Still improving. But today, I'm genuinely proud of how far I've come.
#ForexTrading
We tested Gold Scalper Advanced MT5 EA on XAUUSD using real tick data from #Darwinex.
The EA turned a $1,000 deposit into $5,435 over 5,915 trades.
Sounds good, until you look at the 83.84% drawdown and 1.25 profit factor.
#AlgoTrading #MT5
+1,113% profit. Sounds great, right?
Then you open Myfxbook and see an 81% drawdown.
Same account. Same strategy.
That's why we always look beyond the profit curve. The numbers can tell a very different story.
#Myfxbook #AlgoTrading