PRIVATE R&D UPDATE 🔬⚙️
A few days ago, I shared the first results from the private autonomous trading engine I’m building with my colleague Aurélien.
The project sits between 3D_NEXUS_META, MT5 Liqbot AI, and a full quantitative research laboratory.
Since that first Gold Futures test, the engine has evolved dramatically.
This time, I moved the research to the E-mini S&P 500 Futures contract, using the deeper and more active order-book data from the main contract.
The latest version now includes:
• NEXUS microstructure signals
• LIQUIDITY IMPULSE detection
• INVERSE IMPACT detection
• Signal Arbiter for conflicting signal clusters
• MARKET and intelligent LIMIT execution
• Pending Order Guardian
• netting basket management
• adaptive SMART sizing
• multi-horizon signal tracking from 1 to 300 seconds
• MFE, MAE, expectancy and first-passage analysis
• BUY/SELL market baselines
• real-time feed-quality monitoring
• statistical confidence intervals
• signal stability analysis
• Parquet and SQLite research datasets
• a complete live STATS dashboard
This is no longer just a bot that receives a signal and fires an order.
It is becoming a complete research, execution and validation engine capable of studying every signal, every market state, every rejected opportunity and every execution outcome.
Latest full MT5 demo test 📊
Starting balance:
$10,000
Current balance:
$22,511.50
Closed net profit:
+$12,531.50
Total trades:
248
Winning positions:
69.35%
Profit Factor:
1.20
The system faced several severe drawdowns during the test.
One of them almost erased a large portion of the accumulated performance before the engine recovered and reached a new all-time high.
That was the most important lesson of the entire experiment.
The signal engine was working.
The main issue was not necessarily the edge.
The issue was the interaction between SMART sizing, multiple basket entries and total exposure.
So I reduced the risk.
Current configuration:
• TP: 20 ticks
• SL: 20 ticks
• maximum basket entries: 1
• SMART maximum multiplier: ×3
• no position stacking
• Signal Arbiter enabled
• LIQUIDITY IMPULSE enabled
• INVERSE IMPACT enabled
Since that adjustment, the gains have become smaller and more controlled, but the capital has continued to rise without the same violent exposure spikes.
Less fireworks.
More structure.
More control.
More usable data.
This is exactly what private R&D is supposed to do.
Not hide the drawdown.
Not celebrate a lucky equity spike.
But identify what created the performance, what created the risk, and separate the true signal edge from the leverage applied around it.
The next stage is now clear:
• fixed-risk control tests
• SMART ON versus SMART OFF
• E-mini data with Micro E-mini execution
• signal analysis by side, session and market regime
• multiple-day out-of-sample validation
• conservative basket-risk limits
• preparation of the future ML Shadow Mode
This remains an exclusive private project.
It is not available.
There is no public access.
There is no release date.
But the engine is alive.
The execution stack is stable.
The data layer is becoming institutional-grade.
And for the first time, the system is not simply trading the market.
It is studying its own intelligence. 🧠📡
#AlgorithmicTrading #OrderFlow #MarketMicrostructure #SP500 #ESFutures #FuturesTrading #MT5 #PythonTrading #QuantTrading #AITrading #MachineLearning #ReinforcementLearning #NEXUSMETA #TradingResearch
1
1July 14, 2026 170
