Experienced Python/MT5 Developer Needed to Optimize XAUUSD Signal Engine — Increase Signal Frequency Without Sacrificing Quality

Title

Project Description

I am looking for an experienced developer specialized in Python, MetaTrader 5, algorithmic trading systems, market structure and XAUUSD, to work on an existing professional Gold/Forex AI signal platform.

The project is already developed and operational. I am NOT looking for a developer to build a new trading system from scratch.

The main objective is to improve the existing signal-generation architecture so that it can produce approximately 2 high-quality XAUUSD signals per day on average, while preserving signal quality and avoiding a simple relaxation of all existing filters.

The target is an average, not a guarantee of exactly two signals every calendar day.

Current Problem

The current engine is highly selective.

It uses multiple confirmation layers including:

• Smart Money Concepts
• Order Blocks
• Fair Value Gaps
• BOS
• CHOCH
• Liquidity
• Volume
• Multi-Timeframe analysis
• Momentum
• Trend Strength
• Support/Resistance
• News Filter

The current system can identify meaningful market movements, but many potential opportunities are rejected before becoming actionable signals.

We have already performed extensive historical and shadow testing.

Several obvious approaches have already been tested and should NOT simply be repeated, including:

• Lowering the main score threshold
• Loosening PHASE_B / MTF restrictions
• Changing counter-trend thresholds
• Removing individual confirmation requirements
• Using SuperTrend + SMA as a replacement signal engine
• Quality-gated direct entry
• Entry-zone fallback experiments
• Various alternative candidate families

Some of these increased signal quantity, but the resulting realistic performance was not sufficiently robust.

An important finding from recent realistic Tick-data testing is that entry/execution selection can create adverse selection. Therefore, I do not want a developer to solve the problem simply by generating more signals or weakening every gate.

What I Need

I need an experienced developer to diagnose and redesign the signal-generation pathway intelligently, while preserving the core philosophy of the existing system.

The goal is:

More opportunities + maintained quality + realistic execution performance.

The desired target is approximately:

≥ 2 quality XAUUSD signals/day on average

while maintaining:

• Positive expectancy
• Robust out-of-sample performance
• Reasonable drawdown
• Stable results across different years/market regimes
• Realistic spread/slippage assumptions
• No look-ahead bias
• Closed-candle logic
• No repainting

Important Requirement

Do not simply lower thresholds until two signals/day appear.

I want the developer to identify where good opportunities are currently being rejected, determine whether those rejected opportunities form a statistically valid secondary signal path, and then design the smallest defensible architectural change.

The solution may involve:

• A secondary signal path
• Better separation between trigger / confirmation / context
• Direction-aware interpretation of existing analyzers
• Better handling of mixed vs genuinely opposing evidence
• Improved entry methodology
• Better distinction between signal validity and entry location
• A controlled secondary-quality tier

But I want these conclusions to be based on data and reproducible tests, not assumptions.

Existing Architecture

The project is a Python/FastAPI-based XAUUSD analysis platform with a React dashboard.

The Signal Engine currently uses the following 12 analyzers:

1. Smart Money
2. Order Block
3. Fair Value Gap
4. Break of Structure
5. CHOCH
6. Liquidity
7. Volume
8. Multi-Timeframe
9. Momentum
10. Trend Strength
11. Support/Resistance
12. News Filter

There is also a Signal Manager responsible for things such as duplicate/repeat protection, cooldown and active-signal management.

Development Philosophy

The developer must:

1. First inspect the existing architecture.
2. Understand the current signal flow completely.
3. Reproduce the current Production baseline.
4. Identify the real bottleneck.
5. Propose measurable changes.
6. Test changes in Shadow/Read-Only mode first.
7. Compare against the current Production baseline.
8. Use realistic spread/slippage and, where available, real MT5 Tick data.
9. Validate IS/OOS and yearly stability.
10. Only implement Production changes after approval.

No Blind Refactoring

I do not want:

• A complete rewrite
• A new unrelated strategy
• Random threshold optimization
• Curve fitting
• Overfitting to one period
• Look-ahead
• Repainting
• Future candle information
• Synthetic results presented as live results
• Changing multiple variables without attribution

Every meaningful modification must have a measurable reason.

Required Deliverables

The developer should provide:

1. Architecture audit

Explain exactly:

Market Data → Snapshot → Analyzers → Direction → Score → Gates → Entry → Manager → Signal

and identify where good opportunities are being lost.

2. Baseline reproduction

Reproduce the current Production results before modifying anything.

3. Bottleneck analysis

Show which gates/rejections are responsible for the majority of missed opportunities.

4. Proposed solution

Explain the proposed architecture before implementation.

5. Shadow test

Test the proposed solution without changing Production.

6. Performance comparison

At minimum:

• Signals/day
• Executed signals
• Win Rate
• AvgR
• Profit Factor
• Total R
• Maximum Drawdown
• Confidence interval
• IS/OOS
• Annual performance
• Performance by market regime
• Spread/slippage sensitivity

7. Final implementation

Only after the Shadow results are approved.

Files I Can Provide

After the developer confirms exactly what he needs, I can provide the relevant project files.

The initial review should focus on these files:

And the relevant API/config files if required:

app/config.py
app/api/v1/routes/analysis.py
app/api/v1/routes/signals.py 

I can also provide relevant historical signal/outcome data and diagnostic reports.

If MT5 execution/entry behavior is part of the investigation, the developer can request the relevant MT5 integration files separately.

Important

The project is GoldPulsev6 only.

Do not touch GoldPulse V4.

The existing frontend is not the main focus unless a UI change is specifically required to display the new signal state.

Developer Requirements

Please apply only if you have strong experience with:

• Python
• FastAPI
• MetaTrader 5
• XAUUSD
• Algorithmic trading
• Market structure / SMC
• Backtesting
• Tick-level data
• MTF analysis
• Statistical validation
• Avoiding look-ahead and overfitting

I am particularly interested in a developer who can diagnose an existing production system rather than simply write more code.

Please include in your proposal:

1. Your experience with XAUUSD signal engines.
2. Similar projects you have worked on.
3. How you would approach increasing signal frequency while preserving quality.
4. How you validate against overfitting.
5. Whether you can work with real MT5 Tick data.
6. Your estimated timeline.
7. Your proposed Phase 1 scope and price.

Phase 1 should be analysis + architecture + reproducible baseline + Shadow proposal. No Production changes without approval.