Potential Issues Found During Setup and Testing

[calvinbasumatary-cpu]

1. Python Compatibility Problems

The README mentions Python 3.9, but the project does not prevent installation on newer Python versions.

When using Python 3.14:

• psycopg2-binary==2.9.9 fails to install
• numpy==1.26.3 fails to build
• Installation fails without a clear explanation

Suggestion

Add a startup/version check or update dependency constraints to explicitly support newer Python versions.

---

2. Unnecessary PostgreSQL Dependency

The project appears to use SQLite by default, but installation still requires psycopg2-binary.

This caused installation failures even when PostgreSQL was not being used.

Suggestion

Make PostgreSQL dependencies optional or only install them when PostgreSQL is selected.

---

3. Duplicate / Highly Correlated BTC Signals

During testing, the bot generated multiple actionable BTC signals with nearly identical values across consecutive 5-minute windows:

• Same direction (UP)
• Similar model probability (~55%)
• Similar edge (~4.6%–5.7%)
• Same suggested trade size

This may result in multiple highly correlated positions being opened simultaneously.

Suggestion

Consider limiting exposure across overlapping windows or adding a correlation filter.

---

4. Possible Market Window Timestamp Issue

The API response for BTC 5-minute markets showed timestamps that appear inconsistent with a 5-minute market duration.

Example:

• window_start and window_end suggest a much longer duration than expected.
• Could be a timezone conversion issue or parsing issue from Polymarket data.

Suggestion

Verify timestamp handling and UTC conversions for BTC market windows.

---

5. Calibration Concerns

After initial testing:

• Accuracy: ~40%
• Brier Score: ~0.279

This suggests probabilities may not be well calibrated.

Suggestion

Review probability calibration and consider post-processing methods such as Platt Scaling or Isotonic Regression.

---

Environment

• Windows 11
• Python 3.14 (installation issues)
• Python 3.9 (successful installation)
• Latest repository version
• Backend and frontend running successfully after switching to Python 3.9

[Nikhilrai27]

Hi! I'd like to investigate the Python compatibility and PostgreSQL dependency issues mentioned here.

I'll try to reproduce the installation problems locally and, if confirmed, submit a PR with either documentation updates or a dependency fix.

Please let me know if anyone is already working on this.

[Nikhilrai27]

I investigated the PostgreSQL dependency concern.

Findings:

• The default configuration uses SQLite:
  DATABASE_URL = "sqlite:///./tradingbot.db"
• I commented out psycopg2-binary==2.9.9 from requirements.txt.
• Installation completed successfully on Python 3.10.
• The application also started successfully using python run.py.
• Database initialization and application startup both worked with the default SQLite configuration.

Based on this testing, it appears that psycopg2-binary is not required for the default SQLite setup, although PostgreSQL support still exists for production deployments.

A possible improvement could be making PostgreSQL dependencies optional or moving them into a separate requirements file for PostgreSQL users.