Paper Trading on a VPS: Discipline Over Dopamine

Paper Trading on a VPS: Discipline Over Dopamine

In the world of solo developers and small studio operators, managing resources efficiently is key to staying competitive. When it comes to paper trading—simulating stock market trades without real money involved—the challenge lies in balancing computational power with cost-effectiveness. This is where leveraging a Virtual Private Server (VPS) becomes invaluable. In this post, I’ll share my setup for running Belfort and Scarface on Hostinger VPS, focusing on the importance of discipline over dopamine to maintain a robust paper trading environment.

Running Agents on a VPS

The core of my paper trading setup involves two key agents: Belfort and Scarface. These are designed to simulate stock market trades using historical data, allowing for analysis without financial risk. To keep costs low while ensuring performance, I’ve opted for Hostinger’s VPS service. This choice allows me to run these agents on a dedicated server with minimal overhead.

Agent Configuration

• Belfort: A paper trading agent that runs simulations based on historical market data.
• Scarface: An auxiliary agent designed to complement Belfort by providing additional analysis and insights.

Both agents are configured to operate within the constraints of my VPS environment, ensuring they do not exceed resource limits. This setup is crucial for maintaining stability and reliability during paper trading sessions.

Pre-Market Checklist

Before initiating any paper trading session, it’s essential to go through a pre-market checklist. This ensures that all systems are operational and ready for the day ahead. My checklist includes:

• System Health: Verifying that Belfort and Scarface are running smoothly without errors.
• Data Integrity: Ensuring that historical market data is up-to-date and correctly formatted.
• Resource Allocation: Confirming that the VPS has sufficient resources to handle the load.

By adhering to this checklist, I can minimize disruptions during trading sessions and maintain a disciplined approach to paper trading.

JSONL Log Harvest

To track performance and identify areas for improvement, logging is crucial. My setup involves harvesting logs in JSON Lines (JSONL) format, which allows for easy parsing and analysis. This data is then used to refine strategies and optimize agent configurations.

Logging Benefits

• Performance Analysis: Identifying bottlenecks and optimizing resource usage.
• Error Detection: Quickly pinpointing issues that could affect trading accuracy or reliability.
• Trend Identification: Analyzing long-term trends in market simulations for strategic insights.

By maintaining a robust logging system, I can ensure that my paper trading environment remains efficient and effective over time.

Keeping the Loop Remote

While local resources are powerful, they come with limitations. By keeping the loop remote—running agents on a VPS—I can leverage cloud scalability without incurring high costs. This approach allows me to focus on refining strategies and improving performance rather than managing hardware constraints.

Local vs. Remote Trade-offs

• Local: High computational power but limited by physical resources.
• Remote (VPS): Scalable, cost-effective, and easier to manage for long-term projects.

By striking a balance between local development and remote execution, I can maintain a productive workflow without compromising on performance or financial constraints.

Next Steps

Update (Jun 2026): The stack now has a memory — see Paper Trading Needed a Memory — Hermes on the VPS for nightly trade review, bounded env patches, agent KB updates, and SFT export on the same Hostinger box.

Hermes runs on systemd timers (belfort-coach.timer nightly, belfort-hermes-hourly.timer during RTH). The repo is VPS_Agent. Train local on the MSI, loop remote on the VPS — discipline over dopamine, but now the bots learn from yesterday.