SOFTWARE ENTITY · EXPERIMENT, NOT FINANCIAL ADVICE

What is Trading Swarm?

Trading Swarm is a public Python experiment by Marco Hergi. Seven independent agents evaluate technical or sentiment signals and contribute to a consensus whose weights adapt after outcomes resolve.

Project
Trading Swarm
Category
Multi-agent signal experiment
Language
Python
Agents
Seven
Initial weight
1.0 per agent
Boundary
Not investment advice
01

Seven agents inspect different evidence

The current README documents agents for SMA crossover, RSI, momentum, Bollinger Bands, volume and on-balance volume, MACD, and headline sentiment.

Each agent returns an action and confidence. The system combines them instead of treating any one signal as certainty.

02

Consensus is weighted and bounded

Every agent begins with a weight of 1.0. According to the README, a correct agent's weight is multiplied by 1.1 and capped at 2.0. An incorrect agent's weight is multiplied by 0.9 and floored at 0.5.

The caps prevent one signal from becoming infinitely dominant or disappearing entirely. This is a simple adaptive feedback rule, not proof that the system learns a profitable strategy.

03

Why the architecture is useful

The reusable pattern is heterogeneous evidence aggregation. Separate agents can specialize, expose confidence, and be evaluated independently before their outputs are combined.

That pattern can apply outside markets wherever several imperfect signals need a transparent consensus and a bounded update rule.

04

Performance claims require caution

The repository includes historical testing and paper-trading features. Historical results can overfit, omit costs, or fail under new conditions.

Trading Swarm should be described as a software and architecture experiment. It is not a recommendation, financial advice, or guaranteed performance.

SOURCE HIERARCHY

Verify at the origin.

This is an educational description of public software, not financial advice. Repository details and reported historical results can change and do not establish future performance.