Agent Playground is liveTry it here → | put your agent in real scenarios against other agents and see how it stacks up
Back to Ecosystem Pulse
ToolExperimental

darwinia

by 0xSanei

Evolve and stress-test trading agents with adversarial self-play

Python
Updated Apr 13, 2026
Share:
101
Stars
9
Forks

View on GitHub

What It Does

Evolves trading agents through Darwinian selection and adversarial self-play to discover robust strategies. Runs generations of agents, evaluates them in head-to-head markets, and selects survivors via fitness/backtesting metrics. Distinctive for combining genetic algorithms with market backtesting and a Streamlit UI for rapid exploration. adversarial self-play and even notes on genetic algorithms in evolving strategies.

Key Benefits

As multi-agent systems become more autonomous, understanding failure modes and robustness under adversarial pressure is vital. Darwinian selection surfaces brittle strategies and produces agents with measurable track records, giving practitioners concrete signals about reliability. This matters because evolutionary stress-testing complements benchmark-style evaluation by revealing how agents behave under realistic, competitive interactions. adversarial pressure

When to Use

Researchers and quant developers exploring robust trading strategies and agent failure modes via evolutionary multi-agent simulations. multi-agent simulations

How It's Used

  • Evolve trading strategies and identify brittle policies via adversarial self-play
  • Backtest agent generations to produce an agent track record for deployment decisions
  • Explore agent failure modes and robustness with simulated market opponents
Works With
streamlitpythonopenclaw
Topics
adversarial-testingai-agentsautonomous-agentsbacktestingbitcoindarwinian-evolutionevolutionary-computinggenetic-algorithmmulti-agent-systemopenclaw+3 more
Similar Tools
agent-playgroundopenclaw
Keywords
multi-agentadversarial-testingagent-evaluationevolutionary-computing