Almgren–Chriss Execution Model

An implementation of the Almgren–Chriss model for Optimal Execution, extended with a reinforcement learning agent that learns to minimize Market Impact.

→ github.com/varunbudati/Almgren-Chriss-model

Why build it

The efficient frontier is one of those results that reads as abstract on the page and becomes obvious the moment you can manipulate it. Turn the risk-aversion parameter up and the trade schedule front-loads — sell fast, pay impact, stop worrying about price drift. Turn it down and the schedule flattens toward TWAP.

You don't really believe the impact/risk trade-off until you've watched the curve bend.

What's in it

The analytical A–C solution for the optimal trajectory, a simulation environment for the price and impact dynamics, and an RL agent trained inside it — so you can compare a policy that was derived against one that was learned under identical conditions.

That comparison is the seed of what later became the research at the Dataism Lab.

Stack: Python, NumPy, reinforcement learning.