Simulation Example¶

Here we will demonstrate a simple simulation where agents randomly place orders.

First we import Bourse

import bourse

then define an agent class

class RandomAgent(bourse.step_sim.agents.BaseAgent):
    def __init__(self, i, price_range):
        self.i = i
        self.price_range = price_range
        self.order_id = None

    def update(self, rng, env):
        # Place an order if one is not live
        if self.order_id is None:
            price = rng.integers(*self.price_range)
            side = bool(rng.choice([True, False]))
            env.place_order(side, 10, self.i, price=price)
        # Cancel live order
        else:
            env.cancel_order(self.order_id)
            self.order_id = None

For an agent to be part of a simulation it should define an update method that takes a numpy.random.Generator and bourse.core.StepEnv as arguments.

In this example an agent places a random order if it does not have an existing one, and otherwise attempts to cancel its existing order.

We then initialise an environment and set of agents

seed = 101

agents = [RandomAgent(i, (10, 100)) for i in range(50)]
env = bourse.core.StepEnv(seed, 0, 1, 100_000)

We can then use bourse.step_sim.run() to run the simulation

n_steps = 50

market_data = bourse.step_sim.run(env, agents, n_steps, seed)

market_data is a dictionary of Numpy arrays containing market data recorded over the course of the simulation.