ianmadlenya/pyziabm

Zero Intelligence Agent-Based Model of Modern Limit Order Book

pyziabm

Zero Intelligence Agent-Based Model of Modern Limit Order Book

Charles Collver, PhD

See the Project Website for more details and radicalmarketsimulation for a more general discussion of applying agent-based modeling to financial markets.

If you just want to install and run a quick simulation: conda install -c dennycrane pyziabm

In a Jupyter Notebook or iPython command line:

import pyziabm as pzi

pzi.Runner()

will call the simulation with the defaults. To change the defaults and run several in a loop see the runwrapper files detailed below. To keep all of the data from one simulation in memory, assign to a variable:

sim1 = pzi.Runner()

There are seven files:

1. orderbook3.py
2. trader2017_r3.py
3. runner2017mpi_r3.py
4. runner2017mpi_r4.py
5. runwrapper2017mpi_r3.py
6. runwrapper2017mpi_r3x.py
7. runwrapper2017mpi_r4.py

orderbook3.py

• Contains the Orderbook class.
• Instances track, process and match orders.
• Imported by runner2017mpi_r3.py.

trader2017_r3.py

• Contains the base ZITrader class and six subclasses: PennyJumper, Taker, Provider, Provider5, MarketMaker, MarketMaker5.
• There is also an InformedTrader, but it is yet unimplemented. See pyziabmc for more details.
• Imported by runner2017mpi_r3.py and runner2017mpi_r4.py.

runner2017mpi_r3.py and runner2017mpi_r4.py

• Contains the Runner class.
• Instances set up and run 1 simulation.
• Note: the __init__() in runner2017mpi_r3.py uses all keywords - these are the defaults. User can change them here or override them when instantiating in __main__().
• Note: User should specify proper paths for saving output.
• Imported by runwrapper2017mpi_r3.py or runwrapper2017mpi_r4.py.

runwrapper2017mpi_r3.py and runwrapper2017mpi_r4.py

• Runs the simulations in a loop, does some summary output bookkeeping.
• Note: User should specify proper paths for saving output.

runwrapper2017mpi_r3x.py

• An example of a Unix executable version of runwrapper2017mpi_r3.py.
• Note: User should specify proper paths for the python executable and saving output.

There are two test files:

1. testOrderbook3.py
2. testTrader2017_r3.py

There are several Jupyter Notebooks that should work with minimal changes to the directories.

Written and tested with Python3.6