ABIDES for dummies

In the next sections, we are going to consider a dummy simulation on the RMSC03 config for a total time of two hours. We are going to use the default configuration RMSC03 for this purpose, and in particular its derivation to achieve only two hours of trading simulation.

Run simulation

To run the simulation, run the following:

python -u abides.py -c rmsc03 -t ABM -d 20200603 -s 1234 -l rmsc03_two_hour

Basic plotting

To produce the basic plottings, run the following:

cd util/plotting && python -u liquidity_telemetry.py ../../log/rmsc03_two_hour/EXCHANGE_AGENT.bz2 ../../log/rmsc03_two_hour/ORDERBOOK_ABM_FULL.bz2 -o rmsc03_two_hour.png -c configs/plot_09.30_11.30.json && cd ../../

To assess the results check file ./abides/util/plotting/rmsc03_two_hour.png.

Evaluation

To produce the evaluation metrics on the simulation, create a new directory in ./abides/evaluation called rmsc03_two_hour and run the following:

cd realism && python -u asset_returns_stylized_facts.py -s ../log/rmsc03_two_hour -o ../evaluation/rmsc03_two_hour && cd ..

To assess the results check directory ./abides/evaluation/rmsc03_two_hour.

Parameters tweaking 🔧

Let's suppose we want to run the simulation on a complete trading day (09:30 to 16:30). In this section we are going to explain the parameters we need to tweak in order to do so.

RMSC03 config file

Open the script ./abides/config/rmsc03.py and change the parser argument --end_time and set its default to 16:30:00.

Telemetry config file

Open the directory ./abides/util/plotting/configs, create a new config called plot_09.30_16.30.json and paste the following into the json dictionary:

{
  "xmin": "09:32:00",
  "xmax": "16:30:00",
  "linewidth": 0.7,
  "no_bids_color": "blue",
  "no_asks_color": "red",
  "transacted_volume_binwidth": 30,
  "shade_start_time": "01:00:00",
  "shade_end_time": "01:30:00"
}

New commands

Simulation:

python -u abides.py -c rmsc03 -t ABM -d 20200603 -s 1234 -l rmsc03_all_day

Basic plotting:

cd util/plotting && python -u liquidity_telemetry.py ../../log/rmsc03_all_day/EXCHANGE_AGENT.bz2 ../../log/rmsc03_all_day/ORDERBOOK_ABM_FULL.bz2 -o rmsc03_all_day.png -c configs/plot_09.30_16.30.json && cd ../../

Evaluation:

cd realism && python -u asset_returns_stylized_facts.py -s ../log/rmsc03_all_day -o ../evaluation/rmsc03_all_day && cd ..

Known issues ⚠️

ValueError: low >= high

This error is due to some kind of overflow in the ValueAgent, when generating a random integer with NumPy. Open file ./abides/agent/ValueAgent.py and replace self.depth_spread*spread with np.ceil(self.depth_spread*spread) (line 202):

adjust_int = np.random.randint(0, self.depth_spread*spread)