openmargin 0.0.7

Arrow Markets repository for calculation margin requirements for crypto option portfolios.

Open Margin

Arrow Markets open source repository to compute margin requirements for risky asset portfolios, including crypto options portfolios.

Overview

This repository uses VAR/CVAR methodology to compute the necessary margin required for a portfolio of BTC or ETH options. Negative values refer to the cash margin needed for a given portfolio, while positive margin values count as credits. For further details, please refer to the white paper "Automated Margin Systems in Practice".

Installation

pip install openmargin

Usage

Open Margin calculates the risk of an options portfolio through RiskCalc. A compatible ticker and an options portfolio are the minimum inputs needed to generate the corresponding margin values. Providing the options data as a dataframe to Open Margin substantially reduces margin computation time.

import datetime
import pandas as pd

ticker = "eth"

expiration = datetime.datetime.strptime('2024-09-27 08:00:00', "%Y-%m-%d %H:%M:%S")
strike = 7000.0
kind = "C"
position = -1
portfolio = {'expiration': expiration, 'strike': strike, 'kind': kind, 'position': position}
portfolio = pd.DataFrame([portfolio],columns = ["expiration", "strike", "kind", "position"])

Once the ticker and the portfolio data are ready, margin can be computed with default parameters:

from openmargin.risk import RiskCalc
risk_calculator = RiskCalc(ticker, portfolio)
risk_calculator.get_margin()

RiskCalc has four pillars that can be modified to generate the required margin. All of the examples below assume that minimum inputs and corresponding package imports are complete.

Options Data

Open Margin uses the current options data for BTC and ETH from Deribit. To provide a different dataset, please provide a pandas dataframe containing options data similar to the one shown below:

To download and save the most recent options data:

from openmargin.auxiliary import deribit_option_data

utcnow = datetime.datetime.now(tz=datetime.timezone.utc)
(expiration_datetimes, yearly_times_to_expiration, strikes, log_moneynesses, contract_types, spot_prices, yearly_mark_implied_volatilities, mark_prices, prices) = deribit_option_data(ticker, utcnow)
options_data = pd.DataFrame([spot_prices, expiration_datetimes, yearly_times_to_expiration, strikes, log_moneynesses, contract_types, yearly_mark_implied_volatilities, mark_prices, prices]).T
options_data.columns = ['spot', 'expiration','tte','strike', 'log_money', 'kind','mark_iv','mark_price', 'price']
options_data = options_data.reset_index(drop=True)
options_data.to_csv(f'options_data.csv', index = False)

To load a cahced option dataset and get the corresponding margin values:

options_data = pd.read_csv('options_data.csv', parse_dates=['expiration'])
risk_calculator = RiskCalc(ticker, portfolio, options_data)
risk_calculator.get_margin()

Risk Configuration

Three main risk factors are configured under RiskConfig. Interest rate, sampling frequency and steps are the key parameters currently allowed for user customization. Sampling frequency multiplied by steps determine the trading horizon that will be used to calculate margin values.

from openmargin.risk import RiskConfig, RiskCalc

r = 0.02
sampling_frequency = 4 # hours
steps = 6
risk_params = RiskConfig(r, sampling_frequency, steps)

risk_calculator = RiskCalc(ticker, portfolio, risk_params = risk_params)
risk_calculator.get_margin()

Risk Model

RiskModel class determines the margin method used. Open Margin only allows VAR type margin methods in this public version. To select between VAR and CVAR and to change margin threshold:

from openmargin.risk import VAR, RiskModel, RiskCalc

var_type = 'VAR'
var_threshold = 0.05
margin_method = VAR(var_type, var_threshold)
risk_model = RiskModel(margin_method)

risk_calculator = RiskCalc(ticker, portfolio, risk_model = risk_model)
risk_calculator.get_margin()

Price Paths

Path dependant methods like VAR/CVAR depend on future price paths. 1,000 paths are generated by default by PricePathGenerator. Changes to the default number of paths can be made via:

from openmargin.auxiliary import get_underlier_price
from openmargin.risk import PricePathGenerator, RiskCalc

spot = get_underlier_price(ticker)
number_of_paths = 10_000
price_paths = PricePathGenerator(ticker, risk_params, spot, number_of_paths)

risk_calculator = RiskCalc(ticker, portfolio, price_paths = price_paths)
risk_calculator.get_margin()

The paths can be generated with:

sample_paths = price_paths.generate_paths()

Further customization examples can be found under example.py.

Clone Repo Directly

The repo can also be cloned directly. This is advisable for doing work on the openmargin repo itself. Please see our guidelines for submitting pull requests. Community contributions are welcome!

git clone https://github.com/Arrow-Markets-Research/openmargin
cd openmargin

venv

Initialize a virtual environment (Windows)

python3 -m venv venv
venv/Scripts/activate.bat

Dependencies

From the venv, run

pip install -r requirements.txt

Create a test file in the src/openmargin directory called om-test.py:

import datetime
import pandas as pd
from openmargin.risk import RiskCalc

ticker = "eth"

expiration = datetime.datetime.strptime('2024-09-27 08:00:00', "%Y-%m-%d %H:%M:%S")
strike = 7000.0
kind = "C"
position = -1
portfolio = {'expiration': expiration, 'strike': strike, 'kind': kind, 'position': position}
portfolio = pd.DataFrame([portfolio],columns = ["expiration", "strike", "kind", "position"])

risk_calculator = RiskCalc(ticker, portfolio)
m = risk_calculator.get_margin()

print("portfolio:\n", portfolio, "\nspot:", float(risk_calculator.portfolio.spot), "\ninitial margin:", m[0])

Now, to compute the intitial margin for the chosen portfolio, switch to the src directory and run

python3 -m openmargin.om-test