VaRpy 0.0.3

Tools to backtest your VaR metric

To install, just use pip :

pip install varpy

Required Dependencies are listed below , such :

Dependency	Version
arch	5.0.1
numpy	1.20.1
scipy	1.6.2
pandas	0.12.2
numba	0.52.1
joblib	1.0.1
scipy	0.4
tabulate	3.3.4

There is no dependency verification , so please, make sure to have installed every required one before using the package.

Example

To begin, let’s extract default data:

import varpy
from varpy import EVT_VaR,Student_VaR,Normal_VaR
from varpy.Backtester.bktst import Backtest
from varpy.Backtester.time_Significance import Testing
import matplotlib.pyplot as plt

data = d1()* 100
data

Let’s compute our weekly standard VaR and CVaR

VaR,CVaR = Normal_VaR(data.values.reshape(-1,) ,0.05,7)
print(VaR,CVaR)

Let’s backtest our VaR, to evaluate its consistency throughout time

In each iteration, we choose to use a window of 500 data to evaluate our tail statistic. Additionally, our VaR is evaluated on a weekly basis for an alpha of 5%.

VaR , CVaR = Backtest(data,500,7,0.05,model = 'Gaussian')
ts = Testing(data,VaR,CVaR,500,0.05)
print(ts.summary)

Plot your VaR and CVaR

import matplotlib.pyplot as plt

fig = plt.figure(figsize=(15,5))
plt.plot(data[500:])
plt.plot(VaR)
plt.plot(CVaR)
plt.show()