Testing for stationarity of a time series is important for certain trading strategies, particulary mean-reversion strategies. Stationarity means that the (rolling) mean and variance are fairly constant. There are various Python libraries that provide the necessary functionality. The example below demonstrates the use of the Augmented Dickey Fuller (ADF) test from the statsmodels package on a year’s worth of daily data for the ADABTC trading pair.
import pandas as pd
import numpy as np
import requests
import json
from datetime import datetime
from statsmodels.tsa.stattools import adfuller
root_url = 'https://api.binance.com/api/v1/klines'
symbol = 'ADABTC'
interval = '1d'
limit = '360'
url = root_url + '?symbol=' + symbol + '&interval=' + interval + '&limit=' + limit
data = json.loads(requests.get(url).text)
df = pd.DataFrame(data)
df.columns = ['date', 'open', 'high', 'low', 'close', 'v', 'close_time', 'qav',
'num_trades', 'taker_base_vol', 'taker_quote_vol', 'ignore']
df['date'] = [datetime.utcfromtimestamp(x/1000.0).date() for x in df.date]
df = df.set_index('date')
df['close'] = df['close'].astype(float)
# extracting only the close prices using values attribute of the DataFrame/Series
values = df['close'].values
# passing the extracted close prices to adfuller function.
res = adfuller(values)
# Printing the statistical result of the adfuller test
print('p-value: %f \n' % res[1])
print('Augmneted Dickey_fuller Statistic: %f \n' % res[0])
# printing the critical values at different alpha levels.
print('critical values at different levels:')
for k, v in res[4].items():
print('\t%s: %.3f' % (k, v))
The generally accepted level is 95% confidence level, in this case being a p value of 0.05 as the null hypothesis is that the series is not stationary.