Crypto advanced strategy

²⁰⁵³ views

Kernel: Python 3 (Ubuntu Linux)

Notebook Instructions

You can run the notebook document sequentially (one cell at a time) by pressing shift + enter. While a cell is running, a [*] will display on the left. When it has been run, a number will display indicating the order in which it was run in the notebook [8].

Enter edit mode by pressing Enter or using the mouse to click on a cell's editor area. Edit mode is indicated by a green cell border and a prompt showing in the editor area.

This course is based on specific versions of python packages. You can find the details of the same in this manual. The manual also explains how to use these codes in other versions of python packages.

Hurst exponent strategy

In this notebook, we will create a strategy using the Hurst exponent and the RSI. The strategy works as follows:

Fetch the minute data for Ethereum/USDT from a csv file
Convert the unix epoch time to normal datetime format and set it as index
Compute the hurst value for the entire price series
Compute the rolling hurst values for with a lookback period of 240 minutes
Compute the signals to indicate the persistent nature of the market
Calculate the RSI values
Calcualte the market returns
Calcualte the strategy returns
Compute the cumulative strategy return and plot it
Compute the slippage
Calculate the net profit after slippage

Import data and libraries

In [3]:

import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
#import talib as ta
from hurst import compute_Hc

Out[3]:

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
<ipython-input-3-53613acd3550> in <module>()
      3 import pandas as pd
      4 #import talib as ta
----> 5 from hurst import compute_Hc

ModuleNotFoundError: No module named 'hurst'

In [4]:

# If you do not have the hurst library please install it by running the below code
!pip install hurst

Out[4]:

Collecting hurst
  Retrying (Retry(total=4, connect=None, read=None, redirect=None, status=None)) after connection broken by 'ConnectTimeoutError(<pip._vendor.urllib3.connection.VerifiedHTTPSConnection object at 0x7fa1cb878a50>, 'Connection to pypi.org timed out. (connect timeout=15)')': /simple/hurst/
  Retrying (Retry(total=3, connect=None, read=None, redirect=None, status=None)) after connection broken by 'ConnectTimeoutError(<pip._vendor.urllib3.connection.VerifiedHTTPSConnection object at 0x7fa1cb878c10>, 'Connection to pypi.org timed out. (connect timeout=15)')': /simple/hurst/
  Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'ConnectTimeoutError(<pip._vendor.urllib3.connection.VerifiedHTTPSConnection object at 0x7fa1cb878050>, 'Connection to pypi.org timed out. (connect timeout=15)')': /simple/hurst/
  Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'ConnectTimeoutError(<pip._vendor.urllib3.connection.VerifiedHTTPSConnection object at 0x7fa1cb878610>, 'Connection to pypi.org timed out. (connect timeout=15)')': /simple/hurst/
  Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'ConnectTimeoutError(<pip._vendor.urllib3.connection.VerifiedHTTPSConnection object at 0x7fa1cb878590>, 'Connection to pypi.org timed out. (connect timeout=15)')': /simple/hurst/
  Could not find a version that satisfies the requirement hurst (from versions: )
No matching distribution found for hurst

In [3]:

df1=pd.read_csv("1min_ETHUSDT-Upload.csv").iloc[-100000:]
# Set the timestamp to be the index for the data
df1=df1.set_index('Timestamp')

Convert the unix time stamp to datetime

In [4]:

df1.index=pd.to_datetime(df1.index, unit='ms')

Compute the hurst value for the data

In [5]:

series=df1.Close
H = compute_Hc(series, kind='price')[0]

In [6]:

print("Hurst Exponent H={:.4f}".format(H))

Out[6]:

Hurst Exponent H=0.6204

In [7]:

# Let us visualize the data to confirm if it really was trending for the given time frame
plt.figure(figsize=(10,7))
plt.plot(df1.Close)

Out[7]:

[<matplotlib.lines.Line2D at 0x22384e05c18>]

Compute the rolling hurst values

In [8]:

df1['H']=0

We are using the past 4 hour data to calculate the hurst exponent.

In [9]:

for i in range(240,len(df1)):
    df1.H.iloc[i]=compute_Hc(df1.Close.iloc[i-240:i-1], kind='price')[0]

Out[9]:

C:\Users\udish\Anaconda3\envs\py35\lib\site-packages\pandas\core\indexing.py:189: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
  self._setitem_with_indexer(indexer, value)

In [10]:

# Let us plot the hurst values to visualize it
# We will be using threshold values of 0.65 and 0.35 to avoid trading for the noise
plt.figure(figsize=(10,7))
df1.H.plot()
plt.axhline(0.35,color='g')
plt.axhline(0.65,color='r')

Out[10]:

<matplotlib.lines.Line2D at 0x223871e3710>

Generate the persitence signals

In [11]:

df1['Signal']= 0
df1['Signal']=np.where( df1.H>0.65,1,df1.Signal)
df1['Signal']=np.where( df1.H<0.35,-1,df1.Signal)

Compute the RSI

In [12]:

df1['RSI'] = ta.RSI(df1.Close.shift(1).values,14)

Compute the market return

In [13]:

# Since we use the previous Close price to calculate the RSI and Hurst values, we use the close prices to calculate
# the market returns during the execution of the strategy.
df1['Return']=df1.Close.pct_change()

Compute the strategy returns

We use RSI here to gauge the overbought/oversold condition of the market.

We buy when

RSI is more than 75 and Persistence signal is 1
RSI is less than 25 and Persistence signal is -1

We sell when

RSI is more than 75 and Persistence signal is -1
RSI is less than 25 and Persistence signal is 1

In [14]:

df1['StrReturn']=0
df1['StrReturn']= np.where(((df1.RSI>75) & (df1.Signal==-1)),-df1.Return,df1.StrReturn)
df1['StrReturn']= np.where(((df1.RSI<25) & (df1.Signal==-1)),df1.Return,df1.StrReturn)
df1['StrReturn']= np.where(((df1.RSI>75) & (df1.Signal==1)),df1.Return,df1.StrReturn)
df1['StrReturn']= np.where(((df1.RSI<25) & (df1.Signal==1)),-df1.Return,df1.StrReturn)

Compute the total slipapge cost

In [15]:

# When we execute a strategy based on market orders our strategy can experience a significant slippage,
# if the bid-ask prices are far away. So, we have taken an average bid-ask spread of 0.05 based on the orderbook 
# data from cryptoexchanges.

total_slippage_cost=len(df1[df1.StrReturn!=0])*0.05/df1.Close.mean()
total_slippage_cost

Out[15]:

0.12242743305187771

In [16]:

Cumulative_Returns=df1.StrReturn.expanding().sum()

Plot the cumulative returns of the strategy

In [17]:

plt.figure(figsize=(10,7))
Cumulative_Returns.plot()

Out[17]:

<matplotlib.axes._subplots.AxesSubplot at 0x2238866e400>

Compute the net profit after slippage

In [18]:

Final_percentage_retuns=(Cumulative_Returns.iloc[-1]-total_slippage_cost)*100

In [19]:

Final_percentage_retuns

Out[19]:

6.805456409238202