Use historical data on earning days and measure volatility

Hi Nico Xenox,

We can use the standard deviation indicator warmed up using historical data to get the volatility for any security.

The information about the earnings day can be found in the fine fundamental data for the given security. Refer to the following doc for more information.

We can use the IndicatorExtensions helper methods to get the delayed volatility values. Refer to the following snippet:

self.delayedVol[symbol] = IndicatorExtensions.Of(Delay(7),self.vol[symbol])

Here self.vol is a dictionary keyed by the Symbol object and the STD indicator created using the constructor as the value to update the self.delayedVol we need to update the self.volRefer to the following example algorithm.

Best,

Varad Kabade 

I took some time off and tried the whole thing in the research notebook. Im was trying to archieve what i wrote in the first post by using two approaches: 

1. Get amount of days from T=0 for the lowest number of standarddeviation (T-30)

2. Get amount of days from T=0 in decending value

Thats what i got:

import pandas as pd
import numpy as np
import scipy
import matplotlib as mpl
import matplotlib.pyplot as plt
import math
from datetime import datetime, timedelta


# Set this between 150 to 300 for higher resolution graphs
mpl.rcParams['figure.dpi']= 300


# Load the QuantBook. The QuantBook is the main 
# access point you will use to request data from
# For more information see [https://www.quantconnect.com/docs/research/overview]
qb = QuantBook()
# Aliasing qb as `self` for portability purposes :)
self = qb
#setting time value
#msft_start = datetime(2015, 1, 1)
#msft_end = datetime(2018, 12 ,31)

bb = BollingerBands(10, 2, MovingAverageType.Exponential)
start_date = datetime(2017,1,1)
end_date = datetime(2018,1,1)
msft = self.AddEquity("MSFT", Resolution.Minute).Symbol
msft_10Q = qb.AddData(SECReport10Q, msft)
df1_bb = self.Indicator(bb, msft, 360, Resolution.Daily)
report_history = self.History(SECReport10Q, msft_10Q.Symbol, timedelta(days=360))

report_history.drop(['datasourceid','report'  ], axis=1, inplace=True)
df1_bb.drop(['bandwidth', 'percentb', 'bollingerbands','lowerband','middleband','price','upperband'  ], axis=1, inplace=True)
df1_bb = df1_bb.rename_axis('time')
report_history.index = report_history.index.droplevel('symbol')


'''
Going back 30 days and finding lowest value
'''


newone=[]
actual = []

for date in report_history.index:
    r = 0
    n = 1
    number= 5
    safe = 0
    date1 = date
    print(date1)
    while(r <= 30):
        
        while((date - timedelta(days=n)) not in df1_bb.index):
            
            n = n+1
            r=r+1
            
        if (df1_bb.loc[date - timedelta(days=n)]['standarddeviation'] < number):
            number = df1_bb.loc[date - timedelta(days=n)]['standarddeviation']
            date = date - timedelta(days=n)
            n=1
            r =r+1
            safe = r
            while((date - timedelta(days=n)) not in df1_bb.index):
                n = n+1
                r =r+1
        elif (df1_bb.loc[date - timedelta(days=n)]['standarddeviation'] > number):
            date = date - timedelta(days=n) 
            n=1
            r =r+1
            while((date - timedelta(days=n)) not in df1_bb.index):
                n = n+1
                r =r+1
    n=1  
    newone.append(safe)
    
display(newone)


average2 = [i for i in newone if i != 0]
print(average2)
average2 = sum(average2)/len(average2)
math.ceil(average2)


        
'''
Finding lowest value in descending order
'''

tryme = []

for date in report_history.index:
    i = 0
    n=1
    while((date - timedelta(days=n)) not in df1_bb.index):
        n = n+1
    while(df1_bb.loc[date]['standarddeviation'] > df1_bb.loc[date - timedelta(days=n)]['standarddeviation']):
        date = date - timedelta(days=n)
        n=1
        i = 1+i
        while((date - timedelta(days=n)) not in df1_bb.index):
            n = n+1

    tryme.append(i)
    print('list' +str(tryme))
    
average = [i for i in tryme if i != 0]
print(average)
average = sum(average)/len(average)
math.ceil(average)    

            
    

As you can see i tried it using just one stock and not a whole universe. I tried implementing it into QC but i dont know how to change this line of code.

df1_bb = self.Indicator(bb, msft, 360, Resolution.Daily)

Also if I import pandas and the other libraries into QC i simply can paste the code that i wrote into QC, right?

 

Hi Nico

df1_bb = self.Indicator(bb, msft, 360, Resolution.Daily)

This line won't work in a QCAlgorithm instance of a backtest. This is just a helper method of data visualization of indicator values in the research environment. You'll need a rolling window to store each updated indicator value in a backtest.

And you'll need modifications of the above code to put a backtest in production. Refer to the docs to do so.

Best
Louis

Thanks Guys
I will try that, I really appreciate the help. I'm doing a school project and with your help I'm slowly understanding more and more stuff.