jasonstrimpel/volatility-trading

A complete set of volatility estimators based on Euan Sinclair's Volatility Trading

volatility-trading's Issues

Question about the Value of Parkinson volatility estimator

As Euan Sinclair said in the book "Volatility Trading", we could use Parkinson, RogersSatchell alternative volatility estimator instead of close-to-close volatility estimator and campare the implied volatility with these realized volitilties. However, I found that the Parkinson and RogersSatchell estimators are much larger than the implied volatility.(As is shown in the y-axis in the graphs, they are bigger than 100%, while the close-to-close estimator is more reasonable).

I want to know if we could use alternative volatility estimators like Parkinson and compare the implied volatility with them to trade; or we must compare the the implied volatility with close-to-close volatility?

setup.py, Add pandas, numpy, and other package requirements

Add pandas, numpy, and other package requirements

setup( name='volatility-trading', version='0.0.1', url='https://github.com/jasonstrimpel/volatility-trading/', license='GPL-3.0-or-later', packages=['volatility','volatility/models'], install_requires=[ 'numpy~=1.21.4', 'pandas~=1.3.3', # ... other packages )

YangZhang

If use different window other than 30 days, the following should be revised:

# result = (open_vol + 0.164333 * close_vol + 0.835667 * window_rs).apply(np.sqrt) * math.sqrt(252)
k = 0.34 / (1 + (window + 1) / (window - 1))
result = (open_vol + k * close_vol + (1 - k) * window_rs).apply(np.sqrt) * math.sqrt(245)

We should use variables instead of constants.

ValueError: x and y must have same first dimension, but have shapes (719,) and (725,)

Hi, Jason,

I have run the sample code in a new Main.py file, however, it prompts the value error. Can you help me? Thank you.

"C:\Program Files\Python36\python.exe" D:/PythonWorkSpace/volatility-trading-master/main.py
C:\Program Files\Python36\lib\site-packages\statsmodels\compat\pandas.py:56: FutureWarning: The pandas.core.datetools module is deprecated and will be removed in a future version. Please use the pandas.tseries module instead.
from pandas.core import datetools
C:\Program Files\Python36\lib\site-packages\numpy\core\fromnumeric.py:56: FutureWarning: reshape is deprecated and will raise in a subsequent release. Please use .values.reshape(...) instead
return getattr(obj, method)(*args, **kwds)
Traceback (most recent call last):
File "D:/PythonWorkSpace/volatility-trading-master/main.py", line 48, in
normed
File "D:\PythonWorkSpace\volatility-trading-master\volatility\volest.py", line 715, in term_sheet
benchmark_compare_fig, benchmark_compare_plt = self.benchmark_compare(window=window)
File "D:\PythonWorkSpace\volatility-trading-master\volatility\volest.py", line 588, in benchmark_compare
cones.plot(date, x, label=self.bench_symbol)
File "C:\Program Files\Python36\lib\site-packages\matplotlib_init.py", line 1717, in inner
return func(ax, *args, **kwargs)
File "C:\Program Files\Python36\lib\site-packages\matplotlib\axes_axes.py", line 1372, in plot
for line in self._get_lines(*args, **kwargs):
File "C:\Program Files\Python36\lib\site-packages\matplotlib\axes_base.py", line 404, in _grab_next_args
for seg in self._plot_args(this, kwargs):
File "C:\Program Files\Python36\lib\site-packages\matplotlib\axes_base.py", line 384, in _plot_args
x, y = self._xy_from_xy(x, y)
File "C:\Program Files\Python36\lib\site-packages\matplotlib\axes_base.py", line 243, in _xy_from_xy
"have shapes {} and {}".format(x.shape, y.shape))
ValueError: x and y must have same first dimension, but have shapes (719,) and (725,)

k of Yang=Zhang estimator

Please, check line 34 on file YangZhang.py

Current code is
k = 0.34 / (1 + (window + 1) / (window - 1))

I believe the code should be
k = 0.34 / (1.34 + (window + 1) / (window - 1))

Adopt interface to take OLHC data as numpy arrays

Altering the interface to allow for numpy arrays of olhc prices would extend the functionality to non-equity tickers.

What the realized in the graph means

Hello! I want to know what the realized in the graph means. Is it the volatility of the last day in the period?

The equation of volatility

@jasonstrimpel Hi Jason, I find that the equation of GarmanKlass is different from that of RogersSatchell, as follows:

GarmanKlass: return trading_periods * v.mean()**0.5
RogersSatchell: return (trading_periods * v.mean())**0.5

Are you sure they are different? Thank you.

math domain errors on some models

The models which follow def f(v) methods seem not to calculate the vols correctly.

17 
 18     def f(v):

---> 19 return math.sqrt(252 * v.mean())
20
21 result = rs.rolling(window=window, center=False).apply(func=f)

ValueError: math domain error

Unable to pip the library

Am I missing something?

(envy1) C:\Users\User>pip install volest
ERROR: Could not find a version that satisfies the requirement volest
ERROR: No matching distribution found for volest

(envy1) C:\Users\User>pip install volatility-trading
ERROR: Could not find a version that satisfies the requirement volatility-trading
ERROR: No matching distribution found for volatility-trading

Missing LICENSE

I see you have no LICENSE file for this project. The default is copyright.

I would suggest releasing the code under the GPL-3.0-or-later or AGPL-3.0-or-later license so that others are encouraged to contribute changes back to your project.

AttributeError: 'Rectangle' Object Has No Property 'Normed' Solution

Apparently matplotlib no longer uses the "normed" property. This deprecated property appears in several places in this project, resulting in the error AttributeError: 'Rectangle' Object Has No Property 'Normed' Solution.

SyntaxWarning with python 3.9.6

volest.py:193,285,359,434,453,570,645: SyntaxWarning: "is" with a literal. Did you mean "=="?

Garman and Klass code

log_hl = (prices['Adj High'] / prices['Adj Low']).apply(np.log)
log_co = (prices['Adj Close'] / prices['Adj Open']).apply(np.log)
log_oc = (prices['Adj Open'] / prices['Adj Close']).apply(np.log)

just wondering how this log_oc fits?

tried to modify the code as per book, i.e.
log_hl = (prices['Adj High'] / prices['Adj Low']).apply(np.log)
log_co = (prices['Adj Close'] / prices['Adj Close'].shift(1)).apply(np.log)

this returns, may be because of negative number:
return math.sqrt(252 * v.mean())
ValueError: math domain error

Also, Rogers, Satchell, and Yoon works fine with following lines:
log_ho = (prices['Adj High'] / prices['Adj Open']).apply(np.log)
log_hc = (prices['Adj High'] / prices['Adj Close']).apply(np.log)
log_lo = (prices['Adj Low'] / prices['Adj Open']).apply(np.log)
log_lc = (prices['Adj Low'] / prices['Adj Close']).apply(np.log)

rs = log_hc * log_ho + log_lc * log_lo

Usage and application

Hello,
I've stumbled upon this library while looking for volatility models applicable to the stock market, however I'm a bit confused as to what the purpose of the repo is? Using a Parkinson estimator with SPY as benchmark and United States Steel Corp as the underlying symbol I get an R squared of 0.83. However looking through the code it appears that it is only fitting to the data and not doing any predictions.

My question is how do you apply the results that this repo gives in the real world and would it be correct to assume that these models can be used with the .predict() to get applicable results?

y = mlab.normpdf(bins, mean, std) would not work in the latest Python

In histogram function:
y = mlab.normpdf(bins, mean, std)

should change to

from scipy.stats import norm
y = norm.pdf(bins, mean, std)
this works fine.

great work.

thanks.

GarmanKlass

According to the original formula on the textbook, the codes should be revised as following:

# log_co = (price_data['Close'] / price_data['Open']).apply(np.log)
log_cc = (price_data['Close'] / price_data['Close'].shift(1)).apply(np.log)

rs = 0.5 * log_hl**2 - (2*math.log(2)-1) * log_cc**2 # Instead of log_co

Using today's close/open will underestimate the volatility caused by jump, however, if jumps always happen within a period, GarmanKlass will give a negative value to be rooted. Using the above formula based on the textbook:

def f(v):
    
    #print(245*v.mean())
    # return math.sqrt(252 * v.mean())
   
    return (245 * (v.mean()))
    # q = 245 * v.mean()
    # return q
    # return math.sqrt(v.mean())
    # return math.sqrt(245 * v.mean())
    # China Market has about 245 trading days
    # If we use intraday data, there should be 240 bars or more

we got:

The second pic is JPM's price during June, 2016. We see that too many jumps result in a negative variance. But Jason's formula won't since abs(log(close/open)) is smaller.

BTW, v ** (1/2) will not raise an error for a negative number while math.sqrt() does.

Unable to pass in trading periods

Every estimator can accept a parameter trading_periods (probably should be named annual_trading_days), however when calling it in on the of the VolatilityEstimator methods, there is no way to pass that parameter through. This will cause issues for markets such as FX, crypto and basically anything other than U.S. Equities.

Simple solution is to allow it to be passed through as an option parameter.

Minor edits to run code in Python 3 from volatility directory

volest.py changed to: import models. It was: from volatility import models
models init.py changed to: from models.api import *

data is in tests directory thus: data_file_path = '../tests/JPM.csv'

I would add 2 explicit lines after vol. I needed to reference fig, plt not just plt.
fig, plt = vol.cones(windows=windows, quantiles=quantiles)
plt.show()

I could not get vol.term_sheet to work.

Unable to run example in Jupyter notebook

Installed locally after cloning repo using python setup.py install. Trying to run the sample code in a notebook. Using python 3.8.

from volatility import volest
from volatility import data

# data
symbol = 'JPM'
bench = '^GSPC'
data_file_path = '../JPM.csv'
bench_file_path = '../BENCH.csv'
estimator = 'GarmanKlass'


# estimator windows
window = 30
windows = [30, 60, 90, 120]
quantiles = [0.25, 0.75]
bins = 100
normed = True

# use the yahoo helper to correctly format data from finance.yahoo.com
jpm_price_data = data.yahoo_helper(symbol, data_file_path)
spx_price_data = data.yahoo_helper(bench, bench_file_path)

# initialize class
vol = volest.VolatilityEstimator(
    price_data=jpm_price_data,
    estimator=estimator,
    bench_data=spx_price_data
)

Error thrown

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-15-890e41b22558> in <module>
----> 1 from volatility import volest
      2 from volatility import data
      3 
      4 # data
      5 symbol = 'JPM'

<frozen zipimport> in load_module(self, fullname)

~/Dev/hermes/lib/python3.8/site-packages/volatility_trading-0.0.1-py3.8.egg/volatility/volest.py in <module>
     10 from matplotlib.backends.backend_pdf import PdfPages
     11 
---> 12 from volatility import models
     13 
     14 ESTIMATORS = [

ImportError: cannot import name 'models' from 'volatility' (/Users/prabhakar/Dev/hermes/lib/python3.8/site-packages/volatility_trading-0.0.1-py3.8.egg/volatility/__init__.py)

HodgesTompkins

I feel that HT should be used in all other models, not only for adjusting close-to-close volatility. Overlapping data are used for all models when using rolling windows.