The creation of the conda enviornment as shown in the readme fails on windows with the last version of conda (conda 4.5.11).

conda env create -f py4fi_conda.yml

Solving environment: failed

ResolvePackageNotFound:
  - ncurses==6.0=hd04f020_2
  - requests==2.18.4=py36h4516966_1
  - xlsxwriter==1.0.2=py36h3736301_0
  - tornado==4.5.2=py36h468dda9_0
  - itsdangerous==0.24=py36h49fbb8d_1
  - cycler==0.10.0=py36hfc81398_0
  - jinja2==2.9.6=py36hde4beb4_1
  - gmp==6.1.2=hb37e062_1
  - tk==8.6.7=h35a86e2_3
  - lzo==2.10=hb6b8854_1
  - testpath==0.3.1=py36h625a49b_0
  - appnope==0.1.0=py36hf537a9a_0
  - setuptools==36.5.0=py36h2134326_0
  - cffi==1.10.0=py36h880867e_1
  - jupyter==1.0.0=py36h598a6cc_0
  - numba==0.35.0=np113py36_6
  - pandocfilters==1.4.2=py36h3b0b094_1
  - pyqt==5.6.0=py36he5c6137_6
  - jupyter_core==4.3.0=py36h93810fe_0
  - cryptography==2.0.3=py36h22d4226_1
  - expat==2.2.5=hb8e80ba_0
  - chardet==3.0.4=py36h96c241c_1
  - patsy==0.4.1=py36ha1b3fa5_0
  - zeromq==4.2.2=ha360ad0_2
  - jedi==0.10.2=py36h6325097_0
  - zlib==1.2.11=hf3cbc9b_2
  - sqlite==3.20.1=h7e4c145_2
  - traitlets==4.3.2=py36h65bd3ce_0
  - python==3.6.3=h5ce8c04_4
  - mpc==1.0.3=hc455b36_4
  - werkzeug==0.12.2=py36h168efa1_0
  - matplotlib==2.1.0=py36h5068139_0
  - pexpect==4.2.1=py36h3eac828_0
  - wheel==0.29.0=py36h3597b6d_1
  - html5lib==0.999999999=py36h79312fd_0
  - pysocks==1.6.7=py36hfa33cec_1
  - libedit==3.1=hb4e282d_0
  - jpeg==9b=he5867d9_2
  - mpmath==0.19=py36h9185fea_2
  - nbconvert==5.3.1=py36h810822e_0
  - xlwt==1.2.0=py36h5ad1178_0
  - pyzmq==16.0.2=py36h087ffad_2
  - fastcache==1.0.2=py36h8606a76_0
  - pytz==2017.2=py36h2e7dfbc_1
  - pyopenssl==17.2.0=py36h5d7bf08_0
  - nbformat==4.4.0=py36h827af21_0
  - pcre==8.41=hfb6ab37_1
  - gettext==0.19.8.1=h15daf44_3
  - libiconv==1.15=hdd342a3_7
  - urllib3==1.22=py36h68b9469_0
  - certifi==2017.7.27.1=py36hd973bb6_0
  - bleach==2.0.0=py36h8fcea71_0
  - ipython_genutils==0.2.0=py36h241746c_0
  - numexpr==2.6.2=py36h8fc668d_2
  - prompt_toolkit==1.0.15=py36haeda067_0
  - widgetsnbextension==3.0.2=py36h91f43ea_1
  - hdf5==1.10.1=ha036c08_1
  - sympy==1.1.1=py36h7f3cf04_0
  - qt==5.6.2=h9975529_14
  - flask==0.12.2=py36h5658096_0
  - ipykernel==4.6.1=py36h3208c25_0
  - libffi==3.2.1=h475c297_4
  - gmpy2==2.0.8=py36h7ef02cb_1
  - freetype==2.8=h12048fb_1
  - asn1crypto==0.22.0=py36hb705621_1
  - xlrd==1.1.0=py36h336f4a2_1
  - intel-openmp==2018.0.0=h68bdfb3_7
  - pymc3==3.2=py36h1e7238b_0
  - openpyxl==2.4.8=py36he899640_1
  - readline==7.0=hc1231fa_4
  - pip==9.0.1=py36h1555ced_4
  - libgfortran==3.0.1=h93005f0_2
  - pycparser==2.18=py36h724b2fc_1
  - libgpuarray==0.6.9=0
  - python-dateutil==2.6.1=py36h86d2abb_1
  - six==1.11.0=py36h0e22d5e_1
  - libsodium==1.0.13=hba5e272_2
  - numpy==1.13.3=py36h2cdce51_0
  - simplegeneric==0.8.1=py36he5b5b09_0
  - ptyprocess==0.5.2=py36he6521c3_0
  - wcwidth==0.1.7=py36h8c6ec74_0
  - ipython==6.1.0=py36hf612aae_1
  - bzip2==1.0.6=h649919c_2
  - icu==58.2=h4b95b61_1
  - qtconsole==4.3.1=py36hd96c0ff_0
  - click==6.7=py36hec950be_0
  - jupyter_client==5.1.0=py36hf6c435f_0
  - libcxx==4.0.1=h579ed51_0
  - libpng==1.6.32=h6184301_3
  - dbus==1.10.22=h50d9ad6_0
  - terminado==0.6=py36h656782e_0
  - mako==1.0.7=py36h55379d4_0
  - pyparsing==2.2.0=py36hb281f35_0
  - pytables==3.4.2=py36hfbd7ab0_2
  - et_xmlfile==1.0.1=py36h1315bdc_0
  - nose==1.3.7=py36h73fae2b_2
  - markupsafe==1.0=py36h3a1e703_1
  - tqdm==4.19.4=py36he502594_0
  - entrypoints==0.2.3=py36hd81d71f_2
  - h5py==2.7.0=py36h6400cee_1
  - pandoc==1.19.2.1=ha5e8f32_1
  - scikit-learn==0.19.1=py36hffbff8c_0
  - decorator==4.1.2=py36h69a1b52_0
  - notebook==5.2.1=py36h640abe8_0
  - mkl==2018.0.0=h5ef208c_6
  - cython==0.26.1=py36hd51f8eb_0
  - jupyter_console==5.2.0=py36hccf5b1c_1
  - pandas==0.21.0=py36hfed917e_1
  - jsonschema==2.6.0=py36hb385e00_0
  - mkl-service==1.1.2=py36h7ea6df4_4
  - jdcal==1.3=py36h1986823_0
  - ipywidgets==7.0.0=py36h24d3910_0
  - glib==2.53.6=h33f6a65_2
  - ca-certificates==2017.08.26=ha1e5d58_0
  - pygments==2.2.0=py36h240cd3f_0
  - sip==4.18.1=py36h2824476_2
  - xz==5.2.3=h0278029_2
  - mpfr==3.1.5=h7fa3772_1
  - openssl==1.0.2m=h86d3e6a_1
  - webencodings==0.5.1=py36h3b9701d_1
  - mistune==0.8.1=py36h638d0ca_0
  - pickleshare==0.7.4=py36hf512f8e_0
  - scipy==1.0.0=py36h1de22e9_0
  - libcxxabi==4.0.1=hebd6815_0
  - idna==2.6=py36h8628d0a_1

i'm pretty sure it's a version mismatch - but couldn't figure out a way to fix it

according to an issuei found (https://github.com/ContinuumIO/anaconda-issues/issues/9480#issuecomment-394873967) exporting via conda env export --no-builds > environment.yml should help fix the issue.

Rolling.std(ddof=1, *args, **kwargs) calculates rolling standard deviation, normalized by N-1 by default. It means, by default, it uses the sample standard deviation function, in which the denominator is a N-1. The denominator can be set to 1 in this case, if std(ddof=251), then N-251=1. I guess the multiplying of math.sqrt(252) here may be a mistake of confusing rolling.std with numpy.std. The default ddof in numpy.std is 0, which means its denominator is a N. In that case, multiplying of math.sqrt(252) can offset the denominator of 252 exactly.

Hi @yhilpisch, the fixed for #4 by setting the first time slice (t=0) to 0 does not correctly fixed the issue. On first look, it appears that it has solved the issue. However, on closer inspection of the equation, I notice that even though setting the first time slice (t=0) to 0, a drift will still be included at time slice resulting in an over-estimation.

Hi @yhilpisch,

Thank you for the great work as an undergraduate student I have learn a lot from your book!

Example 3-4 code which estimates the theoretical value of a European call option via Monte Carlo simulation appears to yield different results from that of bsm_call_value() function in Chapter 3.

Using the analytical formula for the valuation of European call option in BSM model for S0 = 100, K = 105, T = 1.0, r = 0.05 and sigma = 0.2, we arrive at a European call price of 8.02135.

Using the code from example 3-4 which estimates the theoretical value of a European call option via Monte Carlo simulation using the same parameter should yield the same result as that of the analytical formula. However, the results yield by mcs_full_vector_numpy.py appears to be significantly different from that of bsm_call_value().

My suspect is that random.standard_normal((M + 1, I)) should be changed to random.standard_normal((M, I)) such that the calculation converge to the true price.

If I understand it right, the Vega of an option, the derivative of the norm cdf is needed. The notation you used in your book is N'(d_1), with a prime. The calculation in bsm_vega has used:

 stats.norm.cdf( ) 

which doesn't appear to be correct. Shouldn't this be just the norm pdf? ie. exp(-x^2) normalized?

The API seems not working. I wonder whether I've got it wrong? This is URL I get from frombook, but the page is empty http://hopey.netfonds.no/posdump.php?date=20140922&paper=AAPL.O&csv_format=csv

This is URL I get from IPython. No data, only column name http://www.netfonds.no/quotes/posdump.php?date=20140922&paper=NKE.N&csv_format=csv

Do anyone figure out the solution ?