• C++ Programming for Financial Engineering
    Highly recommended by thousands of MFE students. Covers essential C++ topics with applications to financial engineering. Learn more Join!
    Python for Finance with Intro to Data Science
    Gain practical understanding of Python to read, understand, and write professional Python code for your first day on the job. Learn more Join!
    An Intuition-Based Options Primer for FE
    Ideal for entry level positions interviews and graduate studies, specializing in options trading arbitrage and options valuation models. Learn more Join!

Analyse Gold Returns Distribution

Hi All


Just want to take your views if you have only Investment in Gold and you want to analyse maximum probable loss in one day , is VaR good for estimation maximum probable loss in 1 day keeping view in mind of Normal Distribution? Please reply soon

good thing is that with gold, we have historical data going back several decades. So the simplist way to estimate the maximum probable 1-day loss (in % terms) is to use the percentile statistics. Look up the bottom 1% percentile historical daily return, and that will be your historical VaR @ 99%.
and why do we believe in VAR as a measure of risk? What if you take the biggest daily loss to have ever occured and add 20%. Should work just as good

Bastian Gross

German Mathquant
Expected Shortfall

I agree with RussianMike on dangerous V@R-measurement.
In fact V@R is NOT a risk measure, I recommend that you should estimate maximum probable loss in 1d keeping view in mind of normal distribution, by compute the expected shortfall also referred to as conditional-V@R. See perhaps this post! Or here.

If you're able to estimate V@R, your expected shortfall estimation will be easy.
If q is your given threshold, you'll compute V@R by solving Normdist(x < V@R) = q and afterwards estimate your expected shortfall ESq = E(x | x < V@R).

In the case of normal distribution:

(X = N(\mu , \sigma^{2}) \\ Value at Risk_{q} = \mu + Quantil_{1-q}\sigma \\\Rightarrow\\ ES_{q} = \mu - \frac{normpdf(Quantil_{1-q})}{q}\sigma)

where (Quantil_{1-q}) is the 1-q-quantil of the standard normal distribution and
normalpdf is the probability density function
(normalpdf(x) = \frac{1}{\sqrt{2\pi}}e^{-\frac{x^{2}}{2}} )


  • Empirical Study of Value-at-Risk and Expected Shortfall Models.pdf
    230.2 KB · Views: 59
  • Expected Shortfall.pdf
    377.5 KB · Views: 54

Bastian Gross

German Mathquant
Here is a list of quantils:

0.8416 for 20% treshold
1.2816 for 10%
1.6449 for 5%
2.0537 for 2%
2.3263 for 1%
3.0902 for 0.1%
Hi Bastian Gross and All thanks for your key views about VaR and so can you plz help in using block Maxima approximation for extreme value theory for Gold Risk Analysis how do we do that please tell me in step by step manner in Excel.

Bastian Gross

German Mathquant

This is a simple and non historical data-sheet:


  • Simple_ExpectedShortfall_____Normaldistribution.xls
    65.5 KB · Views: 83
Thanks !!! alot for giving me knowledge

but now also through some light on block maxima approximations and how do we do that in excel

Bastian Gross

German Mathquant
Block Maxima Method

Okay, block maxima method is an other algorithm. I don't know much about this.

This paper could be helpful:


  • Block_Maxima_Method.pdf
    289.5 KB · Views: 42
Thanks man!!:)
I need favor from you will you help me in my research project and i would like to hear views about emerging economies , and Financial Innovation any suggetion do you have for country like Pakistan where financial innovation in its initial phase

particularly if non performing loans reached to more higher levels beyond expectation will credit derivatives play any role here and how?


Bastian Gross

German Mathquant
This is a simple and historical data-sheet:


  • Expected_Shortfall_historical_data.xls
    745.5 KB · Views: 70