Skip to content

CSS Value at Risk (CSS VaR): Improving Upon Popular Portfolio Optimization Procedures

October 7, 2010

David Varadi with Henry Bee

Having spent the bulk of research efforts at CSS on alpha-generating strategies, we decided to allocate some time towards the burgeoning  field of portfolio allocation and risk-management technology. In this case our approach to the problem was similar to what we have typically done in the past: 1) question key assumptions underlying conventional theory 2) modify unsuitable assumptions to account for the “real-world” of financial markets. Surprisingly these modifications were especially prominent in improving conventional optimization procedures. In this case our goal was to have higher returns with less risk and more importantly–a lower probability of a major drawdown. The “Value-at-Risk” framework was immediately appealing and we created an algorithm around this concept. In this test we used identical return inputs into each model, and thus we did not try to improve optimization through superior return inputs. Instead we focused on the engineering concept itself. We compare our model– CSS Value at Risk or CSS VaR using the industry version of CVaR and conventional Markowitz mean-variance optimization. So far in our testing, our method is universally superior due to some key properties that make it better suited to financial data. In this test, we used five different major asset classes via ETFs with the three different optimization methods. The result was superior return relative to risk and a lower maximum drawdown to the other methods. While this may not seem like an exciting result to traders, it is important to consider that using for example good systems or relative strength as inputs for expected returns would substantially boost the performance of all three models. Ultimately most of us pay too little attention to portfolio allocation and risk management, when this is the only true “free lunch” in financial markets.

Advertisements
7 Comments leave one →
  1. October 7, 2010 7:12 am

    “Ultimately most of us pay too little attention to portfolio allocation and risk management, when this is the only true “free lunch” in financial markets.”

    So true! (and well done for improving on the standard methodologies)

    Have you investigated (with any success) some concepts from Ralph Vince (ie Leverage Space Model) for portfolio allocation?
    That paper is a good intro to it:
    http://parametricplanet.com/rvince/article.pdf

  2. prazor permalink
    October 7, 2010 2:36 pm

    Nice work!

    Still one of the best quant blogs!

    Was this a walk forward optimization?
    Daily data?
    Period for reallocation?

    How does the transaction cost compare btw the methods?

    Thanks!

  3. Bgpl permalink
    October 8, 2010 5:17 pm

    hi David, Henry,
    as in the previous few posts, are you using the rolling 3 years data to do a mean-variance optimization, and your optimization as well ?
    rgds
    bgpl

  4. Carl permalink
    October 19, 2010 9:59 pm

    So, is an Amibroker CSSVaR on its way? 😎

Trackbacks

  1. News Nodules: Offerings At the Altar of Information | TraderDaily
  2. Weekend Reads « Derek Hernquist
  3. ETF Rewind Pro – 10.07.10

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: