Run More Risk Models Faster? … Maybe

Jim Goodnight, the co-founder and CEO of the SAS Institute has suggested (see Penny Crosman’s March 28, 2013 article “The Trouble with Banks’ Risk Models: Q&A with the Chief of SAS,” at ) that the problem with risk models is that they are not run enough times, asserting that a greater number of iterations of running the models will lead to greater accuracy of risk assessments. While this is a valid recommendation, if (and only if) the models and the data are adequate, no amount of iteration will improve the assessment of risk and the subsequent decision-making if the models are wanting and/or the data are inaccurate.

Journalist Mitch Ratcliffe is quoted as once saying “A computer lets you make more mistakes faster than any invention in human history—with the possible exceptions of handguns and tequila.” The same may well be true of risk models in general and those that are used to calculate trading risks in particular.

Mr. Goodnight is without doubt a great leader and innovator in the computer-based statistical analysis field and one should certainly listen when he touts the use of his company’s software to accelerate processing times of risk models. One might also agree with his philosophy that running a greater number of iterations of risk models can enhance the value of those models … but only if those models accurately depict reality. Running inaccurate models faster and more frequently can be very misleading and provides one with a false sense of precision, which in turn can lead to greater confidence in less reliable results.


  1. Kenneth F. Belva May 29, 2013 at 7:58 am | Permalink

    Warren – Good article. Your focus is on the accuracy of the risk model. Another perspective is that you can have a perfect model but, as the old saying goes, “Garbage In, Garbage Out.” The same reasoning in the article applies to GI,GO as well.

  2. Bill Bowes May 8, 2014 at 12:21 pm | Permalink

    A very thought provoking series of articles indeed.
    The concept that the risk model depends on the accuracy of the data on which the model is based is also very relevant. What however has not been brought into any of these equations is the human factor, the ability of the decision making process, being an infinitely variable one, which may, or more probably may not, depend on logical thought processes to reach subjective conclusions.
    Any computer model depends, even relies on everything and every stage being a defined constant, but any human brain can make a nonsense of these models.
    Just a thought!!

Post a Comment

Your email is never published nor shared. Required fields are marked *