Analytic Decision-Making Is Worse than Random More Often Than You Think!

Scientific thought and literatureĀ sometimes contains disturbing concepts. A number of papers have shown, for example, that on financial markets, random stock picking is better than informed choices made by any “professional” investor.

Decision making: should you not throw dices instead of over-analyzing?
Decision making: should you not throw dices instead of over-analyzing?

Here’s another fun example of such study: how Orlando the cat beat professionals in a contest. The article goes on to explain that it is not an isolated study. And that in fact, those investors which we admire because of the high return on their investment they generate, might just have been lucky (over the short period where we observe them). As the article mentions, Daniel Kahneman, Nobel prize in economics, showed over a sample that “the correlation between those fund managers who were successful in one year and those who were successful in the next year was close to zero (0.01) over eight years“.

Financial markets are the archetype of the complex system. In such a system it seems that random decision-making is better then (excessive) analysis.

When I presented this idea to a room full of engineers a few weeks ago they probably took me for a fool. Yet as the complexity of our world increases, it is not certain that the best way to take the right decisions is to enhance further our analytic models, which will be less and less representative of reality.

Have some dices in your pockets. In complex situations, throwing them might be the best decision-making method!

  • Anne Bard

    read the blog of Philippe Silberzahn about non predictive strategy and in particular his post “We have met the enemy and he is, er, forecasting”. He deals as well with the fallacy of analytic predictions and proposes other models.

    • Hi Anne, thanks for the reference, it is a great post. Fully agree with its contents too!

  • Michael Ingenbleek

    Hi Jeremy,

    How are you? Trust all is well?
    Just a quick note.

    Interesting read, got me thinking which is
    a good thing I suppose. I can imagine however that (pending on what kind of
    engineers) the engineers in question come from a more empirical positivist
    background and would need to rely on quantitative data. With regards to the Zilberzahn
    & Jones article referred by Anne Bard, I am not too familiar with
    philosophical stances. The authors appear to indicate a hermeneutic approach to
    data however wish a pragmatism outcome of forecasting?

    Arguably forecasting requires assumptions. The dynamic market theory
    of De Jong (1989) identifies the limitations of assumptions (unless research
    reproduced in a laboratory). Evidence in the literature suggests that ceteris
    paribus is utilized to control the extensive amount of variables (unless
    stochastic programming is your field of expertise). Which raises the question
    how good are the assumptions in question? And to what extend could scenario
    analysis cover these assumption gaps?

    Look forward to your thoughts.

    Best of wishes,


    • Hi Michael
      Good to hear from you, and that we got you thinking šŸ™‚
      Personally I am more and more convinced that a lot of the forecasting you can see everywhere (about the stock market, the economy, the next election etc) is just pure BS because nobody can really understand how variables are connected and there are too many to remain predictable. Of course there are periods of relative stability where people dream they can analyze and predict. But since I have discussed with traders that all assume that the future is not predictable anyway and that analytics like graphical analysis etc which try to trend the past are of very limited value, I have become more an more pessimistic about forecast.

      Probably a good teaching to enjoy the present moment!