Is It Possible To Make Predictions?

ScreenHunter_447 Jul. 25 20.28

Prediction of the future has always been man’s dream. However, there is an overwhelming amount of physical evidence that this is quite impossible. This is because the future is permanently under construction. Therefore, as every second passes, the future is changed. The cause of this are the laws of physics. If the future were predictable with all likelihood we would have different physical laws and life would probably not even exist.
But man is a subborn species. The unhealthy desire to predict the future has pushed mathematicians to devise utterly  unnatural methods which, in virtue of prolonged and often distorted use, are now deeply rooted in the practises in virtually all spheres of social life. Scientists speak of predictive models, just as the economists, the weather man, etc. Some people believe in horoscopes while others buy lucky  lottery numbers.
Much of the contemporary “predictive machinery” is based on statistics – looking back in time, building some model of what has actually happened, extrapolating into the future. The concept of probability plays a central role here. Bertrand Russel is known to have said, back in 1929, that “probability is the most important concept in modern science, especially as nobody has the slightest notion what it means”. In fact, probability is not a physical entity and it is not subjected to any laws in the strict scientific meaning. As a matter of fact, there are no laws of probability. If a future event will take place, it will do so irrespective of the probability that we may have attached to it. If an extremely  unlikely event will happen, it’s probability of occurrence is already 100%.
Predictions are of major interest in the realm of uncertainty. Clearly, one can predict with a high degree of accuracy when an object will hit the ground when it is dropped from a certain height (providing it is not a feather). What we are more concerned with is the desire to predict phenomena and events of interest to economists, investors, managers or politicians.
But there is another problem with predictions. Suppose you do indeed know with certainty that an event of interest to you will happen at a specific time in the future.  You will surely take action based on that knowledge. What this can cause, however, is a change in the chain of events such that you inevitably alter that event. As an example, suppose that you are extremely wealthy and that you know the exact value of certain stocks some months in advance. You will immediately start to buy and /or sell massive amounts of these stocks. This will surely cause other investors to react. Inevitably, the flow of events will be such that the predicted values of these stocks will not be the ones you knew with “certainty”. What does this mean? It means that you can only verify a prediction if you do nothing. The moment you act based on your knowledge of the future you automatically alter it and the prediction cannot be verified.  Consequently, the phrase “predictive model” is an oxymoron. As mentioned, because of the way the laws of physics work, the future is permanently under construction. And if you add Goedel to the picture …… The Creator is indeed very smart!
So, it seems that our efforts to devise some sort of predictive analytical machinery is futile. The current planetary meltdown of the economy eloquently underscores this fact. The severity and depth of this crisis has not been predicted (had this been the case we would have taken measures, right?) and this speaks of the quality of the contemporary economic and econometric models and of their predictive capability. With all respect, their predictive capability is not too exciting.
In actual fact, we still don’t even really understand the crisis and its multiple causes. But how can one speak of predicting phenomena which  are poorly understood? Shouldn’t we change the order of things? Shouldn’t we try to first understand better the dynamics of   highly complex interconnected  and turbulent economic systems and devote less resources to fortune telling and high-tech circle-squaring? How about:
  • Taking a holistic view of things, analysing systems of entities not single entities.
  • Searching for recognisable patterns, not repeatable details. The closer you look the less you see!
  • Moving from sophisticated (and subjective) models to model-free approaches.
  • Developing a new kind of maths, which is less “digital” and closer to reality.
  • Dedicating more effort to understanding the way things work, the way Nature works.
What cannot be achieved should not be pursued. Our efforts and resources should be focused on real problems that admit real solutions. Omnis ars naturae imitatio est.

Established originally in 2005 in the USA, Ontonix is a technology company headquartered in Como, Italy. The unusual technology and solutions developed by Ontonix focus on countering what most threatens safety, advanced products, critical infrastructures, or IT network security - the rapid growth of complexity. In 2007 the company received recognition by being selected as Gartner's Cool Vendor. What makes Ontonix different from all those companies and research centers who claim to manage complexity is that we have a complexity metric. This means that we MEASURE complexity. We detect anomalies in complex defense systems without using Machine Learning for one very good reason: our clients don’t have the luxury of multiple examples of failures necessary to teach software to recognize them. We identify anomalies without having seen them before. Sometimes, you must get it right the first and only time!

6 comments on “Is It Possible To Make Predictions?

  1. “If an extremely unlikely event will happen, it’s probability of occurrence is already 100%.” This concept removes risk resolution from the analysis. Risk management is not open ended, without scope or borders. You have to set those and examine the threat or opportunity within those parameters. Who cares it is 100% that a hurricane will occur somewhere in CONUS in the next 10 years with death and destruction. That does not tell anyone anything of importance. Now setting the resolution in both time and space becomes far more meaningful and the probability is no longer 100%. North Carolina, Outer Banks, 2nd week of August, 2014…not 100%. The weather models based on our history and the limitations of humans making them, which is not perfect, will establish a very reliable and reasonably valid probability and probability density function of impact from which a vacationer can make reasonable plans with appropriate contingencies.

    That models are made from historical data and are not perfect, who cares? What is perfect? Would you tell you your wife, sister, mom to stop doing her yearly mammograms because of the prevalence of types 1 and 2 errors caused by the machine itself and the same errors made by the human interpreting the picture? After all, the study is not only looking at what’s there now but also calculating the risk for this patient in her future, based on differences in the scan year to year but also family historical risk factors…and now apparently the presence of a gene in the DNA. Should we stop going to the doctor?

    And so what if our action post risk identification alters the future. That’s what mitigation and exploitation of threats and opportunities are supposed to do. Alter the future. The inherent issue with your statement above exposes a huge illusion of control. There are thousand of variables, mostly random, impacting us daily. We control but a tiny piece of that. Our ability to alter the future is mostly limited to an illusion, and with some a delusion.


    • David, nobody stops getting her mammogram because it helps catch problems in advance – this is precisely what I am suggesting: high-frequency monitoring. A mammogram catches cancer independently of the probability of actually getting cancer. ratings are issued for ALL listed companies on a yearly basis, not just for those that are potential candidates for default. What i am trying to say is that things happen independently of a measure of probability which, ultimately, means very little.

      I’d like to point you to this recent blog:


      • Mammograms will certainly detect cancer after it is present, not including false negatives, but scans are ordered for women at high risk at a higher frequency than those in a lower risk category. Radiologists not only read the current film but also compare films from the past to look for changes because that could change the risk profile.

        I am not understanding the philosophy behind, ‘things happen independently of a measure of probability….” That seems to suggest that, since we cannot predict the future with 100% accuracy and precision, then all of risk management is nonsense.


  2. The nature of things does not admit too much accuracy and precision. Therefore we should rather speak of relevance, not accuracy.

    Risk management is not nonsense but if you look at the state of the global economy you may want to question the significance and performance of traditional risk rating, assessment and management technology. Not too exciting.

    What is nonsense is to stratify the Probability of Default of companies into 30+ classes as the Credit rating Agencies do. What is the PRACTICAL difference between an AAA and an AA+? What difference does a 0.025% PoD make? What does it really mean? Does it mean that an AA+ company will default before an AAA one? Does it mean the consequences will be more severe? Certainly not. It is modern digital alchemy (astrology).

    Wrt “things happen independently of a measure of probability”. Since “probability” – a more accurate definition is “frequency of an event measured during an experiment” – has no physical meaning (i.e.e there are no laws of probability) events happen independently of the probability we attach to them. If, for example, “it will rain tomorrow”, it will rain independently of the fact that the weather service will say the chnace of rain is 25% or 50% – it WILL rain regardless.


  3. I don’t know much about financial risk ratings but I studied a little bit weather products and forecasting because I fly planes. The only way we can judge the relevance or accuracy / precision of a weather prediction is if the forecast is 100% of sunshine or 100% of a weather event. I cannot imagine drawing a conclusion on how well we forecast weather unless we capture what happened with a 75% forecast of a weather event over many observations, like maybe 100 or so days of 75% forecasts, where we would expect to see, out of 100 days, 75 it rained, 25 it did not, or thereabouts.

    The weather products produces a likelihood based on what is being observed, like the day before, with winds, temperature, and pressure patterns. Our models say 50% because in our history, 50% of the time those conditions did not materialize and 50% of the time it did.

    Still, I don’t get your view, but I’ll continue to read with interest to try to understand from where you are coming. Thanks.


  4. David, I appreciate the comments (and the open mind!). How about a Skype call? At Ontonix we are most interested in how people view risks.

    Not sure if you know the famous statement on probability made by Bertrand Russell: “Probability is the most important concept in 20-th century science, especially as nobody has any clue as to what it means”. I subscribe to this statement 100% as I have tremendous problems with understanding it. I’ve written a few books on the matter and I must say that the concept is extremely difficult to grasp.


Leave a Reply

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

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

Facebook photo

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

Connecting to %s

%d bloggers like this: