In this post I explore several things related to the meaning of prediction. I explore patterns in relation to trends.

It is important to define what people take prediction to mean and what I mean by prediction. I’m so totally uninterested in what Google or some guru out there thinks ‘prediction‘ is! Why – cuz these are my thoughts having done the listening to ‘other people’ over many years.

Broad definitions

In general – because I haven’t done a statistical survey – prediction to most people means one or all of the following (the list is not exhaustive):

1. It’s something to do with the so-called future – that nebulous place and time which doesn’t actually exist.
2. ‘When’ something (usually an event) is happening in ‘the future’.
3. What is going to happen.
4. A declaration that ‘X’ precedes ‘Y’ and  predicts ‘Y’, and ‘Y’ happens at some time in the future, then the declaration that ‘X predicted Y’, is said to be a completed prediction.
5. If ‘Y’ does not happen then people take the ‘prediction’ to be wrong.

I will come to what I think about prediction in the course of this exploration. In order to know what event is being predicted, it should normally be declared and shared with at least one other person. A prediction cannot be truly appreciated if it is not known by at least one person other than the individual making the prediction. There is little point in saying to somebody after the event has happened, “I knew this was going to happen!” – tough, that’s not a prediction! That’s called “Gorilla chest-beating“.

Simple things

The other important issue with predictions is that they must have value, relevance and meaning. Very few people would be interested in a prediction that states, “The sun will rise tomorrow morning.” However, people may become interested in a prediction that, “The sun will not rise tomorrow morning.” If the latter statement was declared and shared with others then it is reasonable to expect that people hearing that sort of prediction, would wish to know how it was arrived at. They might quiz the person making the statement to provide some sound reasoning and evidence in order to add credibility to the statement.

In the example above people naturally come to know from experience that the sun does rise on mornings. They know this because they will have seen this pattern over many years. (Tough – I can’t deal with minor exceptions such as newborn babies!). Therefore a statement declaring that the sun will not rise tomorrow goes against that well-established pattern. It is therefore reasonable to make deeper enquiries if a prediction suggests something out of the ordinary.

Immediately the issue of ‘patterns’ becomes very important and people often say that patterns are predictive. A pattern is not a living entity. It cannot ‘know’ anything or predict anything. It is people who use patterns to make their so-called predictions. A pattern is simply a sequence of events that happens with a certain rhythm and regularity which then gives it some reliability. It is easy to think about many patterns in everyday life. The ebb and flow of the seasons are well known. Simple events such as sunrises and sunsets fall into this simple set of patterns. Increased traffic between 4 p.m. and 630 p.m. around most UK cities is a well-known pattern. People often say that those sort of things are predictable. In reality they are referring to the regularity of a pattern, which creates a higher probability for its re-occurrence and its reliability.

Complex and random things

But life is not all so easy and people often want to know far more than simple things such as sunrises and the seasons. For example, they might wish to know when there might be exceptionally heavy snowfall. Or if there is a storm spotted by the weather people then people want to know roughly when (or whether) the storm might hit their area (or not). Now, storms though happening with greater frequency in certain seasons, are random events. They have the potential to destroy property and life. Therefore, people expect predictions about the path of storms and when they might hit, to come with some degree of accuracy. It is common knowledge that many such forecasts become inaccurate when they focus on a small geographic area. The prediction about the path of a storm may be more accurate for a larger geographic area simply because storms are pretty large in general. A simple example is commonly seen in everyday weather reports where the weatherman might predict heavy showers in large geographic areas of the country but then individuals in a small town are surprised when they are treated to much sunshine and no rain at all. The weatherman obviously cannot predict which small areas may escape having heavy showers poured on them.

I want to focus more on things like storms, tsunamis and earthquakes. If we accept that these events are random in nature and move on, then clearly their randomness cannot actually be predicted with great specificity for date, time and location. The discovery of a storm obviously puts it on a map in some part of the world. It is only then that the people who study these things along with other complex data, can make certain forecasts about the path and strength. But so far to date, no meteorologist has been able to predict the precise emergence of a storm at any point on a map of the world.

At this point prediction for events that have a clear seasonal or cyclical pattern may seem more accurate. Accurate predictions about events that are random in nature are far more difficult. Strangely though, our real interest is in things that are random, which may threaten our safety and security. I mentioned the contrast between predicting events that have an established pattern against those that have no clear pattern, in order to explore the meaning of the word prediction. ‘Meaning’ in itself is a word that suggests value.

To clarify this a bit further, whilst the 4 seasons are not random, certainly the occurrence of storms, tornadoes and tsunamis are random events in those seasonal patterns. The people with specialist skills who study the latter sort of events have a difficult task in analysing and making sense of complex data – even if they use ‘computers’. Their knowledge and skill may give a probabilistic estimate of ‘when’ such events may occur. For example they may inform persons who prepare for disasters, that there is a high probability within a certain timeframe of X number of tornadoes happening. But for sure they are unable to predict exactly where and when a single tornado might materialise.

Trends

I see trends as aggregates of patterns. Trends are therefore bigger and broader than patterns. For example, there is a trend towards global warming – but in that trend is contained several seasonal pattern changes that inform us where the  world’s climate is heading around this time (meaning the next 100 years or so). Trends are therefore even less predictive than patterns. Why? Because it would take a significant number of pattern changes to reverse or change the direction of the trend (containing such patterns). Trends are even less specific than patterns. They simply point a ‘direction’.

Turning now to the issue of stock-market crashes and recessions, a lot of people want to know when these are likely to start how long they are going to continue and how bad they are likely to be. The problem for those who try to predict those things is very similar to the scientists who are expected by the general public to give specific information about severe weather disruptions. Those who try to forecast random events can only give a rough estimate based on specific data when some random event is more probable. Stock markets change in many ways. There are patterns of changes and then trend changes.  Estimating or forecasting what will happen next in a stock market is very difficult. If one focuses just on patterns, those may be short lived and of limited value. If however, one finds a trend containing a number of patterns indicating a particular direction that could be more meaningful – but less accurate in terms of ‘prediction’. So, like global warming, we know it is happening and it is warming up – that’s the big trend (comprising seasonal pattern changes) – but nobody can accurately predict that (for example), in January 2120 what the peak temperature in Sao Paulo is gonna be.

Conclusions

1. Patterns have predictive value in general – though nobody can predict the future with any real degree of specificity.
2. Trends are broad containers of patterns – having even less specificity, though having better overall indication of future direction.
3. Prediction is a nonsensical term simply because it is too simplistic a concept. It is ‘event-focused’ for most people i.e. either the event happens or it doesn’t – in a rather binary way.
4. Prediction, for me, means the application of specialist knowledge, skill and experience in coming  a probabilistic estimate of what some event or outcome may be – within well defined parameters.
5. Predictive value means how valuable and reliable is a probability estimate.
6. Every probability estimate in one ‘direction’ cannot exclude a probability in the opposite direction.
7. The so-called prediction of random events even if nested in cyclical patterns is likely to be perceived as wrong. This is simply due to the fact that nobody can predict a truly random event with any specificity.

And finally, this post does not mean that I am trying to predict when there is going to be a stock market crash. I have to say this simply because ordinary people reading this are likely to extract their own meanings and turn inferences into facts. I could well be trying to predict the next storm in the mind of somebody (anybody) out there – and I’m not! This is how ridiculous blogs, forums and any sort of social media have become! And – my saying any of this is not with the intent of denying that which I might be doing. Chrysst!