Power law distributions

The last text was about Mediocristan and Extremistan. Let's forget the former and focus on the latter. Extremistan is a world in which the enormous outcomes are the results of small causes. 80/20 rule is the best known power-law distribution. It states that 80 percent of the outcomes come from 20 percent of the causes. This is just one power-law distribution and there are uncountable amounts of them. Do not focus on the 80/20 rule when you think about power-law distributions. Covid19-epidemic is an extreme example of a power-law distribution. Even though it is not sure, it likely started when one person ate some part of an animal he should not have. Millions of people will suffer from the virus.

The good, the bad and the irrelevant

It is not important to only understand that a minority of causes or inputs produce the majority of effects and outputs. Understanding that the majority of causes or inputs produce a minority of causes and outputs is equally important. Let's say that your causes and effects have 80/20 distribution. Then 80 percent of your causes and inputs are irrelevant. Their effects and outputs are close to zero. It is safe to say that 20 percent of your causes and inputs have large positive and negative effects and outputs. If your life is good, then your 20 percent has larger positive than negative effects. If your life is bad, the negative effects of that 20 percent are larger.

You do not have to be a rocket scientist to understand you have to enlarge the effects of positive causes and inputs in your life and diminish the negative ones. You can change to get rid of some of the irrelevant 80 percent of the causes and inputs. Change them into positives. It is unlikely you can get rid of all the negative causes and inputs. You can find ways to diminish some of the effects and outputs. You can transform some of them into irrelevant ones.

Long-term predictions are not smart in Extremistan

You can make predictions in Extremistan and be right for a short time, but the overall results of predictions are usually negative. As I mentioned in the last text, you need 100 billion fold data points in the 80/20 world compared to the normal distribution world. When the power-law distribution is 50/1, you cannot have enough data points or computing power to make any useful predictions. Financial markets and economies are parts of Extremistan. Therefore, an economist, financial pundit, or someone else who believes they can make exact predictions are idiots. Their total effect on the economy is negative. The funny thing is that they believe their predictive models are based on solid math. In short, do not believe anyone who believes they can make any useful long-term predictions in any social interactions. The world today is not predictable.

When predicting is not possible, you have to prepare for everything. How to do it is a whole another ball game. One way to deal with it is to reduce tight interactions. Do not use all your resources to achieve something like putting all of your time or other resources like money into one endeavor. Be less leveraged than you can. Have smaller debts than you can take. Be more independent. Rely less on other people and their resources. Have many income streams. Then you rely less on a single one.

Start making small experiments. Make small and numerous adjustments to your inputs. See what happens. There are lots of examples of what you can do. Make small changes in your marketing materials when you can do them in a cost-efficient way. Change a few words or colors, etc. Make small adjustments to your daily habits like your eating. See what happens if you diminish the availability of bad nutrition and increase the availability of a good one. Put some of your sweets (if you have them at home or work) into somewhere you cannot see them without increased effort in finding them.

Do you consider yourself a helpful person? If you found anything useful in this text, teach it to your friends.

Until next time,

-TT

How to predict better?

Everybody makes predictions. Most predictions are irrelevant in the big picture, but everyone makes them. For example, you can predict tomorrow´s weather. Most of the time, it doesn´t matter whether you are right or wrong. But sometimes you have to be outside for the whole tomorrow and then it matters. You make these predictions without a systematic process. You need to have a systematic process to make better predictions. This text is about making them.

Commit to the truth

The first step for making better predictions is to commit to the truth. You have to recognize the influence of your beliefs, and psychological tendencies and you have to question them. You cannot make predictions that confirm your way of seeing the truth when the evidence does not confirm your beliefs or assumptions. You have to be willing to update your beliefs and assumptions when the objective facts tell you different stories than you would like to see. Your beliefs and assumptions have the power to modify your predictions in wrong directions unless you commit to find the truth no matter what that is. Your goal is to make the best prediction you can make with the facts you can gather. If you think you do not have enough facts, you have to commit to finding them. If you have too many facts, you have to commit to separating the relevant facts from noise.

The process of making good predictions

Most people have three different answers to predictions: ”Yes”, ”No”, and ”Maybe.” People who make good predictions live in an uncertain world. In this world, ”Maybe” is the only right answer. This means that you have to make predictions with probability estimates. Your prediction could look like this: ”There is a 70% chance of raining tomorrow.” You also have to second-guess all the people who make vague predictions like ”It may rain tomorrow” or ”The unemployment rate may be less than 5% in the next two years.” You cannot know whether these people have made the right predictions because their predictions can mean anything. If you want to make better predictions, you have to commit to making probability estimations.

The starting point for making a probability estimation is to find a base rate for the thing you are predicting. For example, if you have to predict what is the probability that your football team will win the next home game, you have to start by looking at how many wins they have won in their home games this season. Your sample size should be enough. If the season has just started, you have to look for last season´s statistics, too. Let´s say they won 7 out of their last 10 home games, then the base rate is 70%. This is your starting point. Then you have to use other data to adjust your probability estimate. Why should you start with a base rate? Because your first estimation is just your hunch and the probability of it being right is smaller than using the base rate. The figure you find first will be the most available for your brains. It will be your anchor during your prediction process. Do not forget to define the time frame for your prediction if necessary.

After you have figured out the base rate, you can start forming your view. Break your question to smaller questions like ”What would have to be true for this not to happen?”, ”What kind of information helps me to answer this question?” or ”What I do not know about this question?” You have to figure out what you don´t know and what you do know. After many questions, you should make your probability estimate and write it down and the reasons behind it.

Then, it is time to find out some outsider view about the same question. Consult people who have made predictions about the same question. Find out what the experts think and their reasons. Focus on the differences in reasons between your view and other people. Consult prediction markets, like stock markets, or betting offices that give odds to a thing you predict. You can use polls if you predict the elections, etc. After you have found out the base rate, made your probability estimate, and found out about the outside views, you have to synthesize and make another probability estimate.

Let the time pass and then you can scrutinize your own prediction and make another estimation. You can do it by assuming your first estimate is wrong. Consider why it is wrong and write down the second estimate and reasons for it. And then find out about other outside views again. Then, synthesize again and update your prediction when new important information arrives. This process ends when the time frame closes. This process can feel frustrating. If you want to predict better, you have to update your estimates often and make gradual adjustments to your predictions.

Prepare to be wrong. Even the best predicting professionals are occasionally wrong. Do not mix up the outcome of the prediction and the quality of the process. When you have 80/20 predictions, you are rt only 4 out of 5 predictions when your probability estimates have good accuracy. No matter whether you are right or wrong, have a postmortem. Think about the process and the reasons behind your estimate. Did your process have high quality? Did you find out the right base rate? Did you update your estimate often enough? Did you find out the outside views of the best professionals and prediction markets? How many times you synthesized these views? All these questions help you to predict better next time.

Sources:

Superforecasting, Philip Tetlock

The Signal and the Noise, Nate Silver

-TT