Nothing around you is certain. People walk swiftly to their cars with keys in hand and take an entire two minutes to open the doors. They probably thought they would get it right in ten seconds. Some wake up and walk to the bus stop to get a bus to work. There is a certain expectation that they will get a bus on time and get to work on time. Truth is, sometimes the buses come full or the traffic starts early. We don’t know for sure whether our assumptions about the future are precisely correct, we just apply.
The same scheme of behavior works for greater systems such as the stock exchange. Some companies come on strong on their first day in the market with a legion of investors reaching out for a piece of the juicy IPO share pie. No one is quite sure that the prices will ascend but they still apply for it. And sure enough there have been enough cases where these bullish traders record losses after a year of holding on to their stocks.
In the discipline of probability we say that everything that happens has a probability of 0 to1 of taking place. The more the number of subset cases of the event (which qualify that that event has taken place) the more probable it is the event can happen. The more time that thing is given to take place, the more probable it is it will come to pass.
As humans we go through life studying the conditions that lead to an event in order to maximize the probability of success. We continuously seek to make the best predictions of what might happen if we take certain actions. Nate Silver, the famed pop-statistics journalist, wrote a book about it called the Signal and the Noise. He deduced that there are certain priors in human decision making that guide the way we model the future. Some of these priors are relevant while other are just noise. For instance if we support a preferred side in a football match (which Silver has already become a pro at deducing), what factors led us to believe that this side will win the game? The priors are numerous – and in certain cases innumerable. Even the best statisticians have admitted that no magical mathematical formula exists to substitute straight up experience in any field of prediction. This would be the case when an experience Bayesian statistician tries to beat an experienced poker player at a round table game. While the statistician looks for numerical correlations, the quality of factors he is enumerating may take a back seat to those that the poker officionado is assessing. This essentially is the difference between noise and signals.
But whether you are the intuitive poker champion of the probabilistic stats player, your chances at 100 percent success, 100 percent of the time are impossible. This is because within the conjecture that everything is an app, there is also a set of differential beneath each one thing. For instance, if you want to predict political election results correctly, you are expected to predict all levels of the elections, gubernatorial, senatorial, presidential and so forth. In probability this means multiplying the probability of your correctness at each level to the next. A 99 percent chance for a certain senator times a 99 percent chance for a president leads to a lower chance of both candidates emerging victorious (0.99 x 0.99). And even when you level down to only predicting the senator, you have a number of factors you used to arrive at your decision. For instance what are the chances of he has been in government, has he campaigned, if so how strongly.. and so on. Multiply them together and you will have an even smaller chance.
Whilst improving deduction techniques may be feasible such as was seen in Nates case, perfecting it is obviously impossible due to the complexity of detail. In this case, what becomes relvant is the value of applying your prediction. Markowitz introduced a sound way of looking at risk and investment, when he stated that higher returns come with higher risks. How acceptable is a prediction model in the face of certain costs? Does a 25 percent stake of inaccuracy warrant abandonment of an entire venture? Jumping ship is actually a matter of mathematical analysis just as much as predicting the storm is. If the cost of losing a bet on the future is manageable, the need to invest in greater accuracy should not be overemphasized.
- Nate Silver’s Odds For Big 5 Teams In The NCAA Tournament (philadelphia.cbslocal.com)
- Here’s all the interesting stuff in Nate Silver’s The Signal and the Noise (yourbrainonecon.wordpress.com)
- Bayesian reasoning, agile steering, and the economic governance of software delivery (ibm.com)
- On Topic: Our best guesstimates as told by former statistician Nate Silver (thegazette.com)