This book, lent to me by my brother-in-law and company lawyer Jeremy, is written by a trader in options, a business about which I understand little. His basic thesis is that most movement on the financial markets is randomly generated statistical noise which dupes us into not expecting the highly improbable.
He points out that randomness can lead to interesting scams. Supposing in January I randomly email 50,000 people with my prediction that the market will go up in the next month and 50,000 with my prediction that it will go down. In February I select those people for whom I was right in January and email 25,000 with the prediction that it will go up in February and 25,000 that it will go down. I repeat this process in March (12,500 each time) and April (6,250) and May (3,125). By the end of May there are 1,562 people convinced that I can predict the market (5 months in a row!!!) of whom maybe half would be prepared to subscribe to my premium advisory service for £50. That's nearly £40,000.
Then he analysed a situation which has a 15% return but 10% volatility (that 10% is a standard deviation so that there is a chance that some returns would lose money) and looked at the chance of such an investment making money. He used a Monte Carlo random path generator. The chance of being up on one's investment after 1 second is 50.2%, after 1 day it is 54%, after 1 month it is 67% and after 1 year it is 93%. He then points out that people hate losing money more than the like gaining it. So if you hate losing twice as much as you like gaining, after 1 second you are probably feeling rather glum; after 1 day you have 54 points of happiness to 96 points of unhappiness, after 1 month the figures are 67 happiness, 66 unhappiness so you break even but after 1 year you are in woopy do territory with 93 happy points to only 14 gloom. His conclusion is that people who monitor there investments on a moment to moment basis are going to feel unhappy and get stressed and have early heart attacks!
Then he looks at things that skew our perceptions. One example is the natural tendency of people to ignore
outliers (extreme events); this can be highlighted by the difference between the mean and the median of a set of statistics (the median is an average which ignore extreme events). A second example is the tendency only to see things from history that still apply: this is like marvelling at the wonderful cathedrals from the middle ages and then generalising that "they really knew how to build in those days"; of course we can't see the cathedrals that fell down (or were replaced because they were too ugly). A third is our difficulty at calculating probabilities (he used the famous birthday paradox in which the chances of someone sharing your birthday when there are 23 people in a room is bout 50%). He points out that a lot of people spend a lot of time searching for correlations (he calls it "data mining"); clearly some correlations are caused by random effects so if you mine data long enough you will find such spurious correlations.
But he reserves most of his ammunition for conditional probabilities. If a test for a disease produces 5% false positives and the disease affects 1 in 1,000 and you randomly test a patient for the disease and the test produces a positive result what is the chance that the patient has the disease? Most doctors will answer 95% (100 - the number of false positives). Imagine you test 1,000 people. 51 will test positive but only 1 will actually have the disease. The correct answer is therefore under 2%.
NNT makes his money by betting on unlikely events. He expects to lose money most days because the events he is betting on are unlikely. But he knows that some days he will win a lot.
A very thought provoking book which I must flick through again some time to ensurer I have truly understood some difficult concepts!
Read January 2009; 230 pp
