>So, if someone makes an incorrect prediction, then it is reasonable to assume that the market is unpredictable? I was looking at the Yahoo! chart 5 weeks ago, and my prediction was that it needed a confirmation day before one could say it was in a downtrend. No confirmation occurred, as a matter of fact, it reversed. Since I made the correct interpretation, does that mean the market is predictable?<
Be aware you are making the interpretation after the fact. All kinds of rationalizations can be made after the price pattern is available.
An interesting experiment has been the development of faked price charts using coin flips. Expert chartists cannot distinguish between these charts and real ones! Nice patterns similar to the ones you see in real charts do develop randomly!
Suppose you take the average stock which on average returns 12% annually with a standard deviation of around 20%. Let's say ignoring the geometric nature (compounding of returns) the stocks returns an average of 1%/month. Let's develop a monthly chart (you can develop a daily chart just by scaling the returns to daily returns)
1. Start with a price of 100 at the begining of the year.
2. Flip a coin and if you get:
i) Heads assume the monthly return is 2% (i.e. multiply the closing price for the previous month by 1.02)
ii) Tails assume the monthly return is -1% (i.e., multiply the closing price for the previous month by .99)
As you see the average monthly return would be 0.5*2% + 0.5*1%= 1%.
Try this out and then look at the use TA on the resulting patterns.
I hope you agree that regardless of your conclusions, for these faked charts, the future price pattern will be independent of the past pattern.
Now let's go to the real charts, in analyzing thet price patterns of real stock charts, researchers have not been capable of showing that these prices deviate significantly from those that may be generated by the ramdom process I just described. Small deviations have been found (the now famous January effect among them) but these deviations are not significant enough to make a killing,particularly if you consider transaction costs (remember the January effect is linked to small stocks for which the bid/ask spreads are higher than for large cap stocks the hit you get in the spread is a much larger portion of the transaction cost than the 9.95/trade or whatever discount commission you pay). Also once these anomalities become know investors destroy them by bidding up [down] the prices. I for example don't care at all about the so called January effect. people are now starting to take advantage of it in December!
Since this random walk business has been beaten to death, brigth assistant professors thirsty for publishing so that they can achieve tenure do spend a significant amount of brain power trying to find deviations from randomness, not confirmations of randomness. The outcome from turning "inside" out millions and millions of megabytes worth of price data as that they have not been capable of finding much.
Now in the long term we can indeed predict that stocks return an average of about 12% per year. However given the high standard deviation of return (20%), the actual return in any given year may be +30% (like we have see recently) or -10%, -15% or even lower. This is why people say money needed in the short term should not be placed in the stock market.
Consider however the case of the person retiring 25 years from now. The expected average annual return for the 25 year period is still 12%/year but the law of averages reduces the standard deviation of return (the risk) by a factor of 1/square root(number of years) so the standard deviation for a 25 year period is only 20%/5 = 4%, since the suquare root of 25 is 5.
This is why we say people investing for the long term face a lesser amount of risk!
To understand the magic reduction of variability introduced when we take an average. Consider a classroom for one of my typical undergraduate evening classes. We attract a significant number of non-traditional students so in an evening class you may find a range of ages that goes from the low 20's and into the 50's even 60's.
Suppose the average age for the class is 27 years of age. If I select one student randomly and ask her/his (lady's first) age, the resulting age will be highly variable; I may pick the 18 year old kid or even the 65 year old student. This is A highly risky experimen(in terms of the variability of out comes).
Suppose now that I take a random sample of let's say 20 students and take the average of their ages. I hope you agree the resulting average is much more likely to be closer 27 since my sample may include extremes such as the ones mentioned obove but these washout in the average.
Pancho
PS: excuse all the typos I am too lazy to fix them |
