That is the classical example. There are deeper ones, for example: http://www.upscale.utoronto.ca/GeneralInterest/Harrison/BellsTheorem/BellsTheorem.html nitro
Cool, a challenge! When you assert randomness you assert a negative proposition and thus run into the same problems that stock market random walkers face. If no one understands quantum mechanics, as Feynman suggests, then how can we unequivocally state that quantum mechanics confirms the existence of randomness? What if there is a logical pattern there but our brains are simply too small to grasp it and we are 'like dogs trying to understand calculus'? The problem isn't solved by discounting logic either, because in that case statements become neither true nor false and the entire question becomes moot. Tell me why I'm wrong here (I'm rooting for myself to be wrong, how bizarro is that).
Of course, it is fair to say that quantum mechanics offers the theoretical viability of randomness. But that remains an unproven possibility and not an absolute assertion does it not? We could assert with equal weightiness that QM shows reality to be made entirely of thought (and have no less tough a time proving the assertion).
sigh darkhorse p.s. show me that all possible hidden variable theories have been properly tested and discarded and I will go on my merry way happy with your one syllable answer.
Still think about that ant farm + food pattern. ________________________________ Actually thought I discovered a random walker pattern about 7 years ago. However after 1 year study,discovered Elizabeth just liked to shop.She wasn't on a random walk;just lliked to walk and shop a lot,[discount]
You pose an excellent question! This question illustrates a major issue in trading and the power of screening to mislead traders (that you pose this question suggests that you are aware of what I am talking about). Even if these four stocks have seemingly non-randomly high numbers of downdays it is no guarantee that they are destined for further downdays. <b>When Coins Seem Nonrandom: The Danger of Survivorship Bias</b> The analogy for screening the N100 for down days is to collect 100 different coins, flip each one a 100 times and record how many heads each coin generates. Out of the 100 coins you WILL find that some coins seem to generate far too many heads, while some other coins generate far too many tails. Moreover, if you do a standard test for the statistical significance of the results, you will determine that some coins seem to have a statistically significant non-random results (even if the coins truly are random). The problem is that the usual 2-sigma or 5% significance threshold implies that 1 out of 20 random data sets will pass the test. Thus, testing the data from 100 stocks, one should find about 5 stocks that "look nonrandom" (Note that as with all things statistical, one will not always see exactly 5 significantly nonrandom coins or stocks out of 100). Flip enough coins, enough times, and you will find strange results. So, what is one to make of the coin that produced the most heads on the first 100 flips (out of the original set of 100 coins)?? Probably nothing because the coin is example of survivorship bias -- by random chance some coins (or stocks) will survive a test for nonrandomness. Despite the heady performance of that coin in the first set of flips, its probability of heads will remain 50%. The point is, if you look at enough stocks you are guaranteed to find some that seem to have non-random patterns or fit a trading system extremely well. As one increases the number of stocks in a screening process (or the number of different trading systems in a backtest), one has to implement ever-stricter statistical tests to filter out the high chance that some of the set of stocks (or coins) has, by totally random chance, produced a pattern that looks non-random. <b>Does the First 50-Days Data Predict the Second 50-days Data?</b> A better way to look at the data is to consider whether the number of up-days seen in the first half of the data predicts (or correlates) with the number of up-days in the second half of the data. A positive correlation suggests trending or momentum (so you should short the stocks with excessive numbers of down days). A negative correlation suggests mean reversion (so you should buy stocks with excessive numbers of down days). There are a number of ways to statistically check the relationship between outcomes in the first 50 days and those of the second 50 days: regular correlation, Spearman rank correlation, contingency tables (from 2x2 on upward). The only caution, in doing these tests, is that one should make the data disjoint by skipping a day of data between halfsets to correct for the presence of errors in the closing price data. One might also want to look at an number of successive 50-day periods -- if the pattern of correlation does not hold across a number of successive disjoint dataset, then there is no exploitable pattern. Modern computers, massive data sets, and powerful software give traders enough rope to shoot themselves in feet with. The harder you look for a pattern, the more often you find patterns that do not actually exist. Trade carefully, -Traden4Alpha
true dat and no wonder, given that computers have less common sense than bugs....long live discretionary validation
Traden4Alpha, i agree, what jperl asked is a good question. my question to you is, yes or no, in regards to what jperl asked? hehe
Traden4Alpha is right, this has been an overlooked post, imo. is the answer to this question yes or no? if no, that the data should not influence you, why look for trends then? i think most people here try to identify a trend in some way, with moving averages or whatever. could you not say the 2 are the same? if that data of % up days does not matter, why would price related to a moving average, or the slope of a moving average matter either? or, is it true, that the above data SHOULD influence your decision?