I must admit that the original post was based on an idea that is IMPOSSSIBLE to implement. It is possible to extend a random time series to become a nonrandom. It is impossible to extend a nonrandom time series to become a random time series. I thought it's possible. Sorry..
"Random walk doesn't exist. The fact that there is a correlation between stocks within the same industry group or the broad market as a whole on decisive up/down days puts random walk to rest. If it was truly random walk, would you ever see two graphs in the same industry that look the same? Think about it" OK I thought about it. For about 2 seconds. Your reasoning is flawed. Individual share prices are OBVIOUSLY correlated to other share prices. i.e. So what if all the refiners acted the same way today? If you can glean what one of those refiners share price is going to do tomorrow based on the fact that as a group, they all went up today, then you really have a gift. Todays price movement will have no bearing on tomorrow's price movement.
If I understand you this is possible. I'm probably too tired to explain this clearly but I'll try. The main feature of a random walk is that your expectation for a future price is the current price. You can have mean reversion or autocorrelation at a short frequency but a random walk at a longer frequency. If a price process is mean reverting at the daily frequency, then your expected value for tomorrow is somewhere between today's price and yesterdays. If it is a random walk at the 2 day frequency then your expected value for 2 days from now is the same as today. I don't think anything prevents a price series from having both properties- tomorrow's price expected price is lower/higher than today's price, but the expected price 2 days from now is equal to todays price. I am not saying prices really work this way but I think it is possible to generate a time series that does. If this doesn't make sense I'll think harder about it later.
Very interesting thread. It has certainly attracted some of the more sophisticated and intelligent traders here at ET I personally don't believe the markets are random. It is apparent that they display certain types of price movement, price patterns and price range over and over again in an extremely consistent fashion when measured over periods of time where one period is equal to one trading session of whatever indice you are trading. IOW, for any 20 periods of time (trading sessions) I would bet that any one of you guys could figure out what the market is doing approximately 70 - 80% of the time. The problem lies in the human mind ... the randomness which we think we percieve in the markets are actually a part of our own psyche and lack of discipline. The randomness is in the human mind.
Well, it has become clear to me that the markets are VERY predictably random. Why? My strategy, much like Maestro's, makes 99% profitable trades and never a losing month. How could that be possible? The way this model works is knowing for a certain that whatever direction the market is moving, it will not last forever, period. In other words, just when most humans in the market see a "non-random" looking pattern on the chart and flock to buy into the "trend", it will soon after correct enough that many buyers get stopped out. If you ever traded with real money, then you have experienced that. Well, think. Someone else took the opposite position from you and profited. How soon the market corrects/rallies and how much it corrects/rallies is unpredictable. BUT you can be CERTAIN it will correct or rally. As far are correlation, of course markets correlate across stocks, etc. Absolutely, the events in the news affect companies/currencies similarly. However, let's try a little demonstration to prove that correlation never changes their randomness or unpredictability. Think about it this way... If you truly believe those correlated stocks are non-random, then simply predict accurately whether each of those stocks will be up or down the next hour, the day, and the next week. Look at the charts and write a log for a couple weeks and see how close you come to predicting any of them correctly. If you try it on a number of stocks or currencies for period of time, rest assured you'll find you are wrong most of the time. That's why all of us fail with most of our back tested strategies when we try to trade them on truly unpredictable markets. They rarely ever work because they attempt to predict the future. That's easy when back testing. Not so with the very real and elusive price "tommorow". So the markets all LOOK predictable with some non-random nature but that, my friends, is ONLY in hind sight which is, of course, 20/20. If you think about it, you can use that simple fact to make a phenomenal amount of money in the markets. That means doubling your account regularly. But no one teaches those strategies for free or for profit. They work too well. I, for one, will never divulge the details of mine to anyone, end of story. Wayne