I took a market analysis statistics class a couple of years ago (very interesting class but hard as nails) and the conclusion we came to at the end of the semester was that statistical analysis doesn't work on the stock market because of unit root (I can't recall what this means) and non-stationary problems. The non-stationary problems are obvious -- the market doesn't stop for you to analyze it so picking your end point for the data is almost impossible (impossible may be too strong a word) and picking your beginning point is very difficult. I plotted stock market data on a distribution curve a while back and it looks like a skewed curve with smaller normal curves inside the skewed curve. The trick is to do your analysis within those minor normal curves. If you can manage to do that then you will be very successful. Also, as the day wears on the data becomes more skewed and the sizes of the normal curves become very unpredictable, which is a very important detail. Do the most of your trading during the morning and trade light in the PM or don't trade at all, assuming you're using a statistical based method. Today's action is a perfect example of what the curve looks like. I attached a rough sketch of what market data looks like statistically. I hear a lot of people say they trade well in the morning and give it all back in the afternoon -- this is why. If someone knows what Unit Root Problems mean, and can explain it in understandable terms I'd love to read about it. I looked it up on the internet but the terminology is beyond my statistical comprehension.
Are there more buyers than sellers in a given price range? How is the buyer trying to disguise his order? Is the seller not showing his full hand? What repeating patterns keep happening? What is driving the stock in one direction, or holding it back from breaking out? If a certain event happens, will people panic? Can you be quicker to react to that event than other people? Can you start that event yourself? What is different about the way the crowd is reacting to the order flow now from a previous period in the trend? Is the crowd more euphoric, or more fearful that the trend may be over? AM/PM/Statistical Analysis don't do that great of a job of answering those questions... Don't make the stock market more complicated than it is, or any other market for that matter.
Outliers usually occur in the afternoon. A lot of people trying to buy the bottom sell the top (going counter-trend) during these uni-directional moves get run over. This is one of the reasons ppl blow their morning loot.