The greater the incident deviation from the mean, the greater the regression to the mean. The reason is that paradoxically a low price encourages selling and a high price encourages confidence and buying
Price is not referential. Price does not care what the price was yesterday. Fear is the expression of the possibility that a strongly performing stock will open the next day down 25%, without any catalyst or earnings. That is the behavior of a crash.
When the investor feels that the possibility of a crash is very low, growing and high cash stocks rise without any reference to underlying value. P/E and revenue/price are measures that the market quickly ignores.
When the trader sees this, the behavior is idiosyncratic; the stock approaches a 52 week, 200 DMA or 50 DMA line, and it crosses it in force.
Price remembers itself, and it revisits those levels. In that way technical analysis is correct. The reason is that buyers, sellers and shorters are concentrated at those levels of sentiment.
But it is not referential. Stocks can "discover" completely dissonant prices.
Recently, in Q4 2012, MMR reported a crap well in Mexico and fell from 10 to 7.75, and was then imm bought for 18 with a special warrant.
MCP reports a takeover rumor and jumps from the 9-10 range to 11.5
a week later it falls to 8 b/c it mines nothing
These prices do not represent a connection to anything. I ask the question, let us look at the imputed value of CROX, a plastic shoe company over time;
2006 IPO at 6
2007 Rise to 70
2008 Fall to 1.5
2011 Rise to 33
2012 Fall to 13
Now, this is a dramatic example, but does one mean to tell me that the fortunes of this company, in fundamental terms; were talking about how many shoes that they sell, do you think that they varied more than the proportion of 1 to 5, much less a move
from 1 to 10, from 10 to .35 to 5, to 2...
No, but only a change in sentiment.