One can view the market as a complex man-made system with evolving dynamics which are intertwined with each other. Unlike natural system however, attempt to model it is can be more difficult than existing systems in nature in the way that human is involved as part of the dynamics, and thus a seemingly successful model may soon be exploited and be factored in as another input dynamics, thereby altering the whole system.
There are unlimited ways to cut through the mustard, but in general we can categorize the majority ones into two: technical analysis and fundamental analysis. I would consider market emotion as part of TA. Some count it as the third one. Anyway here is what I think of these two type:
For a graph technician who works in the field of technical analysis, he does not need to know what the company, which the stock represents, does. In fact, it is better not to know, because the extra information will skew his judgment. In performing TA, the technician would mainly consider historic trade transaction data consisting of price movements and volume. And that's it. There are a lot of derived data created from these pair, the most famous being the candle sticks, moving averages and stochastic oscillators, depending on your taste, but fundamentally, these bells and whistles are all coming from these two simple entries: price and volume, on different horizon or time frame.
Now arming with these tools the technician can plot support and resistant levels, determining trend and then make his projection -- at this point this is really more art than concrete mechanics.
Comparing to TA, an analyst performing fundamental analysis has a lot more data to consider. Let me throw in the important ones: company past earnings, debt/liabilities, cash flow, market outlook, credit and the list goes on and on. To put a price on it, there are a few generally accepted models, like the discounted cash flow and the discounted dividend model, with the premise that assuming the company will exist forever, what is the present value of the company considering all the future revenue? There are many weak points in this assumption, but let me point out the one that academia likes to avoid: in calculating the discounted future value, how do you know what the discounted rate (most use the inflation rate) is? Usually they would resort to historic average, but is that it? Can't I just take the historic average of P/E ratio of S&P and assume the board market will drift back to this mean, assuming that these are cointegrated? Wouldn't life be much easier then, ha. But even assuming the rate will remain relatively constant, and that the company would not deviate from its past performance too much, this projected price is only the backbone of price estimation. Overloaded with all the information from earning reports, the analyst must decide and pick the important ones to calculate how much the company is worth based on his projection from the mountains of data. Popular metric includes Return On Equity (ROE), Economic Value Added (EVA) and Price per Earning ratio (P/E), comparing to the company's peers in the industry. These are all levers and knobs to adjust the projection. Unlike TA, these are actually based on sound (at least on the surface) financial theories to calculate how much a company is worth.
In the end as you may have figured, it is all approximation and guess work. I would not say this is logical guess work because you cannot prove any of the above really work. Let's not inflate and rape the rigor of logic.
