Price curves have a very low signal to noise ratio. Therefore one would normally assume that signals or patterns based on the high or low of a bar have no predictive value, except maybe for HFT, or for 24-hour bars of the Japanese rice market. Is price action just trading the noise or is there more to it?
stare at a 15 min ES chart every day for a few years. Soon you'll start to see patterns that repeat themselves. For example, Thursday 3/8/12 's chart pattern I hadn't seen in about 3 months or so. But I recognized it fairly quickly and knew the probability was in my favor to initiate a long position. Thats why its so important to have years of historical data. The way price action repeats itself is not obvious by any means. But after time, you'll be staring at a, say, 15 min ES chart, and you'll think, "holy sh#t" ive seen this before and its most likely going to go to point X. Now you understand PA. Now you can get a trading plan together !
I used to have the same doubts about PA until I studied thousands and thousands of charts, and observed live price action as it happens. There are definitely undeniable pivot points or SR lines where price hesitates then makes a decisive move and I am certain this is not voodoo or an illusion or hindsight bias. In a nutshell this happens because people cannot truly define absolute value, but rather can only decide on the relative value of things. The probabilities are not absolute but are conditional, and they are well-defined at those points. The difficulty is in properly defining the bins to count the statistics on -- this is not easy. The stock market is not all physics -- it is based also on the psychology of the masses. Some examples of this in everyday microeconomics are illustrated in the book "Predictably Irrational" by Dan Ariely.
The key here is probability . PA trading makes sense if you can get a prior estimate of probability. Then it turns into a very powerful method but not easy to execute because it involves quant analysis. One of the PA gurus is undoubtedly Michael Harris. You will find many examples in his blog of PA analysis and very interesting and counter-intuitive results. For example before the open last Tuesday he posted a blog about a pattern in SPY, an inside day breakout and analysis that showed that historically the underline PA had a bullish bias based on two samples of 60 and 98 trades. Here is the link. It is interesting that the market gapped down on Tuesday at the open and since then SPY had a 3-day winning streak for a gain of $2.20 per share. This is a style of PA analysis that I like. There are other styles based on s&r, pivots, etc. but in those cases it is hard to estimate prior probability. Of course in the case of Harris his program helps a lot but it's too expensive for most.
You bring up a good point: there are perhaps 3 main elements to PA. One is to define your probability bins (events, cases) by bar-patterns. This is easy to do and easy to get stats on. Unfortunately most of these are noisy, and barely give you enough positive expectation above commission+slippage. The second is based on the "shapes" of the price trajectory, i.e., lower highs higher lows, trend lines, bull flags, bear flags, pennants, etc. This is the bread and butter for a successful discretionary trader. The problem with this one is that it is difficult to write a program to detect and bin those general "shape occurrences" to count and formulate your conditional probabilities. The third is based on ranges: where you are in today's range, yesterday's range, wrt pivot lines, weekly range, quarterly range. Again this is fairly straightforward to calculate probabilities, but most of these again are not slam dunk cases, and often what is significant in the last 3 years will not be significant for the next 3 years, etc. I think by far, the second item above is the most reliable, but the most difficult to code.
It's interesting that you mention Michael Harris, because I found his software on the web and it's one of the reasons I wondered about price action. At a first glance his website looks like a scam - its purpose seems promoting his relatively simple software at a ridiculous price. But on the other hand, when price action indeed works, then what he's doing would make sense.
The most successful trader I know is a PA trader. I could only dream of ever obtaining 10% his prowess.. ES
Of course it makes sense but the kind that does, the true action that matters, I personally have not seen it posted in public. Some get close but they either choose not to post what follows or they just don't see it and never get there. Once you find about it combined with some statistical probabilities you can do some good damage provided you have capital as over-leverage will kill the very best of traders, including PA traders trading good stuff. Just remember that anything can happen at any given time, and when I say anything, I mean anything whatsoever, without limitations.