what do u like/dislike about it? i quite like the candlesticks and market profile, but dislike some superstitious price patterns.

From what I do know about the subject (after reading a few books on it), the candelstick method seems more hocus pocus than chart patterns, IMO.

A lot of it is pure voodoo. Some of it (momentum) seems to work, but you don't need charts to measure momentum, and using charts to find it is likely to screw you up, because you'll see momentum that may not be there in a statistical sense.The human mind is a pattern recognition system, and it likes to recognize patterns that don't exist.

worst part of TA is the use of charts...

QuoteOriginally posted by: bskilton81A lot of it is pure voodoo. Some of it (momentum) seems to work, but you don't need charts to measure momentum, and using charts to find it is likely to screw you up, because you'll see momentum that may not be there in a statistical sense.The human mind is a pattern recognition system, and it likes to recognize patterns that don't exist.i have read and heard lots of things like this, many books convinced me that TA doesn't work, but you know what, some basic concepts, like breakouts, are working quite often

Some of it does work, particularly in FX markets, but people should mind their t-stats. Backtest a strategy and and convince yourself that the results are statistically significant. Most people using TA make statesments like, "This has happend 10 times before, and 7 out of 10 times we had a breakout to the upside, so this is a good place to buy." That's all well and good, but a series of ten coin flips will give 7 or more heads with over a 17% probability, so basically it isn't really meaningful.

Some of it does work, particularly in FX markets, but people should mind their t-stats. Backtest a strategy and and convince yourself that the results are statistically significant. Most people using TA make statesments like, "This has happend 10 times before, and 7 out of 10 times we had a breakout to the upside, so this is a good place to buy." That's all well and good, but a series of ten coin flips will give 7 or more heads with over a 17% probability, so basically it isn't really meaningful. One of my problems with TA is the way it is used in practice (no tests for significance, very small samples, etc). Also, people like to backtest TA strategies using information they couldn't have known at the time of decision, because the information didn't actually come out until later (like using regressions for instance, when one wouldn't have known the parameters until long after the decision was made). This is cheating.

QuoteOriginally posted by: mitsome basic concepts, like breakouts, are working quite oftenThe only reason breakouts wouldn't work, is if suicidal idiots traded them. The condition that everybody knows about them isn't sufficient to stop them from working, only to get them down to the market rate of return.Here's an essay.There is something nice about breakouts, that is not mentioned in the essay. And that is how relatively easy it is for a group of people using them as a trading signal, to avoid triggering one another in single cascade. If you put small gaps between each groups of traders, from the small and fast to the large and slow, the price jumping that gap will signal something other than just the entry of a faster wave of breakout traders.Or, suppose the entire global population of breakout traders triggers itself at once and avalanches. You look at how big a move that was, and set your system to trigger at X times 1.5, or whatever.

I've tried about every trading method I can think of, and the only one that I have continued success with is breakouts. I would love to have a good overbought/oversold strategy in my portfolio, but I am not even sure if one that I trust even exists.

This is why bayesian inference, boot-strapping, surrogate data, etc are really useful for checking the validity of any TA based indicators..QuoteOriginally posted by: bskilton81Some of it does work, particularly in FX markets, but people should mind their t-stats. Backtest a strategy and and convince One of my problems with TA is the way it is used in practice (no tests for significance, very small samples, etc). Also, people like to backtest TA strategies using information they couldn't have known at the time of decision, because the information didn't actually come out until later (like using regressions for instance, when one wouldn't have known the parameters until long after the decision was made). This is cheating.

QuoteOriginally posted by: crowlogicThis is why bayesian inference, boot-strapping, surrogate data, etc are really useful for checking the validity of any TA based indicators..QuoteOriginally posted by: bskilton81Some of it does work, particularly in FX markets, but people should mind their t-stats. Backtest a strategy and and convince One of my problems with TA is the way it is used in practice (no tests for significance, very small samples, etc). Also, people like to backtest TA strategies using information they couldn't have known at the time of decision, because the information didn't actually come out until later (like using regressions for instance, when one wouldn't have known the parameters until long after the decision was made). This is cheating.I disagree with most of this. The problem with chart analysis, is that you have no idea what you are looking at, what phenomenon is causing the price to move. The only thing that all phenomena have in common is that they happen gradually, they trend. Plus, phenomena share common factors, so some inevitably are correlated at various lags.So if you are looking for something not involving trending or lagged correlation, better and better statistical tricks will just get you more and more unlikely coincidences. And if you are looking for trending and correlation, they are certainly there and cannot hide from your searches, the only reason not is because someone else is already trading them.Maybe you are suggesting that by looking at the distribution of outcomes associated with a given signal, you can tell more than by just looking at the sum of those outcomes. But if you're looking for something other than trending, lagged correlation, and the sound of people trading both, then all this is for nothing. Whereas if something isn't trending or correlated, all you've shown is that someone is already trading it.I don't care what some statistical study tells me, I know that an uptick in bonds is likely to be followed by another uptick. If it doesn't tick up, that's because the next uptick was masked by another trader like me, already selling into it.Basically, there's a whole set of guaranteed money makers out there, and you needn't bother yourself with whether they exist. Microsoft is correlated to Oracle with a lag, don't let any test tell you otherwise! All you need bother yourself with is, of these moneymakers which you know exist without any need for a test, which ones are not already being traded by someone else better than you?Don't kid yourself that you're measuring the nature of the world, of currency, or of interest rates, or of corporate prospects. What you are measuring, when you look for trends, is whether other traders have traded them away.

I agree with most of this because I use the term TA loosely. Basically, all forms of TA are transformations which show the evolution of prices in terms of some sort of phase space projection, and then people look for patterns in this transformed space.The high frequency data I work with is very high frequency so 'upticks' are usually followed by some random passage of time before another tick is realized, e.g. +1 0 0 0 00 0 +1 0 0 -1 0 0 0 0 0 -1When multiple ticks occur in one second I use an average tick to represent that point.Once you identify common points in phase space you can determine the distribution of the trajectories leaving this point in phase space.QuoteOriginally posted by: farmerQuoteOriginally posted by: crowlogicThis is why bayesian inference, boot-strapping, surrogate data, etc are really useful for checking the validity of any TA based indicators..QuoteOriginally posted by: bskilton81Some of it does work, particularly in FX markets, but people should mind their t-stats. Backtest a strategy and and convince One of my problems with TA is the way it is used in practice (no tests for significance, very small samples, etc). Also, people like to backtest TA strategies using information they couldn't have known at the time of decision, because the information didn't actually come out until later (like using regressions for instance, when one wouldn't have known the parameters until long after the decision was made). This is cheating.I disagree with most of this. The problem with chart analysis, is that you have no idea what you are looking at, what phenomenon is causing the price to move. The only thing that all phenomena have in common is that they happen gradually, they trend. Plus, phenomena share common factors, so some inevitably are correlated at various lags.So if you are looking for something not involving trending or lagged correlation, better and better statistical tricks will just get you more and more unlikely coincidences. And if you are looking for trending and correlation, they are certainly there and cannot hide from your searches, the only reason not is because someone else is already trading them.Maybe you are suggesting that by looking at the distribution of outcomes associated with a given signal, you can tell more than by just looking at the sum of those outcomes. But if you're looking for something other than trending, lagged correlation, and the sound of people trading both, then all this is for nothing. Whereas if something isn't trending or correlated, all you've shown is that someone is already trading it.I don't care what some statistical study tells me, I know that an uptick in bonds is likely to be followed by another uptick. If it doesn't tick up, that's because the next uptick was masked by another trader like me, already selling into it.Basically, there's a whole set of guaranteed money makers out there, and you needn't bother yourself with whether they exist. Microsoft is correlated to Oracle with a lag, don't let any test tell you otherwise! All you need bother yourself with is, of these moneymakers which you know exist without any need for a test, which ones are not already being traded by someone else better than you?Don't kid yourself that you're measuring the nature of the world, of currency, or of interest rates, or of corporate prospects. What you are measuring, when you look for trends, is whether other traders have traded them away.

QuoteOriginally posted by: crowlogicthen people look for patterns in this transformed space.I'm just saying that for 99% of what anybody hopes to find, "patterns" is too general. If you find that every time the price draws an "ssm" it then draws a "vnhs," I don't care how many statistical tests it stands up to. In the other hand, you're holding the simple fact that you're looking at a ticker, which is going to send you simultaneous signals from a thousand other places than the one which has the pattern. So it doesn't matter if you know there's a guy on your conference call singing "the sky is my eye." The next time you hear "sky," it's going to be coming from someone else. The only songs that everyone sings the same are trending, correlation or leading indicators, and trading thereof.I know there are traders claiming to be using "patterns." But they probably just mean combinations of leading indicators, or something. Like I think Aaron Brown said "chords of association," when he could have said "patterns of association." If you literally see a pattern - like a picture - in the ticker, there is nothing on Earth that will convince me it has been produced by the same physical phenomenon two times in a row. And the behaviors that are common to different physical phenomena, are not complicated enough to be called patterns in my lingo.

I agree that looking for things like geometric patterns in charts is pretty silly, but if there are enough people looking at geometric patterns and trading off of them then it becomes a self-fulfilling prophecy with different traders judging the shapes, timing, volume, etc differently so they don't all respond in the same way.Of course I cannot sit here and determine what an individual trader is going to do every time, but the aggregregate behaviour of statistically similiar groups of traders can create emergent behaviour that is semi-deterministic on a potentially infinite number of timescales.I know I'll probably get flamed for this, but go pick up a copy of 'A New Kind of Science' and look at some of the ordered behaviour that can arise from the iteration of extremely simple rules, cellular automata, continus cellular automata, coupled flow lattices, coupled map lattices, differential equations, etc.There are geometric patterns that are self-similiar and fractal but they don't occur in finance probably.. but 'random/stochastic fractals' are a different type that only says things are statistically self-similiar and you dont need to model the behaviour of individual traders to benefit from it.QuoteOriginally posted by: farmerQuoteOriginally posted by: crowlogicthen people look for patterns in this transformed space.I'm just saying that for 99% of what anybody hopes to find, "patterns" is too general. If you find that every time the price draws an "ssm" it then draws a "vnhs," I don't care how many statistical tests it stands up to. In the other hand, you're holding the simple fact that you're looking at a ticker, which is going to send you simultaneous signals from a thousand other places than the one which has the pattern. So it doesn't matter if you know there's a guy on your conference call singing "the sky is my eye." The next time you hear "sky," it's going to be coming from someone else. The only songs that everyone sings the same are trending, correlation or leading indicators, and trading thereof.I know there are traders claiming to be using "patterns." But they probably just mean combinations of leading indicators, or something. Like I think Aaron Brown said "chords of association," when he could have said "patterns of association." If you literally see a pattern - like a picture - in the ticker, there is nothing on Earth that will convince me it has been produced by the same physical phenomenon two times in a row. And the behaviors that are common to different physical phenomena, are not complicated enough to be called patterns in my lingo.

Basically, I take plausible explanations over statistical tests. No test in the world is going to convince me that the S&P 500 draws a face on Fridays. And no test in the world is going to convince me that Microsoft and Oracle aren't correlated.And I really only know one plausible explanation: trends, correlation or leading indicators, and people trading them. No search can reveal something new to me, like I never thought of that. Only some unknown combination of those things. And everything can be explained as that plus noise.The statistical tests just fill in the missing numbers in this preconceived structure or topology. So what am I looking for? A pattern? No, I am looking for the correlation between two things or the common factor, the trendiness, and the relative number of people trading it.Traders have patterns, but they probably aren't very stable. Their complex patterns either A) quickly adapt to the underlying phenomena, which are simple and smooth, or B) are wild and unpredictable.