Has technical analysis been thoroughly debunked in the bitcoin community yet?

[bassclef]

I can't understand why smart people who "get" what Bitcoin is have such a problem with TA. It's psychology.

I can take any excited bull market run-up and use TA to estimate how much it will retrace by using fib retracements and/or law of threes. To say there's no merit in it is ridiculous. Will it tell me exactly? No. But I can get a lot closer than guessing.

If you're searching random indicators, reading about them for five minutes on stockcharts.com and thinking you can make money, you're doing it wrong.  

[FelixOliver]

TA isn't psychology or a predictive window into the future. The charts can only relay to you what has already happened. So, in essence, trading using TA alone is like driving a car, and using just the rear view mirror.

That doesn't mean that there aren't some technical wizards making serious bank from recognizing and spotting mere patterns here and there.. But i think that has more to do with luck than any type of skill

[Raystonn]

Neural networks, such as that in your brain, are also mere pattern recognition machines working exclusively with historical data.  Would you say anything humans achieve by applying their brains is also just luck?  Perhaps there is value in assuming some patterns will repeat.

[segeln]

I can't understand why smart people who "get" what Bitcoin is have such a problem with TA. It's psychology.
I can take any excited bull market run-up and use TA to estimate how much it will retrace by using fib retracements and/or law of threes. To say there's no merit in it is ridiculous. Will it tell me exactly? No. But I can get a lot closer than guessing.
If you're searching random indicators, reading about them for five minutes on stockcharts.com and thinking you can make money, you're doing it wrong.  

and psychology of market participants results in chart-Patterns which can be studied and found.then you can say,what direction the market will take based on allways (at least most) the same psychological behaviour
"it is psychology ,stupid"

[Este Nuno]

Neural networks, such as that in your brain, are also mere pattern recognition machines working exclusively with historical data.  Would you say anything humans achieve by applying their brains is also just luck?  Perhaps there is value in assuming some patterns will repeat.

Not necessarily always luck, but humans often fall prey to confirmation bias. In reality proper application of statistical analysis can tell us much more about a data set than simple pattern seeking.

[Este Nuno]

All of these indicators that people use should be able to be easily proved by back testing them in a large enough sample of data.

Is it theoretically possible that someone could come up with a system based on TA that can show a profit when automatically applied over a statistically significant about of time? Yes.

Is it likely at all that someone drawing lines on charts and attempting to seek patterns is going to accurately predict something that when viewed in a short time frame acts similar to a random walk? No.

[aminorex]

All of these indicators that people use should be able to be easily proved by back testing them in a large enough sample of data.

Is it theoretically possible that someone could come up with a system based on TA that can show a profit when automatically applied over a statistically significant about of time? Yes.

Is it likely at all that someone drawing lines on charts and attempting to seek patterns is going to accurately predict something that when viewed in a short time frame acts similar to a random walk? No.

My monkey does pretty well for himself (and me).

[Marbit]

All of these indicators that people use should be able to be easily proved by back testing them in a large enough sample of data.

Is it theoretically possible that someone could come up with a system based on TA that can show a profit when automatically applied over a statistically significant about of time? Yes.

Is it likely at all that someone drawing lines on charts and attempting to seek patterns is going to accurately predict something that when viewed in a short time frame acts similar to a random walk? No.

My monkey does pretty well for himself (and me).

Yeah, I do okay too. I've always managed to profit in both coin and $$ in the long run. If you can pull a 60% win rate, that's pretty damn good, to be honest. It's all about cutting out your bad trades early and letting the good ones run.

Some of you anti-TAers around here are just using empty rhetoric. I realized I had a knack for trading the same way I had a knack for online poker (too bad liquidity is dead there for US players). Reading probabilities and managing risk. That's it.

[IMZ]

I came to cryptos from a background that included reading libertarian and contrarian economic articles. Market Oracle comes to mind. The clincher with Elliiot Double Shoulder Trend Line Uppy Downy Wisdom Corp is that the authors are always incommunicado. Overall, I suggest that they rely on a revolving-door reality of impressed noobies who nibble nibble; win a little or lose a lot; and drift off.

And . . . I don't think their rules would apply anyway to Bitcoin.

Mark Blair, Unicup, Western Australia

[Wandererfromthenorth]

I can't understand why smart people who "get" what Bitcoin is have such a problem with TA. It's psychology.

I can take any excited bull market run-up and use TA to estimate how much it will retrace by using fib retracements and/or law of threes. To say there's no merit in it is ridiculous. Will it tell me exactly? No. But I can get a lot closer than guessing.

If you're searching random indicators, reading about them for five minutes on stockcharts.com and thinking you can make money, you're doing it wrong.  

People here saying "TA is total bullshit" are just noobs that tried to daytrade but lost money because they had no idea of what they were doing.
It's okay though. Basic TA is nothing particularly hard to understand and apply in order to increase your holdings, but some people are just too lazy to learn something new and try to apply it.

They prefer to "HODL", and they think daytrading is just gambling, so that you know, it's not really their fault if they lost money...

HODLING is fine if you don't have time for daytrading and if you want to be a more relaxed investor in BTC, but please don't go around screaming general statement like "TA is bullshit" or "daytrading is gambling! the price just does whatever it wants", that just makes you look like a noob.

[Este Nuno]

All of these indicators that people use should be able to be easily proved by back testing them in a large enough sample of data.

Is it theoretically possible that someone could come up with a system based on TA that can show a profit when automatically applied over a statistically significant about of time? Yes.

Is it likely at all that someone drawing lines on charts and attempting to seek patterns is going to accurately predict something that when viewed in a short time frame acts similar to a random walk? No.

My monkey does pretty well for himself (and me).

Yeah, I do okay too. I've always managed to profit in both coin and $$ in the long run. If you can pull a 60% win rate, that's pretty damn good, to be honest. It's all about cutting out your bad trades early and letting the good ones run.

Some of you anti-TAers around here are just using empty rhetoric. I realized I had a knack for trading the same way I had a knack for online poker (too bad liquidity is dead there for US players). Reading probabilities and managing risk. That's it.

Feeling that you have an intuitive understanding of trading and the market doesn't mean that TA works. Also having a 60% win rate doesn't tell us anything about the EV of a set of trades.

To use a poker analogy, I feel most people who follow TA are the same people who call a naked gutshot on the turn with no showdown value against a bet that doesn't give them odds because they 'feel' it. Except in this case they draw lines and use fancy sounding indicators that they think predict something that is incredibly complex underneath it all.

I think there is lots of potential for people to apply a quantitative approach to any market, but this type of thing is so far beyond your average internet TA guy that it might as well not be considered. Most people doing that sort of thing would never think of sharing their work. Or they work for a company that contractually forbids them from doing so.

I think I can sum up my feelings on TA this way: If you're not a quant, anyone who only looks at the 'what' of the market, without asking 'why' is essentially gambling. You can have 50 fancy lines and curves on your graph but without understanding the underlying movements they are meaningless and only tell you what has happened. They are not predictive. And if they are, they can be statistically proven to have a correlation with the price and anyone not doing the math to show that is just trying to find water with metal rods.

[Wandererfromthenorth]

Is it likely at all that someone drawing lines on charts and attempting to seek patterns is going to accurately predict something that when viewed in a short time frame acts similar to a random walk? No.

You have a wrong idea about what basic TA done right actually allows the trader to do.
TA doesn't just tell "Do THAT" at whatever point in the graph. TA doesn't just allows to draw exactly what the price will do in the future. TA allows you to predict scenarios and react accordingly.

Simple example for anybody to get:   You see a huge triangle pattern forming, soon the price will have to breakout of that triangle.
TA doesn't necessarily tell you if we are going to go up or down after (depends, sometimes it almost does), but if we break out of the pattern and we start to go up it's a strong buy signal, if we start to go down it's definitely time to sell. Of course look at an increase in volume at the moment of the breakout in order to confirm it.
That's it.

If you think TA is supposed to tell you "BTC's price will be at $5032.04 on 12th of March 2015", you have a wrong idea about what TA is supposed to be.

[Wandererfromthenorth]

All of these indicators that people use should be able to be easily proved by back testing them in a large enough sample of data.

Is it theoretically possible that someone could come up with a system based on TA that can show a profit when automatically applied over a statistically significant about of time? Yes.

Is it likely at all that someone drawing lines on charts and attempting to seek patterns is going to accurately predict something that when viewed in a short time frame acts similar to a random walk? No.

My monkey does pretty well for himself (and me).

Yeah, I do okay too. I've always managed to profit in both coin and $$ in the long run. If you can pull a 60% win rate, that's pretty damn good, to be honest. It's all about cutting out your bad trades early and letting the good ones run.

Some of you anti-TAers around here are just using empty rhetoric. I realized I had a knack for trading the same way I had a knack for online poker (too bad liquidity is dead there for US players). Reading probabilities and managing risk. That's it.

Feeling that you have an intuitive understanding of trading and the market doesn't mean that TA works. Also having a 60% win rate doesn't tell us anything about the EV of a set of trades.

To use a poker analogy, I feel most people who follow TA are the same people who call a naked gutshot on the turn with no showdown value against a bet that doesn't give them odds because they 'feel' it. Except in this case they draw lines and use fancy sounding indicators that they think predict something that is incredibly complex underneath it all.

I think there is lots of potential for people to apply a quantitative approach to any market, but this type of thing is so far beyond your average internet TA guy that it might as well not be considered. Most people doing that sort of thing would never think of sharing their work. Or they work for a company that contractually forbids them from doing so.

I think I can sum up my feelings on TA this way: If you're not a quant, anyone who only looks at the 'what' of the market, without asking 'why' is essentially gambling. You can have 50 fancy lines and curves on your graph but without understanding the underlying movements they are meaningless and only tell you what has happened. They are not predictive. And if they are, they can be statistically proven to have a correlation with the price and anyone not doing the math to show that is just trying to find water with metal rods.

Again, your idea of the way TA is supposed to predict markets is wrong IMHO.

[Wandererfromthenorth]

If your question about TA was "Using TA, can we have a pretty good approximation of what the price of BTC will do in 2015?", the anwer is clearly no, but that is obvious, and it doesn't seem to be what you are talking about.

If your question instead is: "Using TA (done right), can we predict and/or react to certain price movements in order to profit?" the anwer is yes.

[Este Nuno]

No, I don't think that at all.

Even if you use your TA to give a price range and corresponding probabilities I still think it's no better than random unless you can statistically prove that its not.

[Wandererfromthenorth]

I significantly increased my BTC holdings using basic TA applied to BTC trading, a lot of other people here too.
Are you just telling me that you think that overall there are no traders that are having better performances than others? that basically if you stick to trading long enough every trader's probability of making a profit tends to 0? just like playing the slot machines at a random casino?

Because that's what your statements imply.

[oda.krell]

No, I don't think that at all.

Even if you use your TA to give a price range and corresponding probabilities I still think it's no better than random unless you can statistically prove that its not.

I get where you're coming from. "If it works, it should be possible to _show_ that it works." (where "show" means something like "up to current academic publication standards"). Did I get that right?

Let me make a countering case. I suggested the analogy before in some other thread, so apologies for the repetition:

Imgagine you're a computer scientist in the 1970s. You implemented a computer chess algorithm on the fastest then-available machine, nothing more complicated than alpha-beta pruning (or variations of it). You still fail to beat a competent human GM. Consistently.

So, you ask the GM how he plays chess. He'll throw an entire library of opening theory (or endgame theory, or whatever) at you. You read it, and will quickly realize that this completely unimplementable. Worthless, from your perspective. Yet, still, that GM beats your machine nearly every single time.

End of analogy. In my opinion, TA as she is practiced by the more competent traders in here, is as much of an art as it is a science (thanks to sgbett for that phrasing). It's difficult (though not impossible) to show that it produces statistically significant results. And even then, what is tested is only a tiny subset of what traders actually use for the TA: the purely algorithmic part, while in reality the algorithmic methods go hand in hand with intuition/human-optimized pattern recognition/etc. (like in the chess analogy above).

It really runs down to the following question, in my opinion: do you only accept knowledge and results that are produced by the full rigor of academic methodology, or do you allow for "conditional knowledge"... insights that appear reasonable to you, that you have personal annecdotal evidence for, and that you hope can eventually be proven to be correct in a more formal way.

Myself, I follow the latter approach: I believe I have tentative evidence that TA (as I practice it) works well enough, but I don't expect to be able to show beyond a doubt that it does work, because (as you already pointed out) the system underlying it is too complex (and not well enough understood yet) to formalize it to the point where the question can really be answered once and for all. Until then, I will continue using those methods, with tight risk control in place to avoid catastrophic failure should the methods "stop working" for me.

[RyNinDaCleM]

No, I don't think that at all.

Even if you use your TA to give a price range and corresponding probabilities I still think it's no better than random unless you can statistically prove that its not.

I get where you're coming from. "If it works, it should be possible to _show_ that it works." (where "show" means something like "up to current academic publication standards"). Did I get that right?

Let me make a countering case. I suggested the analogy before in some other thread, so apologies for the repetition:

Imgagine you're a computer scientist in the 1970s. You implemented a computer chess algorithm on the fastest then-available machine, nothing more complicated than alpha-beta pruning (or variations of it). You still fail to beat a competent human GM. Consistently.

So, you ask the GM how he plays chess. He'll throw an entire library of opening theory (or endgame theory, or whatever) at you. You read it, and will quickly realize that this completely unimplementable. Worthless, from your perspective. Yet, still, that GM beats your machine nearly every single time.

End of analogy. In my opinion, TA as she is practiced by the more competent traders in here, is as much of an art as it is a science (thanks to sgbett for that phrasing). It's difficult (though not impossible) to show that it produces statistically significant results. And even then, what is tested is only a tiny subset of what traders actually use for the TA: the purely algorithmic part, while in reality the algorithmic methods go hand in hand with intuition/human-optimized pattern recognition/etc. (like in the chess analogy above).

It really runs down to the following question, in my opinion: do you only accept knowledge and results that are produced by the full rigor of academic methodology, or do you allow for "conditional knowledge"... insights that appear reasonable to you, that you have personal annecdotal evidence for, and that you hope can eventually be proven to be correct in a more formal way.

Myself, I follow the latter approach: I believe I have tentative evidence that TA (as I practice it) works well enough, but I don't expect to be able to show beyond a doubt that it does work, because (as you already pointed out) the system underlying it is too complex (and not well enough understood yet) to formalize it to the point where the question can really be answered once and for all. Until then, I will continue using those methods, with tight risk control in place to avoid catastrophic failure should the methods "stop working" for me.

Well put!

I will also ask, who does the burden of proof fall on? Those of us that successfully use TA have all the proof that we need. The naysayers want proof that it works... It's very difficult if not impossible to prove either way. But it's the dismissal of something they don't understand which gets me. If you had a bad experience trying to use TA then maybe it's the execution rather than the idea.

The indicators we use are (for the most part) lagging, but most still show the change in momentum and force well before the actual change in direction of price. Many indicators show that one outcome is more probable than another. There is psychology behind the patterns that we pick out. EW is heavily based on psychology. Fibonacci (which I feel I have proven to work in my thread) seems random, but until it fails to produce, I will continue using it.

I think the burden of proof falls on the naysayer. This isn't like a "Prove the existence of God" thing. At least TA has something you can actually look at if you take the time to try.

[Marbit]

How about when very successful traders like Thomas Bulkowski give their statistical results? For example, a falling wedge results in an upward breakout 68% of the time. A descending triangle results in a downward breakout 64% of the time.

http://thepatternsite.com/fallwedge.html
http://thepatternsite.com/dt.html

Is successfully trading those patterns "random"?

@Este Nuno -- what are your thoughts? From what you've said (just rhetoric, really), and in reality, empirical proof is lacking and must be. But that goes both ways.

Yeah, I do okay too. I've always managed to profit in both coin and $$ in the long run. If you can pull a 60% win rate, that's pretty damn good, to be honest. It's all about cutting out your bad trades early and letting the good ones run.

Some of you anti-TAers around here are just using empty rhetoric. I realized I had a knack for trading the same way I had a knack for online poker (too bad liquidity is dead there for US players). Reading probabilities and managing risk. That's it.

Feeling that you have an intuitive understanding of trading and the market doesn't mean that TA works. Also having a 60% win rate doesn't tell us anything about the EV of a set of trades.

To use a poker analogy, I feel most people who follow TA are the same people who call a naked gutshot on the turn with no showdown value against a bet that doesn't give them odds because they 'feel' it. Except in this case they draw lines and use fancy sounding indicators that they think predict something that is incredibly complex underneath it all.

I didn't say it proved TA works. I don't think empirical proof is possible. I said I started trading because I was successful in poker, which depends on similar statistical analysis. In my data set, Fib retracements, moving averages, certain candlestick patters, and momentum/volume flow indicators all hold statistical significance. And actually, I'm under the impression that a win rate does tell you about the EV of a given set of trades, assuming proper risk management.

Comparing TA that incorporates significant backtesting to "calling a naked gutshot on the turn" is empty rhetoric. That's plain for all to see.

[Wary]

No, I don't think that at all.

Even if you use your TA to give a price range and corresponding probabilities I still think it's no better than random unless you can statistically prove that its not.

I get where you're coming from. "If it works, it should be possible to _show_ that it works." (where "show" means something like "up to current academic publication standards"). Did I get that right?

Let me make a countering case. I suggested the analogy before in some other thread, so apologies for the repetition:

Imgagine you're a computer scientist in the 1970s. You implemented a computer chess algorithm on the fastest then-available machine, nothing more complicated than alpha-beta pruning (or variations of it). You still fail to beat a competent human GM. Consistently.

So, you ask the GM how he plays chess. He'll throw an entire library of opening theory (or endgame theory, or whatever) at you. You read it, and will quickly realize that this completely unimplementable. Worthless, from your perspective. Yet, still, that GM beats your machine nearly every single time.

End of analogy. In my opinion, TA as she is practiced by the more competent traders in here, is as much of an art as it is a science (thanks to sgbett for that phrasing). It's difficult (though not impossible) to show that it produces statistically significant results. And even then, what is tested is only a tiny subset of what traders actually use for the TA: the purely algorithmic part, while in reality the algorithmic methods go hand in hand with intuition/human-optimized pattern recognition/etc. (like in the chess analogy above).

It really runs down to the following question, in my opinion: do you only accept knowledge and results that are produced by the full rigor of academic methodology, or do you allow for "conditional knowledge"... insights that appear reasonable to you, that you have personal annecdotal evidence for, and that you hope can eventually be proven to be correct in a more formal way.

Myself, I follow the latter approach: I believe I have tentative evidence that TA (as I practice it) works well enough, but I don't expect to be able to show beyond a doubt that it does work, because (as you already pointed out) the system underlying it is too complex (and not well enough understood yet) to formalize it to the point where the question can really be answered once and for all. Until then, I will continue using those methods, with tight risk control in place to avoid catastrophic failure should the methods "stop working" for me.

If it's impossible to prove workability of TA by analysing TA itself, it still can be possible by analysing success rate of TA traders: if an TA trader can consistently demonstrate better than random results, TA works. Although such proof would be hard to get in practice, since most of btc traders won't be happy to submit their trade history to a nosy researcher.  

If the question of TA profitability is not theoretical, but practical one (should I trade using TA?), the answer is easy: you (OP) shouldn't. Because number of coins is limited, this is zero-sum game and for you gaining a btc only if somebody have to lose his. So, you can trade (doesn't matter, with TA or without) only if you are sure that you are playing with bigger fool less talented person, than you are. If you don't know who's the fool in the game, the fool is you (c).

Edits: grammar and politeness.