what points are you using to define this channel? top line seems to have two, but bottom line only has one, unless you can argue for the significance of the top of that bull trap.

i'm glad we're mostly in agreement

you're right of course, all this is fine and dandy but i've got to show that the specific methods of technical analysis are effective at 'uncovering' these patterns. i find that standard indicators like the MACD do indicate simple trends well, as i mentioned earlier, and the properties of price momentum are really effectively displayed by oscillators.

not only do real-world things like closed banks cause regularities, but patterns in human behavior as well. the convenient thing is, we don't have to hypothesize about social psychology -- we just have to rigorously show a pattern in the data.

for the trend-following example, we tend to overreact, and push the price into 'overbought' and 'oversold' states (these are metaphors, admittedly, but give a decent understanding of what the oscillators represent). recognizing that we are in these unstable states helps you anticipate price movement, given other market conditions. i'm reminded of a striking example of this during the rally to $160:

-===-

it is just like physics ain't that cool?

i'd hate to be a talking head, but i'll put in my two cents if you insist.

imho, determining the odds of the price being at a certain point "ever again" is pretty futile. i liked ichthyos model of a gaussian curve for a model of the probability of 'ever' reaching a price point 's' standard deviations away from the present price, or the mean historical price. in this model, the probability of a price point s deviations away approaches zero as s -- > infinity, but is nonzero for all such points.

for time-scales:

short-term: unlikely
mid-term (this year): possible
long-term: see above

executing that on historical price data would be simple. the question is, is it a functioning indicator? is the height above the peak of the distribution indicative of deviation from a trendline? time-scale matters here as well.

yes, i kind of stitched together bullet points of a sort of thesis i'm trying to develop. i apologize for the disorganization.

let me try to build a bridge between the ideas you just mentioned:

the emphasis on information is because trading is a limited information game. this is separate from the price function, which is stochastic, essentially random, which is exactly why no methods of technical analysis can produce 100% confidence.

take the 'weekend dip'. once a thread about it was posted and trafficked to saturation, the effect stopped happening. why? because traders selling early friday and buying back in sunday, profited off of this information asymmetry, anticipating that others were going to sell and undercutting them.

these regularities have been shown to exist, and i do believe they violate the strong form of the EMH, which would imply that a good is at its 'true price' at all times in the marketplace. in practice this falls quite short of ideal. remember how the price doubled suddenly after a long period of consolidation a few months before the block reward halving? the suddenness of the onset of the price discovery phase suggests that the way that price-relevant information is factored into the price is far from ideal, and if you have meta-information -- information of another players' future actions -- you can undercut them profitably.

i actually appreciate this kind of discussion, johnblaze made a good point, and showed he had thought about it critically. it just bothers me when someone comes out of the woodwork and posts a dismissive comment with no constructive criticism at all

i'm sorry if it was dense. and i realize now i didn't exactly connect it to what i quoted. let me try to summarize more clearly:

"""TA can work even if the two conditions you noted are not met. i propose the following:

a trader using the methods of TA attempts to extract more information from the available price data than others who are not looking for regularities (not doing any analysis).

this creates a difference in the information available to each 'player'. this is profitable for the same reason 'insider trading' is profitable."""

i hope i was able to communicate my meaning this time.

straight to zero, of course

edit:

this is a really good model. the normal distribution is a probabilistic representation. i wonder how one would go about calculating the skew?

do you rage at your weatherman, too, because he's not 100% accurate, yet reports on his projections every day?

excellent explanation, rynin! thanks for the write-up

keeping it short, but hopefully clear:

the effects of 'insider trading' suggest that price-relevant information is not factored into the price in an ideal way, but the Efficient Market Hypothesis suggests that most relevant information is 'encoded into' or somehow represented in price -- this information including market sentiment, greed/fear, risk-aversiveness of other players, and the like.

this leads to the idea that profitable information asymmetries between players can form for short periods of time.

from a mathematical point of view, it has been shown (see the Hurst exponent) that autocorrelations exist in price data -- that is, regularities which can be profitably anticipated.

as such, in a simple way, technical analysis is an analysis done on price data to find patterns -- essentially, identifying how the price is deviating from its natural stochastic (random) nature.

and this demonstrably occurs! trend-following is very successful in certain market conditions (see goomboo's journal). this leads to the idea of 'momentum', that it is statistically unlikely for the price to suddenly violate a trend. another simple strategy is measuring the 'rate of change' of price, because 'momentum changes before price changes' follows from basic rules of derivatives in calculus.

i hope i have showed you how it is possible to get at information that may be 'coded' into the price, that others do not have access to, that creates a profitable information asymmetry between players.

--arepo

sorry about that, i must have just read a post of yours when i was writing that. at least i didn't name electric -- he hates triangles hah

edited in the original

--

i always find it fascinating how many layers of self-similarity there are in the price function. this is the third 'public test' of fractal analysis, which i'm a little suspicious of because it is lacking in rigor in that it is just as prone to apophenia as elliot waves. i've been outlining a number of shapes that i can define in rigorous terms, and once that list is complete i intend to do statistical analysis on price history data to hone the method's precision.

until then i've been focussing on the 'bubble shape', a type of bearish reversal that occurs across varying scales in many bullruns in bitcoin price history. i've identified three versions of this pattern in price data since just April 1, and used their shape with great success in anticipating price movements in similar market conditions. it's just an amazing idea that the larger shape of price can mimic smaller structures that have already formed within its history, exactly like a mandelbrot set!

since the 'fractal hypothesis' is still in its infancy, i encourage feedback from the sharper minds of the community i've learned so much from other posters here, that i've come to see it as a community of very intelligent people, and an excellent collaborative atmosphere for new ideas,

--arepo

bear markets make people really unhappy also tons of newbies who haven't acclimated to etiquette and such. i'm not surprised proudhon left 

we don't have to be enemies... this is a community i have nothing against you, fitty.

*10-day hourly scale*



-===-

looking awfully oversold on the short-term scale. the downward momentum has been exhausted. i expect a downwards breakout from the pennant evolve outlined, but we won't break under the $100 support.

-===-


-===-

instead, the oversold signal confirms some fractal analysis that suggests we'll retest but bounce significantly into a larger bearish pennant, peaking and then breaking downwards again to test the $100 support in the next 48 hours.

-===-



-===-

will $100 hold then is not yet clear.

--arepo

same person? i smell an agenda.

market sentiment lags price perfectly.

good point -- as i like to say, if there are methods that work, they can be considered functional analysis. if there aren't, why is everyone here wasting their time with a cointoss? there's satoshidice for that

i think the bear market is rustling all the newbies' jimmies

there are many others threads on this board dedicated to the question of whether or not methods of TA can be considered 'effective'. if you'd like to have that conversation, you can continue it there, or at least supply some kind of new argumentation besides the bald assertion that it doesn't work because it doesn't.

as i said before (i'd hate to be the tyrant who self-moderates) so discussion and presentation of TA ITT please. (assertions don't count as discussion )

--arepo

edit: it's hilarious how quiet this thread was as my many projections last week unfolded, but as soon as i get one thing wrong, the TA-naysayers are back