How to properly use Technical Analysis | Page 5 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

How to properly use indicators 213 replies

How much Memory is needed to properly run MT4 Terminal 10 replies

Do you backtest your systems PROPERLY? 0 replies

Custom Indicator not Drawing Properly? 0 replies

Attachments: How to properly use Technical Analysis
Exit Attachments

How to properly use Technical Analysis

  • #81
  • Edited 11:25pm Jan 1, 2012 10:37pm | Edited 11:25pm
Quoting GEfx
Hi Red, sorry I've been watching football today and just saw your reply. OK, I see what you're talking about, and to be honest, I've never done that analysis. I will think about how I would conduct that analysis and also check with my trading team members to see what they think. As for my 4H and 1H analysis, here is a quick summary of my charts re-drawn on naked charts:

Attachment

Attachment

I think based on these two charts, you could make a case for going long to targets above, and use your red line as a game...

you are more than welcome to contribute

so how do you use your fibs GEFX? and the trend lines.

at nubcake- true that everything has two sides especially in a bear market sell orders will line up, however, fib Retracements are meant to enter while extensions to look for possible targets to exit. in other words you should already be in the trade, and looking at your, chart geometry, assessing whether continuation is probable. For instance, limit orders might very well be placed at that extension level to which you have 3 choices.

1- liquidate position and enter shorts
2 liquidate positions and await for set ups or for your analysis to see what to do next
3- hold to see how far the retrace goes and if in effect you are at risk of profit depletion liquidate, if the market continues to take this up the n ride it.
AVT INVENIAM VIAM AVT FACIAM
  • #82
  • Jan 2, 2012 11:14pm Jan 2, 2012 11:14pm

Quoting nubcake
fibs are the biggest wank in the history of trading wankery.

... It's Sooo True...

A lot of people are struggling to comprehend this very small hurdle.
And if 95% can't overcome this small hurdle,
then let it be so,
that the 95% will never make a living in this endeavor.


Quoting the redlion
i actually don't know, i am finding it hard to come up with parameters that will give me objective data.
suggestions would be welcomed.
Throw it in the bin.
  • #83
  • Jan 3, 2012 2:20am Jan 3, 2012 2:20am
Quoting nubcake
don't know dude. i'm sure you are 100 times more schooled in statistics than i am.

the first issue is how do you define a swing point to start from. do you give it a time-based criteria such as "if price does not break a high by n bars it is considered a top"?

i'd probably then look at just polling how far price moves between the 2 outer 'levels'. e.g. in terms of the 38 level if price retraces down past 50% (by even just a point) then how much further does it go before going back up through the 50% point, or did it simply go all the...


Nub,

I'm proud of ya bro. You are really excelling and making very cogent arguments. You are absolutely right. It does not matter how a person wants to trade, the most important first step is to define exactly what you are doing. It's impossible to figure out if something works (or has worked) if the rules are not absolutely clear.

As for fibs, where does a person draw them for the sake of testing? My suggestion, start simple--even by asking critical questions. Why should fibs for example (arbitrary percentages) have any special place within price data? We've all heard the quick reflex answers, but you better not stop there. Start testing!

A quick example, I just did a quick and simple test with this criteria:

When the daily high price is greater than the previous days high, what percentage of yesterdays range did price break above yesterdays high? This is basically applying a fib to the daily ranges. My test only took into consideration the very next days PA. Now, it is not uncommon for traders to apply fibs to daily ranges. If there is anything special, for example, about the 1.382 or 1.618 projection I'd expect to see this evident in statistical data. For this test, both of those fib projections are really nothing special. In fact, you'll notice what one might easily hypothesize without having done any testing at all and that is this: The probability for price moving rapidly above the yesterdays high quickly falls off every 10% of yesterdays range...The chart here speaks for itself. I did this test using daily EU price data from December of 2006 to October 2011. It's a big enough sample to at least get an idea and I'd posit that the picture presented here would remain largely unchanged with a larger sample...Anyways, I will shut up now, just wanted to give an example of what I mean when I say "testing."

Happy New Year BTW!! Good trading to all.
Attached Image (click to enlarge)
Click to Enlarge Name: % Break Chart.JPG Size: 100 KB
  • #84
  • Jan 3, 2012 3:36am Jan 3, 2012 3:36am
Quoting Scotty B
I did this test using daily EU price data from December of 2006 to October 2011. It's a big enough sample to at least get an idea and I'd posit that the picture presented here would remain largely unchanged with a larger sample...

  1. Are you assuming that there will be no change from 2012 to 2020?
  2. So, what if from 2012 to 2020 became a reversal trend against your test results?
  3. So, you've never taken into account about change in trend/randomness and black swans?

Have you even considered the above points?... Because if you have, it would render your test useless in such events. All those test will be useless. I'm pretty sure you've heard people sulking about how "the market has changed" ?

  • #85
  • Jan 3, 2012 3:59am Jan 3, 2012 3:59am
Quoting nubcake
.... then i'd take a look at all of the resulting data, scratch my head, and toss it all away since i'm sure it would all make no sense to me at all.
hmm...thats what happens to me a lot, im absolutely wank when it comes to funky stats like that.

Quoting the redlion
i actually don't know, i am finding it hard to come up with parameters that will give me objective data.

suggestions would be welcomed.
Have you considered gathering statistics based on highs and lows of 'a' candle?

Price will break the high/low of previous candle, and continue.

Price will bounce from the high/low of previous candle, and reverse.

Everything else seems confusing to me, since the above is what i see happening over and over again, irrelevant of how far back in time one goes.
  • #86
  • Jan 3, 2012 9:33am Jan 3, 2012 9:33am
scotty B is right, but so is crucial point.
but how can this be.

the only way to obtain objective data analysis is by statistical data, however the market will not move as expected all the time, in fact the CHART is MERELY A TRANSACTION RECORD, having said that statistics give you a clear picture of BASELINE expectations to which your observations determine whether price is out of line with past historical behavior.

Martin D Evan's publications on the Market micro structure, a research paper from Georgetown University.

and i tend to agree, with his model

the market moves in average about 40% according to Fundamental analysis, due to the economic conditions, and the climate it creates in a broad spectrum broad term view however it is 60% approx moved by information.

1 interbank trading, risk offsetting, covering position, shoring up inventories, etc..... in a inter dealer (dealer meaning banks, market maker, brokers etc)
2 the individual orderflow is not centralized and does not IMEDIATELY OR IPSO FACTO impact the market, however in position managing of the inter dealer trading.

information used is their individual orderflow, and their in house books also sometimes force dealers to seek trade in currencies to satisfy their demand.

the market more or less moves in tandem due to fear of arbitrage so that keeps dealers honest.

and in effect does explain the trend and their intraday variations.

also if you are familiar with the random walk theory and the various experiments creating purely random charts that create formations, and interact well with current technical analysis tools that efficient market theorist put forth as argument that TA is moot. That in fact sheds light on the fact that it is more important to look at the market in the sense of prevalence of orderflow (not the Dark star orderlfow theory but the real orders executed) as opposed to put to much importance in individual variations.


there fore when trading i take the view of the market being NOT RANDOM BUT UNPREDICTABLE, my analysis are grounded in statistics but my money management ASSUMES THE MARKET IS PURELY RANDOM. i invite you to contribute to the discussion of these ideas.
__________________
AVT INVENIAM VIAM AVT FACIAM
  • #87
  • Jan 3, 2012 10:56am Jan 3, 2012 10:56am
The market is not random, that's a proven fact. You can differentiate between a random walk chart and a real chart - the real one has heteroskedasticity. There's a famous experiment when somebody sent to a bunch of technical analysts a series of charts and asked them which one are real and which ones are fake. Most failed to identify them, however a few were aware of the volatility clustering effect and succeeded in differentiating most of the charts. But unknown to them, the sender was expecting this and also included a few fake charts which simulated clustering and tricked even them

I'm also of the view that orderflow is very important. That's why I like to trade imbalances (ie: be a liquidity provider). But this is easier said than done.

http://dl.dropbox.com/u/190212/ff/imbalance.png
  • #88
  • Jan 3, 2012 10:58am Jan 3, 2012 10:58am
Quoting CrucialPoint

  1. Are you assuming that there will be no change from 2012 to 2020?
  2. So, what if from 2012 to 2020 became a reversal trend against your test results?
  3. So, you've never taken into account about change in trend/randomness and black swans?

Have you even considered the above points?... Because if you have, it would render your test useless in such events. All those test will be useless. I'm pretty sure you've heard people sulking about how "the market has changed" ?

Yeah, I don't think that the market will shift too much towards the right (speaking in regard to the chart I posted) with further data. We have to start with what we know about the market, thats what I'm doing. All my test shows is that the market, on breakouts , makes only between a 0 and 20 percent break of the last days range the most. This doesn't mean that the market will not continue to make spectacular breakouts, it will, but it does suggest that maybe a trailing stop would be useful.

Instead of dwelling on theoreticals, counter my evidence with better evidence. Show me where the market is changing specifically so that we can all see how it relates to the results I posted.
  • #89
  • Jan 3, 2012 11:11am Jan 3, 2012 11:11am
Quoting Adal
The market is not random, that's a proven fact. You can differentiate between a random walk chart and a real chart - the real one has heteroskedasticity. There's a famous experiment when somebody sent to a bunch of technical analysts a series of charts and asked them which one are real and which ones are fake. Most failed to identify them, however a few were aware of the volatility clustering effect and succeeded in differentiating most of the charts. But unknown to them, the sender was expecting this and also included a few fake charts which simulated...
I've said a few times lately that the market taken as a whole (all price data) does approximate a random walk, however, the market is not random. There are what I have found to be informative periods of trade regarding near term direction. The true randomness of the market dilutes the signals if all price data is counted.

You're da man, Adal.
  • #90
  • Jan 3, 2012 1:12pm Jan 3, 2012 1:12pm
Quoting Adal
The market is not random, that's a proven fact. You can differentiate between a random walk chart and a real chart - the real one has heteroskedasticity. There's a famous experiment when somebody sent to a bunch of technical analysts a series of charts and asked them which one are real and which ones are fake. Most failed to identify them, however a few were aware of the volatility clustering effect and succeeded in differentiating most of the charts. But unknown to them, the sender was expecting this and also included a few fake charts which simulated...
i never thought of price being heteroskedastic, most statistical and volatility studies assume the homoscedasticity of price dispersion, i mean, it makes perfect sense, as information driven orders will create RELATIVE volatility or price variance that changes according to CURRENT order distribution.

would you care to expand more on the matter, this is really interesting stuff.

so then Heisenberg's uncertainty principle is even more relevant to the market than i thought. Maybe I am mistaken but then Range can be viewed as relative and General Direction is much more important, so the less precise the greater the chance of being right?
AVT INVENIAM VIAM AVT FACIAM
  • #91
  • Jan 3, 2012 1:23pm Jan 3, 2012 1:23pm
however even if price is Random like, but not random with heteroskedastic distribution periods.

in order to control the ONLY thing we can control in trading, and that is RISK. ONE must, base exposure on Average price behavior, not only to have a baseline for analysis but to have defined risk parameters.

so to measure strength or weaning strength of current impulse waves then we are going to have explore the MOMENTUM analysis. Even with General direction through order consensus, when participants no longer feel as strongly in buying up or selling the underlying security, we see a delay in shift in market conditions.

the loss of momentum, be it by, MACD, Trend lines, or linear projection, pattern recognition of distribution and consolidation, some how this is evident.
AVT INVENIAM VIAM AVT FACIAM
  • #92
  • Jan 3, 2012 1:49pm Jan 3, 2012 1:49pm
Quoting Adal
volatility clustering effect
...
I'm also of the view that orderflow is very important. That's why I like to trade imbalances (ie: be a liquidity provider). But this is easier said than done.

http://dl.dropbox.com/u/190212/ff/imbalance.png
"liqudity provider" -- too familiar with that

I'm unsure what your chart is to represent, volatility clustering and overflow imbalances I guess. To me, though, you drew a bunch of small trend angles.

Those have fascinated me for a good while: They're highly intuitive to read and make for a challenge to build technical studies around.
Here's an untested approach to interpreting them. It's proportion-, rather than balance- or momentum-, based.
Attached Image (click to enlarge)
Click to Enlarge Name: imbalance Adal -- adjusted.png Size: 97 KB
  • #93
  • Jan 3, 2012 1:55pm Jan 3, 2012 1:55pm
i am unfamilar with trend angle studies, what are your parameters and the rationale behind trend angles? is it possible you give a detail explanation, if it is not too much to ask Ha-pattern.

and also i know you are a pretty knowledgeable guy when it comes to pattern recognition. MY question for you is when looking at patterns what is it that you are looking for exactly?
AVT INVENIAM VIAM AVT FACIAM
  • #94
  • Jan 3, 2012 2:05pm Jan 3, 2012 2:05pm
Quoting the redlion
i am unfamilar with trend angle studies, what are your parameters and the rationale behind trend angles? is it possible you give a detail explanation, if it is not too much to ask Ha-pattern.

and also i know you are a pretty knowledgeable guy when it comes to pattern recognition. MY question for you is when looking at patterns what is it that you are looking for exactly?
I don't know, sometimes Adal and others post highly original ideas and I get inspired. Not a lot of charts go thru the nontrading Trading Discussion threads, so it's easier to filter for them.

My point was, I'm also fairly unfamiliar with them; the chart shown is only a narrowly-considered approach.
I've toyed with the notion of trend angle studies, only I've never gotten past balance-and-momentum groupings, which is a poor substitute.

I have made progress away from your individualized, any-situation type of patterns -- triangles, head and shoulders and what not -- and toward defining them in terms of simpler geometric terms (dot, line, curve; etc.) to tie them with other ways to draw on the chart.

Thing is, I keep ending up with nice lists of ways to draw, and my trading never improves due to it. Strategy/tactics on how they're used is far more important, something I'm still a rank newbie at.
  • #95
  • Jan 3, 2012 3:28pm Jan 3, 2012 3:28pm
Quoting ha-pattern
I don't know, sometimes Adal and others post highly original ideas and I get inspired. Not a lot of charts go thru the nontrading Trading Discussion threads, so it's easier to filter for them.

My point was, I'm also fairly unfamiliar with them; the chart shown is only a narrowly-considered approach.
I've toyed with the notion of trend angle studies, only I've never gotten past balance-and-momentum groupings, which is a poor substitute.

I have made progress away from your individualized, any-situation type of patterns -- triangles, head and...
I think Adal is actually using orderflow information on those charts, those are not trendlines. Correct me if I'm wrong Adal, but are those not periods of net positive and negative flows. I notice during one of those consolidations that there are several indication of net selling and then a strong drop. Cool stuff.
  • #96
  • Jan 3, 2012 3:29pm Jan 3, 2012 3:29pm
Quote
i never thought of price being heteroskedastic, most statistical and volatility studies assume the homoscedasticity of price dispersion
Price heteroskedasticity is not new, it was discovered in the 80s, if not even earlier. I think you've been reading the wrong studies. Instead of reading papers coming from the economists side, which assume EMH, rational investors and the like, try reading from the quant side, which actually measure the real market and study behavioral patterns.

Today almost all volatility models assume it, with the GARCH model (and it's offspring) being the most famous.

This stuff is very important in the options world, but you can also use it for trading spot, for example for setting stop or profit targets. Some people already do this, using ATR-based stops. You can consider ATR a poor's man GARCH.

Quote
would you care to expand more on the matter, this is really interesting stuff.
I'm not sure what to expand on. Here is a video explaining the basics:

Inserted Video


Quote
so then Heisenberg's uncertainty principle is even more relevant to the market than i thought. Maybe I am mistaken but then Range can be viewed as relative and General Direction is much more important, so the less precise the greater the chance of being right?
Maybe in a philosophical way. In physics you will never be able to defeat the uncertainty principle. But in trading, this can be just a metaphor. Because theoretically there is no reason why you can't build a computer in which to enter ALL the positions, stops and take profits of all the traders and dealers in the world, add all the news input from all the decision makers in the world and then you'll probably be able to perfectly predict the market. Also, note that high frequency traders are able to predict price direction even if they work on very short time windows, which goes against the "financial uncertainty principle".

Quote
"liqudity provider" -- too familiar with that
I think you know the idea is to take back the liquidity after a while, preferably at a better price

Quote
I'm unsure what your chart is to represent, volatility clustering and overflow imbalances I guess. To me, though, you drew a bunch of small trend angles.
The picture sort of shows average momentum, what redlion talked about. No volatility stuff in it. Note that the lines are an indicator I'm currently working on. They were not drawn manually by me, but by an analytics script.

Quote
Thing is, I keep ending up with nice lists of ways to draw, and my trading never improves due to it. Strategy/tactics on how they're used is far more important, something I'm still a rank newbie at.
Like they say in statistics, your analysis needs to be falsifiable. That's why I don't like TA, because no matter what the price does, the analysis is never wrong. Price might slice through the support line, OR it might bounce. Right. So what do I do then? Or, an option barrier will be heavily defended, OR heavily attacked. Again, no matter what happens, the analyst was never wrong.

This is why I try to backtest my ideas. That's also why I avoid anything with "might" in it. I like stuff with "will" in it. "will bounce" instead of "might bounce". If a method can't say for certain that "this will" happen, it's useless to me.

Quote
I think Adal is actually using orderflow information on those charts, those are not trendlines. Correct me if I'm wrong Adal, but are those not periods of net positive and negative flows.
Average momentum can be considered positive/negative flows. I don't say they contain orderflow information, because to me that means actually seeing the orders (ie, being a bank).
  • #97
  • Edited 4:17pm Jan 3, 2012 3:38pm | Edited 4:17pm
Quoting Scotty B


Nub,

I'm proud of ya bro. You are really excelling and making very cogent arguments. You are absolutely right. It does not matter how a person wants to trade, the most important first step is to define exactly what you are doing. It's impossible to figure out if something works (or has worked) if the rules are not absolutely clear.

As for fibs, where does a person draw them for the sake of testing? My suggestion, start simple--even by asking critical questions. Why should fibs for example (arbitrary percentages) have any special...
cheers.

even a broken clock is right twice a day, so it had to happen eventually i guess.

truth be told, i've touched on this kind of thing before with my own coding (automagically identifying s/r, plotting potential and valid trendlines from these points, and dealing with removing broken lines in an efficient manner that doesn't hang mt4 when there are more than a dozen lines to deal with).

i would be dead-set mind-blowing surprised if anyone doing a proper analysis on fibs, under any imaginable ruleset, would ever see anything resembling a bell-curve around fib points. i'd still be super surprised if there was even an identifiable blip of extra data on or near these points that wasn't just random noise.

if your data is speaking a language i'm understanding then your particular analysis on the broken high etc ruleset suggests to me that there isn't anything workable with that ruleset since it appears to nicely fade-off evenly and suggests to me that a broken high is generally nothing more than a broken high and expect price to continue ranging for the next day.

as for what crucialpoint says on the matter i don't know what to make of it. i can't imagine the future being that much different and negating your current data. for that to happen we'd have to start experiencing extended bull/bear runs that have real balls behind it to push it in one direction for extended periods of time, instead of just a bit of a push here and a bit there every few days, but meandering shit in between.

pipalicious, i hate you because i see myself in you

edit : actually, what would be interesting in the broken daily high test is the other side of the coin.... how far it comes back into the previous daily range. i imagine it probably just mirrors what was already presented.
Congratulations ScalaFX for coming out of the closet!
  • #98
  • Edited 5:58pm Jan 3, 2012 5:29pm | Edited 5:58pm
Quoting Adal
Like they say in statistics, your analysis needs to be falsifiable. That's why I don't like TA, because no matter what the price does, the analysis is never wrong.
...
This is why I try to backtest my ideas. That's also why I avoid anything with "might" in it. I like stuff with "will" in it. "will bounce" instead of "might bounce". If a method can't say for certain that "this will" happen, it's useless to me.
I remember that after a while, I was trading (demo) regularly and got into a groove understanding the chart,
so that I could pick a small chart area and add just a little technicals to it, to give reason for something to happen much, much later.
I merely had to accept that this good reason had a TP/SL that was both exact and less than perfect, to reap the benefit.

Those using prepackaged indicators can easily define for each one a certain chart variable (trend, volatility, cycle speed, range, pivot, etc.) and give TP-SL results for that one. Contrarily, one can draw on the chart from anywhere, even if there are candle- and shape- (temporary with smaller patterns fit together, and permanent/traditional) patterns to distinguish where more effective ones are, and get some kind of 'might' happen from it. With drawing, an approach doesn't exist to fitting together variables that divide the entire chart up into a jigsaw puzzle.

So usually, a pattern or three is recognized as fitting its definition better than other possibilities, and one projects a line from it.
The trouble is that one doesn't know if these recognized patterns comprise a total set of variables to judge the chart with. That's why I continually work with a list of patterns and their projected lines, to create such a set.
Rather than choose set variables, though, I'm leaning toward accepting the temporary characteristics that defining the TP-SL reason each of several devices/methods would provide. Ideally, I'd find how several devices/methods go on the chart and then how they interrelate to define the TP-SL performance of each one's projection.
  • #99
  • Jan 3, 2012 7:17pm Jan 3, 2012 7:17pm
Quoting Scotty B
Yeah, I don't think that the market will shift too much towards the right (speaking in regard to the chart I posted) with further data. We have to start with what we know about the market, thats what I'm doing. All my test shows is that the market, on breakouts , makes only between a 0 and 20 percent break of the last days range the most. This doesn't mean that the market will not continue to make spectacular breakouts, it will, but it does suggest that maybe a trailing stop would be useful.

[i][color=Navy]Instead of dwelling on theoreticals,...

Courtesy of ScottyB

Attached Image



I can not dispute your argument nor can I counter your evidence with better "statistical" evidence.

All I can contribute with our conversation is that I studied great trading failures. With so many evidence of trading failure in the past in all different ways/aspects/types of trading, there stood one cluster with all of them.
All great trading disasters had a "flawless statistical evidence" to back their system. Well it seems flawless, but it is not flawless, because it failed. What I found is that; It is not that the statistics are wrong, but a lack of understanding what they are trying to statistically represent/research and accomplish. That is where the error to be found.

If you refer back to your graph, it looks lovely. Your graph clearly shows the greater statistical divergence in breakouts. I'm pretty sure everyone here who looked at it came to somewhat a conclusion. And I'm pretty everyone arrived to somewhat similar conclusions.

I've said it before in previous post, for every bounce I can show you equally the same amount of breaks. I'm pretty sure a lot of people are fixated on the first 10 bars. And I'm pretty sure a lot of people can not critically analyze this stats.

So let me contribute my experience here and shed an insight with some of my mental diarrhea

Attached Image


See the circles in red? the <10% reaching to 90 occurrence? To a lot of people, it means a number and nothing more. Here is where Sherlock Holmes comes in That <10% represents fakeouts, consolidations, exhaustions, etcera, etcera... and a whopping 90 occurrence!!! Bar 2-10 are rendered useless if you can not overcome this hurdle. Not to mention, this statistic occurred in a financial turmoil where heavy volume of trades occurring. Handling fakeouts, consolidations, exhaustions are big enough hurdles for traders, give it a double whammy with low ranging days,... 95% of trading methods/strategies have no chance in such environments.

So it really boils into one thing... can your system detect/counteract fakeouts, consolidations, exhaustions? because if it can, you don't really need to be testing for prolonged periods. Your system will simply be telling you to stay the f*ck out
  • #100
  • Jan 3, 2012 7:44pm Jan 3, 2012 7:44pm
Quoting CrucialPoint
What I found is that; It is not that the statistics are wrong, but a lack of understanding what they are trying to statistically represent/research and accomplish. That is where the error to be found.


would you agree that this is usually because someone is trying to 'prove' something, instead of disprove something (which is what i took this case to be)?

Quote
So it really boils into one thing...
Quote
can your system detect/counteract fakeouts, consolidations, exhaustions? because if it can, you don't really need to be testing for prolonged periods. Your system will simply be telling you to stay the f*ck out
so you suggest that something like low volatility days are skewing the data where price obviously has a tendency to not go anywhere by it's very nature, and eliminating the shit days might paint a different picture? actual question since it was scotty who did the original analysis and has the data.

obviously the end-goal for everyone is to ride the trends, whether that is a 10 pip or 1000 pip trend. removing the shitty nowhere times is surely what everyone wants and can only help the bottom line.

i'm very tired, so these might not even be questions. lsdkfjlsdfja
Congratulations ScalaFX for coming out of the closet!
0 traders viewing now