Menu
Sister Sites

Options

Bookmark Thread

First Page First Unread Last Page Last Post

Print Thread

Similar Threads

If you could go back - one Forex thing you'd tell youself? 64 replies

Trade with Naked Eyes 1,757 replies

Looking on a naked chart, can you pick the best strategy? 32 replies

Can you tell how the fundamentals are from the monthly chart? 3 replies

Trading as naked as you can get 0 replies

Attachments: Can you tell with naked eyes?
Exit Attachments

Can you tell with naked eyes?

Kk007,

In the book the black swan by nassem taleb, u can learn about two types of variable,

Mediocristan - outliers will not break the distribution n may be ignored
extremistan - outliers may render distribution meaningless, outliers value hs extreme impact

its a good book.
Quoting luqmanz
that sounds logical to me, but to be practical in trading, i assume all setup will fail so im always prepared for failure. This is all trader need to worry about.
The ability to distinguish what "appear to be" real (in fact random) and "actually" real is something imporant, I think.
Quoting kk007
If you frame a movement with a guassian/normal distribution, it is not completely random. Say, adults' height in a population is normally distribution, can you say that the height of an arbitrary adult in the population is completely random?
Attempting to predict the height of a single person selected at random from a population is not possible with any degree of accuracy. Population heights are actually quite interesting because they do fallow a normal curve. Male heights in the U.S. have a mean of 5' 9.3" (69.3inches) and female heights have a mean of 5' 4" (64 inches). Both have a standard deviation of 2.8 inches on either side. This means that if the statistics hold, 97.5% of men 74.9 inches tall or shorter. If someone offered me a bet where I win $1 for every male height drawn out of a bag that was less than 74.9 inches and I lost $1 for every male height drawn out of a bag that was more than 74.9 inches, I would take that bet as often as possible. It is true that I could end up losing 10-20-30 in a row because of an unlucky streak - the draws are indeed random. However over time if we did thousands of draws from the bag, I would end up with all my friend's money. Thus the individual draws are random, but the distribution of heights follows a gaussian curve.


Quote
In the MIT paper, subjects learned the real charts in the training phase and they were told which is real and which is random. In the testing phase, they are given the same real charts and newly generated random charts again, and this time they have some ability to distinguish between the two types. It could be because they remember the exact shape of the price curve in the real charts.

Are you sure you read the paper? Perhaps you overlooked the following from section 2 Experiment Design: http://web.mit.edu/alo/www/Papers/arorassrn.pdf

From the design it doesn't appear that memory can have much if any of an impact as half the data was new to each participant, and the other half (the daily data) was data shifted to hide its identity. From my own experience it was easy to quickly identify the components that make up a real price series from the fake. Are you unable to do the same?

To evaluate the robustness of our experimental design, we varied various parameters of
the experiment across data sets, as indicated in the Results section below. In addition,
we presented subjects with data charts in two different ways. For half of the data sets
corresponding to transaction-by-transaction (or “tick”) data, each subject was shown a fresh
set of charts, based on a sequence of returns disjoint from the sequences shown to other
subjects. For the other half of the data, corresponding to daily data, the charts shown to
each subject were based on the same sequence of returns.6

6However, the data was shifted by a random amount for security reasons, i.e., to avoid the possibility that two subjects could coordinate their guesses, for example by simultaneously playing the same charts on
two nearby machines.


Quote
You have no argument to justify this weakness and just ask me to re-examined my belief. Mine is not a belief. Mine is a valid argument. And you? you have no argument.
Again, I don't need to come up with an argument because the methodology looks sound and my own experience backs it up.

Quote
I did not read the book you mention, I don't know whether it is a good book or not yet. But it does not mean that you are knowledgable and other people are ignorance.

I will quote from wikipedia - by the way, no personal insult was intended:

Quote
Ignorance is where someone or something is uninformed. This should not be confused with being unintelligent, as one's level of intelligence and level of education or general awareness are not the same. The word "Ignorant" is an adjective describing a person in the state of being unaware. The term may be used specifically (e.g. "One can be an expert in math, and totally ignorant of history.") or generally (e.g. "an ignorant person.") -- although the second use is used less as a descriptive and more as an imprecise personal insult.

Quote
P.S. Also you made up the intention of FQ as if he had told you why he posted.
I'm sure FQ will speak for himself if he so chooses.
kk007, in the study, those who had previous financial markets experience fared only fractionally better than those who had no previous financial market experience. It was in the range of 1-2% higher.

Also, no, there's no way they could 'memorize' the data sets. It goes over this specifically in the study.

Quoting luqmanz
Btw to answer ur question. someOne did an experiment to see if trader can distinguish randomly generated chart and real chart.
Result: they ccouldnt tell the difference.

This result tells nothing bout market tradability. Dont take it too seriously
I'm not sure what study you are referring to, but there is a study linked and discussed in this thread in which people were able to distinguish between random and real charts given that they have experience seeing how the real chart moves.

Traders can distinguish between a chart of a market they know and a random chart. Traders can NOT distinguish between a real but unknown market and a random market at first glance. Why? Because different markets exhibit different, but repeated, characteristics (which would manifest in price patterns).

Quoting kk007
The ability to distinguish what "appear to be" real (in fact random) and "actually" real is something imporant, I think.
If you are a price pattern trader, I agree. And as the MIT study shows, given enough experience watching a market, one CAN distinguish between said market and a random price chart. Posting random but real price charts vs. just random price charts is utterly meaningless. There's no way to differentiate which is real without knowing what market it is. I also think the idea of just posting one picture and comparing it to another with no context is pretty stupid.

Now, want to do try something relevant? Howabout instead of posting two random charts, one real and one fake, why not post two real charts of two different markets and see if people can distinguish between the two? Post a EUR/USD and gold chart, but don't label either, and then see if people can distinguish the differences. That's a more fruitful endeavor imo.
Quoting kk007
Since I cannot find some good advice which can help me to enhance my discrimination ability in identifying real price movement from random price movement, I will just learn it by practise. My plan is I will keep reading the randomly generated charts and real price charts, make note on their possible difference, until I get some conclusions.
Your whole premise if faulty.

Market generated price charts are not random. Each price represent a value proposition for both the buyer and the seller.

Put another way; everything every trader knows about the market is reflected in the price. Everything the market knows about itself is reflected in its price.

So here's the rub:

Even though you "can't" immediately ascertain a "random" chart from a market data derived chart, the underlying elements that make up the individual price points are what makes one tradable (the real one) and one not (the computerize one). True price patterns do have elements of market psychology imbedded within them. It stands to reason that they might fail on random charts by the vary nature of the fact that no real human element has produced them.

You can't "prove" a naked chart (no indicators) can't be traded by taking out the very thing that makes them tradable.
Wyckoff VSA: (1) Supply vs Demand (2) Effort vs Result (3) Cause vs Effect
OP,

I see you missed the newbies post.

I gave you an answer that would convince you or anyone else
(and for those that mentioned mandelbrot, you should agree).

The simple way to discern real price from random (in the gauss
sense) is convert the price series to changes, you will visually
be able to see the difference! I guarantee it.

Will that help you any? How so.
Quoting Craig
Wrong, consider a stationary Gaussian Random walk, it's generated via random numbers (or as random as a computer can), when it gets more than 2 standard deviations from the mean, what are the chances of the next step being closer to the mean? (hint: it isn't 50%).
I agree with Craig. It is possible to predict the random outcome with certain accuracy if you can figure out the probability distribution of the randomness.

So ONLY if it is random and you dont know the probability distribution, then it is not predictable.
Quoting kk007
Hi CP,

It was not my intention to test people's ability in the first 2 posts on this thread. I posted 10 randomized charts in the first two posts to ask people if they can tell they are randomized. You made me know that even a guy with 6.5 years + high IQ cannot distinguish.

cheers,
It was just a bit of fun for me to test myself... but hey, I got 50% of it right. 50% correct with money management is good enough for me

My take on the issue:
Fake or real/random or non-random... if one cannot exploit real charts, then what are the odds/chances of one's ability to exploit fake charts... even if one has the ability to distinguish fake from real. It doesn't do you any good.

It's like an appendix... doesn't do you any good, but you're in big trouble once it gets infected and then explodes.

I admire that you are thinking outside of the box. I hope you are find what you are looking for. I love to research, but for me... (in my humble opinion) there are no solid argument to pursue such research.
Quoting FXEZ
kk007 - you owe it to yourself to read up on the subject. After reading your last statement it's clear that even you don't understand what you wrote. Earlier in the thread, one of the posters was having a bit of fun at your expense - ForexQuant. http://www.forexfactory.com/showpost...1&postcount=16

Sorry FQ to spoil your little game but the OP owes it to raise himself/herself from ignorance by reading Mandelbrot's book, "the Misbehavior of Markets". In that book you'll find on page 17 the pictures that FQ posted, and a...
Well i have no intention to make fun of anybody.

To me his original question is more like find out the difference in between random and non random chart but I just suggested to rephrase the question to "spot the real and fake chart".

For those who had read mandelbrot's book should know that the randomness is never a main issue of spotting the real and fake chart, what does matter is the statistical behaviour, ie price distribution. Personally I do not agree to focus on the random or not random issue because i do not think it will bring any benefit or calculated edge even if you can measure the randomness correctly!!! IMHO the price distribution is more important and can bring us the calculated edge!!!

One of the main reason that why most mathematical market models failed is not they cant figure out the randomness but they ASSUMED market is normally distributed.

Therefore discussion on randomness is futile, IMHO.
Quoting Craig
Wrong, consider a stationary Gaussian Random walk, it's generated via random numbers (or as random as a computer can), when it gets more than 2 standard deviations from the mean, what are the chances of the next step being closer to the mean? (hint: it isn't 50%).


Hi Craig,

I couldn't stop laughing when I read this. (not mocking you, but agreeing with you)

I'm laughing because its what brings in the money for me. I say, "you either get it or you don't."

"when it gets more than 2 standard deviations from the mean, what are the chances of the next step being closer to the mean? (hint: it isn't 50%)."

OMG! This is my bread and butter. This is what I exploit again and again... and again.
[quote=markmm;3637227]It is tradeable, but not using patterns, at least not the traditional patterns.[/quote
i agree
if you can trade randomness you can trade any instrument. its easy, find the trend and trade with the trend. at the same time make sure your winners are larger than your losers.
Judging from various threads and posts from OP, i guess he wanted to know why the price pattern exist even in a random generated chart while his original intention was to prove that real market price is not random.

Personally I think that the existence of repeated price pattern (ie triangle, double top, support & Resistance) does not prove that the market is random or not random. Mathematically, the reason that price pattern is kept repeating because the market is fractal and it is a non-deterministic chaotic process. The fractal nature of market will occasionally turn into some sort of patterns in the eye of human. However it is common that different people may have different interpretation of pattern. It is just like different people who looking at the same cloud, some may see a bear but some may see a rabbit.

However when most people were educated to recognize the price pattern then a miracle happened. The self-fulling prophecy may cause the randomness reduced significantly because everyone does the same action(Everyone makes money in strong bull market). When the balance of supply & demand breakdown, we will see the price 'trend' or move to the right direction as the pattern suggested, therefore we 'think' price pattern can give us hints on future direction. However that particular pattern may or may not give us edge because we dont know the probability of success and its risk reward ratio until we measure it. I would say it is the human behaviour that create an edge but not the pattern itself.

Some may argue that you can spot the traditional chart patterns in random generated chart (No human behaviour involved) and the price moves to the direction as suggested. therefore the market is random, or not random depending on the position of your viewpoint. No offence but i would say this is a hintsight because you are not looking at the right edge of the chart. Yes I do agree that the random generated charts do have pattern but can you tell the expectancy for each pattern? How you know that the breakout of triangle pattern will provide you a mathematical edge? Have you ever measured it or you just made your conclusion based on a few observations?

Damn i think i dont know what am i talking now, I guess i lost somewhere else. I better stop comment after my last comment.

Focus on randomness is futile and it does not bring any benefit to trading. A focus on price distribution could be a better way to find the calculated edge.

As usual I could be deathly wrong!

P/S: Damn it i forgot to put my entry order because i didnt look at my chart while writting this!!!
Quoting CrucialPoint

[/b]OMG! This is my bread and butter. This is what I exploit again and again... and again.
[/color]
This is my bread & butter too! The probability theory is the foundation of my quantitative trading model.

P/S: The price surged to half of my 50pips TP while risking 5pips only. no mood to trade now
So it is a study contradict to MIT paper's finding. So, depending on how researchers set their experimental design, they get different findings.

Quoting luqmanz
Btw to answer ur question. someOne did an experiment to see if trader can distinguish randomly generated chart and real chart.
Result: they ccouldnt tell the difference.

This result tells nothing bout market tradability. Dont take it too seriously
Quoting kk007
So it is a study contradict to MIT paper's finding. So, depending on how researchers set their experimental design, they get different findings.
The research that I mentioned was a very old one mentioned in one of the classic trading book I read long time ago. Perhaps the MIT study is more up-to-date.
Quoting luqmanz
The research that I mentioned was a very old one mentioned in one of the classic trading book I read long time ago. Perhaps the MIT study is more up-to-date.
That MIT study is a crap, the smoothed charts in "mandrill", "bulls", "bears" and other are the real, in "reindeer" is the opposite and in other is the chart with the fast price movement... I can't distinguish if the charts posted here are fake or real and I'm first in one contest and second in other two, you can see it here: http://arora.ccs.neu.edu/v4/tool/menu.jsp judge for yourself.
So, you agree one cannot call a value taken from a guassian disturbution completely random?

Quoting FXEZ
Attempting to predict the height of a single person selected at random from a population is not possible with any degree of accuracy. Population heights are actually quite interesting because they do fallow a normal curve. Male heights in the U.S. have a mean of 5' 9.3" (69.3inches) and female heights have a mean of 5' 4" (64 inches). Both have a standard deviation of 2.8 inches on either side. This means that if the statistics hold, 97.5% of men 74.9 inches tall or shorter. If someone offered me a bet where I win $1 for every male height drawn...
Isn't this obvious? If I had not read the paper, how could I critize it? The point is not whether I am skilful enough to get high score in the experiment test. The point is the crucial weakness in the experimental design. If the real charts given in the training phase is the same as the real charts given in the testing phase. Subject learned the "instant" of real charts rather than the abstract rules. Experimenters did not address how to prevent instant learning.

Quoting FXEZ
Are you sure you read the paper? Perhaps you overlooked the following from section 2 Experiment Design: http://web.mit.edu/alo/www/Papers/arorassrn.pdf

From the design it doesn't appear that memory can have much if any of an impact as half the data was new to each participant, and the other half (the daily data) was data shifted to hide its identity. From my own experience it was easy to quickly identify the components that make up a real price series from the fake. Are you unable to do the same?

[b]To evaluate the robustness of...
  • Post #119
  • Quote
  • Edited 11:34pm Apr 15, 2010 10:37pm | Edited 11:34pm
I don't have a direct practical application initially. It is my belief that if I dig an issue interesting to me deep enough, some benefits will eventually come. A direct practical application is not what I am looking for to begin with this thread. Nevertheless, this one below I just come up with may be a practical application?

False Positive (Type I problem): My trading system trigger a trade signal based on a historical price movement, but the price does not move as predicted.

False Negative (Type II problem): My trading system does not trigger a trade signal based on a historical price movement, but the forward price movement can be predicted.

False Negative is less a problem than False Positive. If my system produce a false negative, the consquence is missing an opportunity. If my system produce a false positive, I expose to risk unnecessary.

OK. If using randomized charts as inputs to my system, and my system keep giving trade signals, my system is a false alarm system. Right? For a perfect trading system, the alarm should not be on using randomized charts as input. Right?

In other word, these randomized charts can act as a "touchstone" for the validity of my system right?

Quoting HiddenGap
Your whole premise if faulty.

Market generated price charts are not random. Each price represent a value proposition for both the buyer and the seller.

Put another way; everything every trader knows about the market is reflected in the price. Everything the market knows about itself is reflected in its price.

So here's the rub:

Even though you "can't" immediately ascertain a "random" chart from a market data derived chart, the underlying elements that make up the individual price points are what makes one tradable (the real one) and...
  • Post #120
  • Quote
  • Edited 11:46pm Apr 15, 2010 10:46pm | Edited 11:46pm
Quoting kk007
So, you agree one cannot call a value taken from a guassian disturbution completely random?
A sample taken from a gaussian distribution would be called gaussian random. A sample taken from a completely distribution would be called completely random.


Quote
Isn't this obvious? If I had not read the paper, how could I critize it? The point is not whether I am skilful enough to get high score in the experiment test. The point is the crucial weakness in the experimental design. If the real charts given in the training phase is the same as the real charts given in the testing phase. Subject learned the "instant" of real charts rather than the abstract rules. Experimenters did not address how to prevent instant learning.

I realize that there may be a language barrier in play but read the quote I posted again from the paper - fresh means new, as in data that the subject hasn't seen before. Personally I'm not employed by MIT to handle such complaints about their scholarly papers, and if I were I would claim my salary as being far too low for the aggrivation. If you have a problem with the research perhaps you should write a rebuttal. I'll just leave it at this here and in the thread.

As this thread was ostensibly about you learning to differentiate between random and non-random charts, which may be reflected in how well you did in the experiment, perhaps some who decide to stick it out to the bitter end on this topic will offer some pointers or you'll be able to apply the many pointers that have already been offered. Best of luck!
0 traders viewing now