Hi,
I got an interesting idea for a quiz...
Which of the following charts is/are the real ones? Which of them is/are fake (pseudo random generated)? And why? What makes you sure about it?
*links down*
I guess it's not only me who is very curious about your (both newbie's and expert's) answers.
A little experiment...
Moderator: moderators
- blubbb
- rank: 150+ posts
- Posts: 219
- Joined: Mon Sep 29, 2008 8:07 pm
- Reputation: 0
- Location: Europe
- Gender:
A little experiment...
Last edited by blubbb on Tue Feb 17, 2009 12:45 pm, edited 1 time in total.
Please add www.kreslik.com to your ad blocker white list.
Thank you for your support.
Thank you for your support.
- blubbb
- rank: 150+ posts
- Posts: 219
- Joined: Mon Sep 29, 2008 8:07 pm
- Reputation: 0
- Location: Europe
- Gender:
Well, I'm a newbie myself. My point is, with all the experts saying that you have to develop "an eye" to see the patterns / the trend and stuff I thought- they might tell the fake/real ones apart just easily. However, I'll find it most intriguing should someone spot a trend where there is none- in a randomly generated "chart".
Of course I'll eventually post the correct answers.
Of course I'll eventually post the correct answers.
Please add www.kreslik.com to your ad blocker white list.
Thank you for your support.
Thank you for your support.
A,B,C,H looks like Forex chart. Long-tail distribution, some patterns visible.
D has a long-tail distribution, it might be real, but there is something different about it.
E might be random generated, looks like normal distribution generated, but it's not sure.
About F I'm very unsure.
G might be forex or something generated with distribution other than normal.
If I had to sort them from the most probably real to least probably real it would be:
A,B,C,H,G,D,F,E
But those are only my "feelings" about it - based on my look on those charts and my memory of past charts that I looked on.
And I'm not a real trader. I'm a programmer specializing in two areas: web application development (PHP, J2EE) and trading support systems (automated trading, risk management).
Probably there is no way to exacly tell which one is real - looking on visual charts (especially chart without open,close,high and low, candles, scale etc.).
One can generate chart that will be very similar to forex chart - basing on the similar probability distribution of returns.
It would be a little bit easier to distinct real one from random one when there would be exact data present (for example in CSV file to download).
You can look on distribution of returns, periodogram, autocorrelation and fractal dimension of timeseries and maybe tell that someone is different from few "typical" markets.
But you will never be 100% sure.
For example forex is pretty noisy - looks much more like random-walk than stocks (but with distribution other than normal - it's long tail distribution).
There can be chosen such frames of time when it is undistinguishable from random walk (if we understand "random walk" as a something that could be generated by random returns with some random distribution, not exactly Gaussian distribution).
D has a long-tail distribution, it might be real, but there is something different about it.
E might be random generated, looks like normal distribution generated, but it's not sure.
About F I'm very unsure.
G might be forex or something generated with distribution other than normal.
If I had to sort them from the most probably real to least probably real it would be:
A,B,C,H,G,D,F,E
But those are only my "feelings" about it - based on my look on those charts and my memory of past charts that I looked on.
And I'm not a real trader. I'm a programmer specializing in two areas: web application development (PHP, J2EE) and trading support systems (automated trading, risk management).
Probably there is no way to exacly tell which one is real - looking on visual charts (especially chart without open,close,high and low, candles, scale etc.).
One can generate chart that will be very similar to forex chart - basing on the similar probability distribution of returns.
It would be a little bit easier to distinct real one from random one when there would be exact data present (for example in CSV file to download).
You can look on distribution of returns, periodogram, autocorrelation and fractal dimension of timeseries and maybe tell that someone is different from few "typical" markets.
But you will never be 100% sure.
For example forex is pretty noisy - looks much more like random-walk than stocks (but with distribution other than normal - it's long tail distribution).
There can be chosen such frames of time when it is undistinguishable from random walk (if we understand "random walk" as a something that could be generated by random returns with some random distribution, not exactly Gaussian distribution).
I also have some quiz.
Just three charts, only one real.
Which one and why?
http://rath.pl/EU1.png
http://rath.pl/EU2.png
http://rath.pl/EU3.png
Just three charts, only one real.
Which one and why?
http://rath.pl/EU1.png
http://rath.pl/EU2.png
http://rath.pl/EU3.png
You are right.
First one was generated with normal distribution (but with the same standard deviation as EURUSD) and third one was generated with distribution of returns that EURUSD has (but distribution taken from longer period of time - 50k bars).
Markets have at least four things that make them differ from normal random walk:
1) Long tail distribution.
2) Fractal dimension a litter bit lower (random-walk has 1.5, markets have lower)
3) Standard deviation changes in time more than in just random manner (moving standard deviation varies more than in random-walk series)
4) There are some patterns that appear more frequently than in random generated series (for example retracements after big move)
First one was generated with normal distribution (but with the same standard deviation as EURUSD) and third one was generated with distribution of returns that EURUSD has (but distribution taken from longer period of time - 50k bars).
Markets have at least four things that make them differ from normal random walk:
1) Long tail distribution.
2) Fractal dimension a litter bit lower (random-walk has 1.5, markets have lower)
3) Standard deviation changes in time more than in just random manner (moving standard deviation varies more than in random-walk series)
4) There are some patterns that appear more frequently than in random generated series (for example retracements after big move)
Please add www.kreslik.com to your ad blocker white list.
Thank you for your support.
Thank you for your support.