Do you see patterns in Random Walks?

[ssrrkk]

Quote from HATEtheRisk:


And the key to this is like always STATISTICS.

You mentioned statistics a number of times, but I believe what's more important than statistics is statistical significance. I think what cooltrader is saying is that you can't really prove statistical significance of those patterns, even though the statistics appear to indicate a tradable pattern. For example, if you looked through the last 3 years of data, and you find a specific pattern that occurred 50 times, and 40 out of those 50 times, the pattern was followed by a significant up-trend (at least 10 points). Would you say that's a 80% chance?

A statistician would not say that outright. He would try to construct a statistical test and to find the significance level of that occurrence, i.e., what is the probability that such an 80% statistic would be observed by chance. Unfortunately due to the fractal nature of price trajectories, most patterns can and do happen by chance, even if they appear to indicate an 80% probability of preceding a significant move.

In other words, your back test probability might say 80% but when you forward test it, the statistics might quickly go down to 50/50.

 

[Random.Capital]

Quote from HATEtheRisk:

You know, i am the true Master of patterns in the markets.

Who are you trying to convince - us, or yourself?

 

[HATEtheRisk]

Quote from Random.Capital:

Who are you trying to convince - us, or yourself?

Nobody.

I just told you how it is.

Shouldnt every technical trader thinks this way ?

But i know its true, i am damned good in my business, its all mine, i created it, its mine, its mine - hehehehehehe

 

[HATEtheRisk]

Quote from ssrrkk:

You mentioned statistics a number of times, but I believe what's more important than statistics is statistical significance. I think what cooltrader is saying is that you can't really prove statistical significance of those patterns, even though the statistics appear to indicate a tradable pattern. For example, if you looked through the last 3 years of data, and you find a specific pattern that occurred 50 times, and 40 out of those 50 times, the pattern was followed by a significant up-trend (at least 10 points). Would you say that's a 80% chance?

A statistician would not say that outright. He would try to construct a statistical test and to find the significance level of that occurrence, i.e., what is the probability that such an 80% statistic would be observed by chance. Unfortunately due to the fractal nature of price trajectories, most patterns can and do happen by chance, even if they appear to indicate an 80% probability of preceding a significant move.

In other words, your back test probability might say 80% but when you forward test it, the statistics might quickly go down to 50/50.

I understand your point.

But i must tell you, it is not true for my trading patterns.

My statstics tell me i have a 90 percentage of success, if i follow my rules to 100 percentage. If i make a mistake the percentage goes down.

Its not just so, that there is a pattern on one timeframe and i would just bet on that - on No lord that would be too easy.

I told you, its about pattern in pattern in pattern, in timeframes in timeframes in timeframes. And i told you its very complictaded and complex, like a matrix, or boxes stuffed in boxes in boxes.....

Then to figure out, how and when exactly this first timeframe pattern, will work out or not, is the other part, where you must look on the other timeframes and patters, and so on.

About all this i have made and always make statistics, so that i know, when a pattern will work out or not.

My view of the markets is just too complex for the most people.

Well, i have learned it this way and it works out very well.

And i say it again, Statistic makes the money, nothing beats statistics, if they are perfect.......

If something have worked the last 100 times, i would bet a lot of my money that it will also work a 101 time, that is the easy description of what i do.

I just believe in the constant repetation of this patterns and i have never been disapointed. I also know, when the pattern will not work out perfect, what very rare can happen, thats the 10% of uncertainity, but i wont lose much or anything then.

This all goes into statistics to predict the perfect odds.

This is the job of an odds manager while gambling with risk.

Thank you

and

Good trading.

 

[CoolTraderDude]

Quote from eusdaiki:

On fractals and scaling in finance page 39 Mandelbrot suggests that price changes over fixed clock intervals are more chaotic than price changes from successive transactions or based on volume variations...

So the key to your riddle may be to move away from time based charts and towards tick or volume based charts...

Mandelbrot is a pretty smart guy. Fractal Geometry is all about finding order in chaos... The patterns are in fact ordered, to a certain extent... The question here isn't really to find the order. I can find the "W" double bottom on a chart within a bigger chart... The problem here is that (in my opinion) I don't think that there is any correlation to future patterns from the current pattern.

Quote from HATEtheRisk:


You are right, patter in pattern in pattern, the more the better.

And yes its the same product at the same time, but its not the same chart, every timeframe is a own chart, thats why you have them.

The difficult thing is to know, when all important timeframe patterns move into each other and break through the blockage of energy cycles restistance, after the law of the lowest resistance, then price moves, in other words, you can clearyl analysis when the big players decide when to buy or sell, exactly the timepoint, because nothing in the market happens without a reason.

The big players know everything and on the charts are in reality all informations written in, you need to know, to predict the future price move odds.

And the key to this is like always STATISTICS.

You can test and prove every idea with statistics and see if it have any value or not.

About your problem for patterns in patterns, you must know this is pretty difficult and also for me sometimes i am wrong with the exactly correct timepoint right before prices move.

But i found many tools to determinate it so exact as possible.

I give you one clue about this problem, there is always one bigger timeframe what rules the patterns on the lower timeframes, as long you dont have this bigger picture right, price will not move, cuz how i said, they are always pattern in pattern in pattern.

And before price moves, there is always a lot blockage of energy what must be tested and retested in a trading range of high and lows, before price breaks out.

But the markets are truly rigged and move in regular orders.

But how i said its pretty difficult to figure all this stuff out and i can understand that you are confused about this.
So, the best for you might be your hot potato trading style.

In the end the game is only about making money and doing a successfull business.

How you do it, doesnt matter in the end.

Good luck

If you can make money by it then more power to you!

Quote from ssrrkk:

You mentioned statistics a number of times, but I believe what's more important than statistics is statistical significance. I think what cooltrader is saying is that you can't really prove statistical significance of those patterns, even though the statistics appear to indicate a tradable pattern. For example, if you looked through the last 3 years of data, and you find a specific pattern that occurred 50 times, and 40 out of those 50 times, the pattern was followed by a significant up-trend (at least 10 points). Would you say that's a 80% chance?

A statistician would not say that outright. He would try to construct a statistical test and to find the significance level of that occurrence, i.e., what is the probability that such an 80% statistic would be observed by chance. Unfortunately due to the fractal nature of price trajectories, most patterns can and do happen by chance, even if they appear to indicate an 80% probability of preceding a significant move.

In other words, your back test probability might say 80% but when you forward test it, the statistics might quickly go down to 50/50.

You're right. All patterns appear in hindsight... It's really hard to use in real time because your "pattern recognition" has to speculate on what the pattern may become not on what the pattern is... It's Hindsight bias, hence its lack of predictive value.

 

[The_Tourist]

Quote from ssrrkk:
You mentioned statistics a number of times, but I believe what's more important than statistics is statistical significance. I think what cooltrader is saying is that you can't really prove statistical significance of those patterns, even though the statistics appear to indicate a tradable pattern. For example, if you looked through the last 3 years of data, and you find a specific pattern that occurred 50 times, and 40 out of those 50 times, the pattern was followed by a significant up-trend (at least 10 points). Would you say that's a 80% chance?
A statistician would not say that outright. He would try to construct a statistical test and to find the significance level of that occurrence, i.e., what is the probability that such an 80% statistic would be observed by chance. Unfortunately due to the fractal nature of price trajectories, most patterns can and do happen by chance, even if they appear to indicate an 80% probability of preceding a significant move.
In other words, your back test probability might say 80% but when you forward test it, the statistics might quickly go down to 50/50.

Interesting point, is probability really the same as statistics?
Probability is static but statistics could be dynamically revealing a bias/physical defect in a dice. What analysis can be done to tell if shorter term statistics is going to revert to the probability mean or is there really a bias?

 

[HATEtheRisk]

Quote from The_Tourist:

Interesting point, is probability really the same as statistics?
Probability is static but statistics could be dynamically revealing a bias/physical defect in a dice. What analysis can be done to tell if shorter term statistics is going to revert to the probability mean or is there really a bias?

LOL

WTF. Of course they are not the same.

But only with statistics you can determine probabilities.

There are 2 kind of statistics: 1 = history statistic backtesting.
2 = Real time Trades statistic.

The gameplan is to make no. 2 to no. 1. This is the objective of the game. When they have become one, they are the perfect statistics what give you then the best possible probabilites and the rules for that kind of setups and strategies.

With them, you can predict all kind of market moves, short term and long term.

 

[phattails]

Quote from CoolTraderDude:

http://en.wikipedia.org/wiki/Pattern

Random Walk



http://en.wikipedia.org/wiki/Random_walk

So if it's a real "Random Walk" then you can't have patterns... Because patterns have to be repetitive... On the other hand, if there are repetitive patterns then it isn't a real Random Walk.

My answer is to combine the two... I'll say that you do have "patterns" in the market but that their intervals and intensities are random which makes them relatively useless for predictions. In a sense, Fractal Geometry applies really well to the market...

While I was using technical analysis to trade my "pattern recognition" would kick in and then... The pattern I thought would emerge turned into something else. I haven't had good experiences with patterns in the markets. It's really hard to assign significance to every formation. But... If you've found a way to do that then you should be able to make a lot of money off of it. [/B]

If you scroll down that wikipedia page you'll see that the "purest" version of a random walk assumes indendent and identically distributed R.V.'s at each step BUT, there are nontrivial variants which are very relevant to this discussion: multivariate models (think cross-covariance), non-gaussian dist., the order book, predictable higher orders, any exogenous variables, etc.

Some of these include "patterns", but all of them provide potential for profit.

BTW, could you please site proof that markets are fractal (if that's what you're implying). I've started to associate the word fractal with religious life insurance salesmen that sit down next to me on an airplane.

 

[ssrrkk]

Quote from The_Tourist:

Interesting point, is probability really the same as statistics?
Probability is static but statistics could be dynamically revealing a bias/physical defect in a dice. What analysis can be done to tell if shorter term statistics is going to revert to the probability mean or is there really a bias?

Statistics are used to estimate population parameters from samples. Probabilities are assigned for inference, either for predicting future values or assigning confidence / significance for parameter estimates.

Bayesian inference is one way to equate past statistics with future probabilities.

In the context of statistical tests, I am referring to the case where probabilities are used to assign confidence or significance to the parameter estimates. So if your past statistics indicate that 80% of the time, a certain pattern resulted in a profitable upswing, then I would like to know what is the 95% confidence interval on that 80% statistic. If it is 80% +/- 40%, then I am less confident about betting on it. Alternatively I could come up with a hypothesis test and ask: is this 80% (profitable) significantly different from 50% (not profitable, money-losing with commission+slippage). The null hypothesis says it is indistinguishable from the 50% case. If the p-value from this test is 0.3, then that means that 30% of the time, you could have gotten the 80% by chance, i.e., that it is just a fluke, there is no mechanism behind it, or it is a curve-fit.

 

[CoolTraderDude]

Quote from phattails:


BTW, could you please site proof that markets are fractal (if that's what you're implying). I've started to associate the word fractal with religious life insurance salesmen that sit down next to me on an airplane.

Cite proof...? Check out the links bellow...

http://classes.yale.edu/fractals/randfrac/welcome.html

http://classes.yale.edu/fractals/randfrac/Market/Market.html

Video...

<iframe frameborder="0" width="480" height="360" src="http://www.dailymotion.com/embed/video/x6xqd2"></iframe><br /><a href="http://www.dailymotion.com/video/x6xqd2_fractals-and-the-stock-market_news" target="_blank">Fractals and the Stock Market</a> <i>by <a href="http://www.dailymotion.com/socionomics"

A fractal has been defined as "a rough or fragmented geometric shape that can be split into parts, each of which is (at least approximately) a reduced-size copy of the whole,"

http://en.wikipedia.org/wiki/Fractal