| Author |
Message |
     stoian
Member
Username: stoian
Post Number: 14
Registered: 03-2004Rating: 
Votes: 4
|
| | Monday, May 30, 2005 - 05:06 pm: |  
|
Hi forumers,
Charts are doing a good work, but not a very good one. Why?
Because when You are picking up values for Your charts, You are picking up (only some time with a greater, and every time with a smaller treshold/limit of certitude) A NONSIGNIFICANT VOLUME of for Your research needed values. If You want to be able to predict/foresight/forecast the future evolution of the values of the phaenomena You need to pick up A SIGNIFICANT VOLUME for Your research.
Do You do this? Are You every time consciously that EVERY KIND OF CHART MEANS ALSO A PROBABILITY FIELD,since when K.F.Gauss, Laplace, R.A Fisher have defined to be every chart, and NOT only as a commonplace/trite/simple grafical representation?
All the scientist of the world together with them of the Massachussets Institute of Technology are preserving this matter as a secret=freemasonry.
When charts are treated from You, only in the simple way i have mentioned and not AS A PROBABILITY FIELD, than charts are functioning literally like an human eye that can not see very, very well what happend in the past, and what is happening yet=curently, and therefore have only a very, very little chance to see any TRUTH about the future evolution.
Kind regards,
Your Doru Stoian
|
hilarius
Member
Username: hilarius
Post Number: 764
Registered: 04-2004Rating: N/A
Votes: 0
|
| | Monday, May 30, 2005 - 05:48 pm: |
|
Stoian
Due you use statistics predictively?
Hilarius
I come in peace to share my thoughts and to shine my candle light on possible long term opportunities
|
redbellie
Member
Username: redbellie
Post Number: 4
Registered: 04-2005Rating: N/A
Votes: 0
|
| | Monday, May 30, 2005 - 05:44 pm: |
|
Doru,
Could you please explain this again for me and the laymen among us please?
Cheers RB
|
david_louisson
Member
Username: david_louisson
Post Number: 60
Registered: 02-2004Rating: N/A
Votes: 0
|
| | Monday, May 30, 2005 - 08:54 pm: |
|
Hi Stoian
I am trying to understand your post. When you talk about volume, I assume that you are referring to the fact that you need to run tests over a large sample of trades, in order to attain statistical significance. Correct? If so, I certainly agree.
I don't understand the reference to freemasonry, or exactly what is being kept secret. Can you explain further?
In your other post
http://forum.incrediblecharts.com/messages/191443/188197.html#POST66734
you talk about SMAs (statistical/mathematical analyze) and probability fields. Do these refer to an approach like Steidlmeyer's Market Profile, i.e. creating a probability distribution of recent price behavior, in order to forecast likely future direction, or are you referring to a statistical analysis of your actual trading test results (i.e. number of wins, losses, expectancy, etc)?
You also say that "very accurate forecasts" are necessary, in order to make profits. Can you explain in more detail how you obtain these?
Many thanks
David
|
stoian
Member
Username: stoian
Post Number: 18
Registered: 03-2004Rating: N/A
Votes: 0
|
| | Monday, May 30, 2005 - 09:30 pm: |
|
To Redbellie and David-Louisson especially, but also for all forumers,
Yes, i am researching in a LIKELIHOOD MANNER an not creating creating a probability distribution of a price evolution. You can called it also, a research of the investor`s=speculator`s behavior regarding the very different price levels, in order to forecast likely=the most probable future direction, and not only, so as i said already many times before in my postings, i estimate in a likelihood manner the MOST PROBABLE LEVEL of the risk of increase/decrease(+/-)=volatility of any evolution for titles from the Stock Markets, from the Foregin Exchange Market, or from any other markets.
Kind Regards,
Doru Stoian
|
stoian
Member
Username: stoian
Post Number: 19
Registered: 03-2004Rating: N/A
Votes: 0
|
| | Monday, May 30, 2005 - 09:53 pm: |
|
Also to redbellie and david_louisson,
I am never referring to a statistical analysis of any actual trading test results (i.e. number of wins, losses, expectancy, etc), because ONLY FROM THIS SINGLE REASON, RESEARCHERS ARE SAYING THAT "HISTORY=WINS MADE IN/FROM THE PAST CANNOT JUSTIFY=SET UP A CONFIDENT BASE FOR FUTURE WINS".
But a am strongly sustain that idea:" that only SIGNIFICANT VOLUME OF HISTORY VALUES is making the future values", or that "the future was build almost completely in the past, as ideas linked to other ideas mentioned also today, about the FEATURE, THAT PRICE EVOLUTIONS ARE NOT EVERY CASES A DUMB STRING OF VALUES, BUT INTELLIGENT=ORDERED STATISTICAL SERIES OF VALUES, which are posessing AN OWN MEMORY AND CAN BE ABLE TO TRANSFER AN INHERITANCE TO THE FUTURE EVOLUTION (they don`t forget=never forget the levels reached in the past when they curentlly describe an evolution). That also refferes too and describes in my words THE THEORY OF EFFICIENT MARKETS, where FUTURE INFLUENCES THE PAST AND VICE VERSA.
Regards,
Doru Stoian
|
david_louisson
Member
Username: david_louisson
Post Number: 61
Registered: 02-2004Rating: 
Votes: 2
|
| | Tuesday, May 31, 2005 - 06:01 am: |
|
To Mr Stoian: Thank you for your reply. I have just located another thread where you provide examples (XLS spreadsheets analyzing EUR/USD) of what I assume is your SMA method of analysis. This has given me a much improved idea as to how this works. It has certainly answered some of my earlier questions.
To those wishing to understand this further: load Mr Stoian's examples from this thread:
http://forum.incrediblecharts.com/messages/23/210494.html
As best as I can understand, this is what is being attempted: a best-fit regression line, using either a polynomial (as used in the EUR/USD example) or exponential formula, is being drawn through the price data points. The assumption that is being made is that the polynomial function can be extrapolated to predict future price direction.
There is an excellent piece of software called 'Curve Expert' (which can be downloaded from http://curveexpert.webhop.biz/) that allows over 30 different curve fitting models (polynomial, exponential, power law, yield density, growth, sigmoidal, hyperbolic, sinusoidal, gaussian, and numerous others, including user-defined) to be applied to any series of X-Y data points. Curve Expert allows data to be imported directly from Excel. By using an iterative process, it will then fit the best curve (i.e. lowest standard error) to any model that you specify, and even compare fits between all of the different models.
Once the formula has been established, it is then possible to extrapolate the function to obtain future values of Y (price) given any value of X (time).
In the early stages of my study of TA, I trialled this product in an attempt to see how well these models were able to predict future price action. What I did was take a historical section of prices, apply the fitting, extrapolate the curve, and compare the result with what actually occurred. I soon gave up, as I found that each model gave a completely different result, in the context of the required level of accuracy (i.e. to produce workable buy/sell signals). It also raised the critical question: how to determine which model was most applicable for any given stock.
I also discovered that, when attempting to fit a polynomial, it was obvious that its ability to forecast the 'medium to distant' future was severely limited, simply because of the very nature of the function itself. An 'n'th degree polynomial always has n-1 turning points. After the final turning point is passed, it proceeds indefinitely in one direction to either positive infinity or negative infinity. Obviously price will never do this :-)
Also, many of the models, if extrapolated forward far enough, gave negative values (i.e. dipped below the x-axis). All of the above gave me the impression that such a study would be useful, if at all, only for very short term forecasting.
I could have (and possibly should have) researched this more extensively, but my knowledge of statistics (and therefore the ability to judge the usefulness of statistical curve fitting) is not sufficiently good. However, my intuition told me that I had seen enough to suggest that this was not a particularly viable option.
A couple of other reasons:
1. The markets are driven largely by emotion, and are buffeted severely by news (esp in the short term), and it is impossible to measure emotion statistically. There are also severe effects caused by 'sector rotation' as applied 'heavyweight' market funds, effects on price of ex-dividend announcements, sudden changes in economic outlook, and (potentially) a number of other events whose effect, I would suggest, defies any kind of precise mathematical analysis.
2. If least-squares-regression-curve-fitting gave a significantly better forecasting result than 'conventional' TA techniques, then I'm sure that it would be gospel by now, and would have made a significant impact on all of the Trading Systems and literature that is available to us today. In other words, the 'Market Wizards' of today would be using it, and advocating it, over conventional TA. Plenty of brilliant minds have already had many years worth of opportunity to evaluate it.
Like Elliott Wave theory, this kind of statistical analysis is attractive in that it doesn't only purport to forecast price direction, but also future turning points, i.e. how far the price is likely to move.
Another interesting type of statistical analysis method is 'Market Profile', developed by Peter Steidlmayer. I'm afraid that I never evaluated it fully. Anybody interested might like to 'do a Google' and study it further.
Mr Stoian, I do not wish to denigrate your work in any way. Perhaps I have misunderstood what you are saying. If that is so, I apologize.
Finally, I fail to see how a reference to Freemasonry or the Illuminati should add evidence as to why a 'statistical' method should warrant consideration over any other method. Again, the 'secular' academic world has had ample opportunity to evaluate such a method, and would surely be applying and recommending it, if it really was the optimum approach. There is really not that much mystery about it (assuming that one's knowledge of statistics is a lot better than mine :-) However, again I apologize if I have missed the point.
David
(Message edited by david_louisson on May 31, 2005)
|
hilarius
Member
Username: hilarius
Post Number: 765
Registered: 04-2004Rating: N/A
Votes: 0
|
| | Tuesday, May 31, 2005 - 10:09 am: |
|
David
I am one who is fascinated (as you appear to be) by the use of statistics to make predictions ... if such a thing is possible
I have a couple of questions :-
(1)
I routinely use moving averages as a means of eliminating noise from data series to focus on the "underlying trend"
Using Guppy Multiple Moving Averages one frequently discovers instances where the short and long term moving averages are in conformity ... and other instances where they are in conflict
Contraction of these averages and movement after crossovers appear to offer interesting predictive opportunities
Have you used GMMA's and what is your view of them?
(2)
A distinguished member of this forum has frequently referred to the rubber band effect ... that is the tendency of the actual price to overshoot or undershoot and in due course respect a selected (preferred) moving average before embarking on another overshooting or undershooting move in the same direction
As I understand it he recommends use of this phenomenon as a tool for timing entries and exits in company with other selection criteria
A problem with this is that some crossovers occur without denoting the end of the major trend while other crossovers are seen with hindsight to be genuine precursors of a major trend change
So my question is how is an insignificant crossover distinguished from a significant one ... presumably the answer lies in the subsequent price action, but what are the subsequent signals we can use to find that the crossover failed to be predictive, and at what specific point should the signals be acted upon?
Finally can we make use of a combination of TA and FA in order to determine overbought and undersold situations where market perception has departed radically from underlying value?
For example at $ 2.50 the Australian stock One Tel was miles away from its underlying true value, as was HIH for many months ... so is there a statistical way to measure predictively the difference between current perception and underlying reality ... and is such use of statistics limited to TA or FA alone, or can both be combined?
If we could agree on the answers to these questions we could write a book together and make our fortune 
With Best Wishes
Hilarius
I come in peace to share my thoughts and to shine my candle light on possible long term opportunities
|
stoian
Member
Username: stoian
Post Number: 22
Registered: 03-2004Rating:
Votes: 1
|
| | Tuesday, May 31, 2005 - 05:23 pm: |
|
To david_louisson and to hilarious,
I am very sure You do not wish to denigrate my work in any way. It is also very sure You have misunderstood what i am saying. But despite this situation You have not to apologize. I explain myself in two steps. The first step will be partly with your words/ideas, and the second step with a question to all of You from me:
1.Only the curve-best-fitting it`s truly and only giving a significantly result but only a partly result, THAT WILL PROVIDE HELP FOR THE NEXT STEP FOR THE FORECASTING METHOD=technique, and is not giving results for a better forecasting itself.
Another interests for the statistical analyse, is to determine the MOST EXACT PROFILE=BEHAVIOUR IN TIME OF THE MARKET OR MARKET PARTICIPANTS (investors and speculators).
If the market has a profile of "NORMALY-RISKED MARKET" where the risk=volatility=variability for every short interval of time is NOT GREATER than +/-15% (it has every short time a risk between 0% and maximum +/-15%), or if we have to do WITH AN "OUT OF NORMAL-RISKED MARKET" where the same indicator called risk=volatility=variability for short intervals of time is GREATER THEN +/-15%.
2.I am repeating for You, THAT EVERY CHART MUST BE TREATED AS A PROBABILITY FIELD (where each daily event in part, HAD NOT ONLY FOR THE PAST BUT WILL ALSO HAVE FOR THE FUTURE, a certain level of probability FOR HIS APPEARANCE=IN PRODUCING ITSELF) and when you are treating a chart without this FEATURE, than YOU WILL HAVE AN IMPERFECT EYE=INSTRUMENT TO SEE AND TO UNDERSTAND REASON OR EMOTIONS WHICH ARE CONDUCTING TO PRODUCE ALL THE MARKET MOVEMENTS ON HER DIFFERENT PRICE-LEVELS, and which is not allowing you also TO FORESEE THE NEXT EVOLUTION FOR THE PRICE-LEVELS.
Where, to whom and why have You leaved this feature?
Kind Regards,
Doru Stoian
|
stoian
Member
Username: stoian
Post Number: 23
Registered: 03-2004Rating:
Votes: 1
|
| | Tuesday, May 31, 2005 - 06:05 pm: |
|
To david-louisson,
All the questions i have posted for You are related eachother. So You have to pay great attention and discover all the links between this ideas.
1) Why do You have nothing mentioned about R.A Fisher with his two extremly important types of differences, called also dipersion (to be calculated through the Least-Squares- Method):
a) The dispersion that exists between each INDIVIDUAL=LOCAL EMPIRYCAL AND HISTORYCAL VALUE and EACH THEORETICAL MEAN VALUE, OF THE STRING OF MEAN VALUES, CALCULATED AS A STRING OF AVERAGE=MEAN VALUES (string of values, which are given through the BEST-FITTING-REGRESSION-CURVE which is calculated on the basis of the already mentioned individual=local empirical values);
b) The dispersion that exists between each of the THEORETICAL MEAN VALUES FROM THE THEORETICAL STRING=SERIES OF VALUES and THE STATISTICALLY (in a likelihood manner) ESTIMATED REAL MEAN.
2) Why do You don`t mentioned nothing about THE GALTON`S OGIVE AND HER REASON OF EXISTENCE?
Do You see that on the vertical axis are represented the probability levels and on the orizontal axis are represented the levels of the GAUSSIAN COEFFICIENT OF GUARANTEE for each level of probablity in part?
Do You understand the meaning of this instrument which the Galton`s Ogive is?
Do You understand that the Galton`s Ogive is showing us the GENERAL LAW OF GAUSS, ABOUT THE GENERAL MODEL OF THE REPARTITION-DISTRIBUTION IN PROBABILITY FOR THE APPEARANCE-PRODUCING OF THE EVENTS OF EACH PHAENOMENA FROM THE UNIVERSE?
Do You understand that LAPLACE HAS DISCOVERED - ON THE BASE OF THE GAUSS`S GENERAL LAW OF REPARTITION - WHAT ARE NORMAL-RISKED PHAENOMENS (like the human behaviour on the financial markets is) AND HOW THEY ARE LOOKING FROM THE SIGHT/ANGLE/ OF THE PROBABILITY LEVELS IN COMPARISON WITH THE GENERAL LAW OF REPARTITION IN PROBABLITY OF GAUSS?
|
cjb
Member
Username: cjb
Post Number: 58
Registered: 02-2003Rating: N/A
Votes: 0
|
| | Tuesday, May 31, 2005 - 10:18 pm: |
|
Hey Sto,
After having a read am I correct in saying you are calculating Pivot points / support-resistance levels? Or have I missed it??
CB
|
ricardon
Member
Username: ricardon
Post Number: 20
Registered: 09-2002Rating:
Votes: 1
|
| | Wednesday, June 01, 2005 - 12:22 am: |
|
stoin, afraid your post make no sense to me at all. Perhaps I need a dinglehopper to comb out the tangles or a babelfish. Perhaps if all the values in the probalistic eigenmatrix were rotated around the number 42 I would see the connection to the gravimetric index.
|
|
