You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Chart Pattern recognition formulas
- Thread starter Bouds
- Start date
Quote from Bouds:
Hi
I'm looking for formulas (metstock, tradestation or any other similar format) for recognising chart patterns such as head and shoulders, wedges, triangles, flags etc.
Any formulas or help should be greatly appreciated.
Thanks
Bouds
MetaStock has some John Murphy plug-ins which will automatically identify several for you.
I stopped with Metastock V5 and it is years since I didn't use it anymore so I won't help you with custom formulas. But here's a site with some patterns detection example:
http://www.paritech.com/education/technical/custom/indicators/default.asp
http://www.paritech.com/education/technical/custom/indicators/default.asp
Quote from Bouds:
Hi
I'm looking for formulas (metstock, tradestation or any other similar format) for recognising chart patterns such as head and shoulders, wedges, triangles, flags etc.
Any formulas or help should be greatly appreciated.
Thanks
Bouds
You may see almost any pattern at
http://www.amibroker.com/library/formula.php?id=105
and
http://www.amibroker.com/library/detail.php?id=166
See also the most recent application at
http://groups.yahoo.com/group/amibroker/message/34847
They work just fine for any OHLC data.
http://www.amibroker.com/library/formula.php?id=105
and
http://www.amibroker.com/library/detail.php?id=166
See also the most recent application at
http://groups.yahoo.com/group/amibroker/message/34847
They work just fine for any OHLC data.
Attachments
The P,V relation can be used to define all formations no matter what the pace or fractal on which you select to trade.
Sooner or later you will want to combine all features of patterns using this.
What you find out that is improtant is that as pattrens form they eliminate the possibility of other patterns coming up instead.
Since patterns also have signals associated with them, the signals are really what you are looking for.
If you are able to get to the place where patterns work for you, then you will begin to see their corresponding signals.
The citations for the formulae are a cogent statement of that fact that most programmers are not traders who make any money on the markets.
One thing to consider is how all these patterns and their signals can be melded into a dynamic "filter" as it is termed in ET. It's not really a filtering process actually; it is, instead, a focussing process.
To have it operate (the focussing process) continually you need to consider entry and exits. What comes up is a dynamic sequencing element that continually eliminates possibilities for either entry or exit whichever is the contemporary need.
The eliminating thing is a tree thing, i.e., a series of branch points and some feedback loops (these emphasize the nature of the volatility of the price over time).
The pattern equations will have several common elements one to anothr. This is the overt clue that a tree is the optimum mechanical devise.
Making up a system of pattrn equations will begin to point out to you what I mean.
The alternative to patterns (which fall into trends) is to build a trend set of functions.
End effects of trends (both beginning and ending) is where it is at.
You can google the seven I made for TC2000 ver 3.0. One of them removes large block trading which disguises all equations from being useful.
If you want another degree of fineness you can add A/D to the P and V equations variables. Do not conider any equations that do not use volume; these are only a wrm up drill for making money if the only include price as a variable. A stands for accumulation and D stands for distribution.
You will find the periodicity of each of the three variables is a doubling of the slower. This makes it obligatory to consider the variables in a given order of importance. P then V then A/D.
Once you get to a place where you can transform discontinuous raw data to a continuously behaving analogue function then you an move to getting leading signals from the maths you generate.
You will see a "calculus" like processing of statistical data is called for.
Once you make some progress you will find yourself in a place where you have results that are "too good to be true" by the standards of conventional practitioners.
Were I you I would skip what you are doing and move into some serious stuff for making money. formations were invented before the computer and copiers and the like.
Focus on end effects of trends and how the dynamic probabilites of possibilites come into play. It makes investing and trading behave like slow motion.
Sooner or later you will want to combine all features of patterns using this.
What you find out that is improtant is that as pattrens form they eliminate the possibility of other patterns coming up instead.
Since patterns also have signals associated with them, the signals are really what you are looking for.
If you are able to get to the place where patterns work for you, then you will begin to see their corresponding signals.
The citations for the formulae are a cogent statement of that fact that most programmers are not traders who make any money on the markets.
One thing to consider is how all these patterns and their signals can be melded into a dynamic "filter" as it is termed in ET. It's not really a filtering process actually; it is, instead, a focussing process.
To have it operate (the focussing process) continually you need to consider entry and exits. What comes up is a dynamic sequencing element that continually eliminates possibilities for either entry or exit whichever is the contemporary need.
The eliminating thing is a tree thing, i.e., a series of branch points and some feedback loops (these emphasize the nature of the volatility of the price over time).
The pattern equations will have several common elements one to anothr. This is the overt clue that a tree is the optimum mechanical devise.
Making up a system of pattrn equations will begin to point out to you what I mean.
The alternative to patterns (which fall into trends) is to build a trend set of functions.
End effects of trends (both beginning and ending) is where it is at.
You can google the seven I made for TC2000 ver 3.0. One of them removes large block trading which disguises all equations from being useful.
If you want another degree of fineness you can add A/D to the P and V equations variables. Do not conider any equations that do not use volume; these are only a wrm up drill for making money if the only include price as a variable. A stands for accumulation and D stands for distribution.
You will find the periodicity of each of the three variables is a doubling of the slower. This makes it obligatory to consider the variables in a given order of importance. P then V then A/D.
Once you get to a place where you can transform discontinuous raw data to a continuously behaving analogue function then you an move to getting leading signals from the maths you generate.
You will see a "calculus" like processing of statistical data is called for.
Once you make some progress you will find yourself in a place where you have results that are "too good to be true" by the standards of conventional practitioners.
Were I you I would skip what you are doing and move into some serious stuff for making money. formations were invented before the computer and copiers and the like.
Focus on end effects of trends and how the dynamic probabilites of possibilites come into play. It makes investing and trading behave like slow motion.
Jack, if I understood what you were talking about I would agree 100%, for sure. Just kidding.
I mentioned in another post that trends are defined by their ends and got trashed, sort of. It's just a y=mx + b representation, a simplification for sure, but a useful one. I did focus on the end points of trends and got some satisfying results at one point.
NeoTicker has a feature where you can capture part of a chart and tech analysis overlays and use that as backtesting criteria. Never tried it myself but on the surface it seems like a real easy way to get underway with testing pattern ideas.
Max
I mentioned in another post that trends are defined by their ends and got trashed, sort of. It's just a y=mx + b representation, a simplification for sure, but a useful one. I did focus on the end points of trends and got some satisfying results at one point.
NeoTicker has a feature where you can capture part of a chart and tech analysis overlays and use that as backtesting criteria. Never tried it myself but on the surface it seems like a real easy way to get underway with testing pattern ideas.
Max
Quote from maxpi:
Jack, if I understood what you were talking about I would agree 100%, for sure. Just kidding.
I mentioned in another post that trends are defined by their ends and got trashed, sort of. It's just a y=mx + b representation, a simplification for sure, but a useful one. I did focus on the end points of trends and got some satisfying results at one point.
NeoTicker has a feature where you can capture part of a chart and tech analysis overlays and use that as backtesting criteria. Never tried it myself but on the surface it seems like a real easy way to get underway with testing pattern ideas.
Max
Thanks for the reference. I agree with you that end effects is a very fertile area.
To get o the opportunity you have to recognize that ends do not occur on EOD data. I tune through seven fractals to find the optimum one for a given market pace.
Then I pair the fractal with the next fastest one for using the maths for anticipation.
Between these two fractals that slip and slide with market pace changes, you arrive in end effects land.
it is neat to think about when the back testing industry is going to get a wakeup call that says what they are doing isn't quite useful at all.
Some of those backtesters actually change their trading algorithms to improve their backtesting results. strange to say but the backtesting approaches drive the situation and that prevents iterative refinement because they get into curve fitting.
Capturing a part of a chart and then fixing the trading algorithm sounds terrific to me. Back testing after the fix is then a possibility.
That was funny when you remarked about not getting very far with the general linear equation.
Trying to get a response from the thread originator on the ROI ball park was about the same.
Obviously he is dealing with a lousy set of stocks to begin with and his backtest period is pure humor as it goes through the markets that occurred.
Can you imagine what would happen if he worked into delineations like market sectors??? I use excel to optimize my universe by emp[loying sector analysis. It is really neat to see how money is made as the sectors rotate and cycle.
Here is a neat representation of sector rotation over the business cycle.
http://stockcharts.com/charts/performance/SPSectors.html
Max
http://stockcharts.com/charts/performance/SPSectors.html
Max