Hey kerosen! You didn't post any news for a long time. Its interesting for me to get how development process going on.
Hi Graff
I didn't post any news for long time because we still work for that CS (communications server) which ioo (in our opinion) is the most important think. is like the spinal column of the system. second, as I said it before, in AI we have two major problems: gathering crystal clear data and(which mean that all the data should be collected from a few servers and syncronize them in order to avoid intrerupt internet connection and power failure) and to select the features.
i read a lot about how people try to use AI but i didn't find someone to understand that most important beyond their AI techniques is the acurracy of the data (pattern inputs) and the features(what kind of inputs e.g. indexes, technicals indicators and so on) that usualy it is used. what I can say is that the team got bigger(we are 6 for now but with so little time for this project-having in view that is not payed project), we did install some tools for the team (as wiki) and we do get more organized.
As I promise I will keep you post it
cheers
I didn't post any news for long time because we still work for that CS (communications server) which ioo (in our opinion) is the most important think. is like the spinal column of the system. second, as I said it before, in AI we have two major problems: gathering crystal clear data and(which mean that all the data should be collected from a few servers and syncronize them in order to avoid intrerupt internet connection and power failure) and to select the features.
i read a lot about how people try to use AI but i didn't find someone to understand that most important beyond their AI techniques is the acurracy of the data (pattern inputs) and the features(what kind of inputs e.g. indexes, technicals indicators and so on) that usualy it is used. what I can say is that the team got bigger(we are 6 for now but with so little time for this project-having in view that is not payed project), we did install some tools for the team (as wiki) and we do get more organized.
As I promise I will keep you post it
cheers
Understanding the PAST, we can watch the PRESENT and look into the FUTURE
usually such project must imply more than one person and one of the reason is that has a few area where you must know what how and why you do that specific thing.
for example the data must be clear, the feature must be chose it, the feature must be optimized(e.g. technical indicators, ANN hidden layers and so on), using data mining applications and techniques and so on. so you must cover a wide areas such as programming, AI, data mining, trading, finance
imho the reason why most of the people didn't got any outstanding result is because they just try to apply ANN at a very basic level(e.g. price pattern and in rare cases with some Technical Indicators-and not even those indicators to be optimized for some specific situations) and the list of reason why is not one man "show" can continue. believe me, I did read hundreds of white papers(if not over thousand of it), and nowere those scientits didn't approach the problem like beeing a multidimensional problem.
I do believe in my signature:
(Understanding the PAST, and watch the PRESENT, we can look into the FUTURE)
when you try to understand the past, you do have to get as many aspects of that past as posible in order to let data mining to extract some rules.
for example the data must be clear, the feature must be chose it, the feature must be optimized(e.g. technical indicators, ANN hidden layers and so on), using data mining applications and techniques and so on. so you must cover a wide areas such as programming, AI, data mining, trading, finance
imho the reason why most of the people didn't got any outstanding result is because they just try to apply ANN at a very basic level(e.g. price pattern and in rare cases with some Technical Indicators-and not even those indicators to be optimized for some specific situations) and the list of reason why is not one man "show" can continue. believe me, I did read hundreds of white papers(if not over thousand of it), and nowere those scientits didn't approach the problem like beeing a multidimensional problem.
I do believe in my signature:
(Understanding the PAST, and watch the PRESENT, we can look into the FUTURE)
when you try to understand the past, you do have to get as many aspects of that past as posible in order to let data mining to extract some rules.
Understanding the PAST, we can watch the PRESENT and look into the FUTURE
Building a Trading Strategy
Most of the traders, technical ones, try to build a strategy taken in consideration technical indicators and my question is: how they determine for what kind of entry point they built that strategy and how they set up the variables within indicators.
I do believe that usually they pay to little attention to the variables for whatever indicators they decide to use and more, they do not look for an entry point, they look for an indicator’s value(they don’t try to identify points into the chart and try to set up the indicator for that specific points).
Is like when you decide to go fishing:
“What I want to fish? Anchovy”. So for anchovy I do not think that would be a good idea to take a fishing rod with a 0.5 wire thick, with a whale hook, to keep up to 150kg. I think you should get a trawl net but not one-eyed with 24x24 cm, one with frequent mesh.
So my opinion is:
Try to find some points that we are interested into it, choose some indicators we want to use, and find a way to set up the indicators for that specific entry points. After we decide those points we have to index them in order to analyze(zig zag indicator). We take all the technical indicators and start to optimize all the indicators for those points. We start to optimize for buy positions, sell position or range separately. Reverse setups I am quiet sure it wouldn’t work that why we have to set it up for each state position. Out from those 30 indicators we will get just a few that match quiet well. For example if we take Stochastic indicator, we will find out that it have like 5 variables and some of them are free setups (you can insert what ever number you want %K period, %D period) and some of them, have presets (price field Low/High-Close/Close). So here we do need an Artificial Intelligence Optimizer method in order to optimize them.
of course there are indicators that they redraw but those we should not take into consideration.
“Give me an Entry point and I might Rule the world”
Most of the traders, technical ones, try to build a strategy taken in consideration technical indicators and my question is: how they determine for what kind of entry point they built that strategy and how they set up the variables within indicators.
I do believe that usually they pay to little attention to the variables for whatever indicators they decide to use and more, they do not look for an entry point, they look for an indicator’s value(they don’t try to identify points into the chart and try to set up the indicator for that specific points).
Is like when you decide to go fishing:
“What I want to fish? Anchovy”. So for anchovy I do not think that would be a good idea to take a fishing rod with a 0.5 wire thick, with a whale hook, to keep up to 150kg. I think you should get a trawl net but not one-eyed with 24x24 cm, one with frequent mesh.
So my opinion is:
Try to find some points that we are interested into it, choose some indicators we want to use, and find a way to set up the indicators for that specific entry points. After we decide those points we have to index them in order to analyze(zig zag indicator). We take all the technical indicators and start to optimize all the indicators for those points. We start to optimize for buy positions, sell position or range separately. Reverse setups I am quiet sure it wouldn’t work that why we have to set it up for each state position. Out from those 30 indicators we will get just a few that match quiet well. For example if we take Stochastic indicator, we will find out that it have like 5 variables and some of them are free setups (you can insert what ever number you want %K period, %D period) and some of them, have presets (price field Low/High-Close/Close). So here we do need an Artificial Intelligence Optimizer method in order to optimize them.
of course there are indicators that they redraw but those we should not take into consideration.
Understanding the PAST, we can watch the PRESENT and look into the FUTURE
I will disappoint you but tick data would be the greatest. is like watching over a microscope ) and nowadays having in view that CPU's are powerful and is quiet easy to do a second hand computer farm because you need power processing(you find pentium 5 on the garbage yard 
) and use grid computing would solve this issue.
I don't remember quiet well the reason why we give up to JMS(we toke that in consideration).
not because i consider I am right in posts here but if you take each post of mine from here you may get a general idea which can help you to your next system building.
thx for appreciation
cheers
) and nowadays having in view that CPU's are powerful and is quiet easy to do a second hand computer farm because you need power processing(you find pentium 5 on the garbage yard ) and use grid computing would solve this issue.I don't remember quiet well the reason why we give up to JMS(we toke that in consideration).
not because i consider I am right in posts here but if you take each post of mine from here you may get a general idea which can help you to your next system building.
thx for appreciation
cheers
Understanding the PAST, we can watch the PRESENT and look into the FUTURE