Concept for an Automated Strategy Development Process

Quote from traitor786:

Dont worry I think that was my last long post on this tread.
More...

Thanks for the reply. I already crawl free data with which my databases are filled:

Code: 
Mi 13. Mär 11:31:47 CET 2013

TABLE_NAME                             table_rows  data_length  index_length  Size_in_MB
CompanyEntity                          11324       22429696     2359296       23.64
CurrencyEntity                         210         16384        16384         0.03
CurrencyExchangeRateEntity             995349      77185024     152567808     219.11
DetailedReportBalanceEntity            99790       40960000     0             39.06
DetailedReportCashFlowEntity           101183      36782080     0             35.08
DetailedReportEntity                   114204      18923520     30097408      46.75
DetailedReportIncomeEntity             128907      52527104     0             50.09
DetailedReportOptionalValueEntity      2149117     289767424    342556672     603.03
DetailedReportOptionalValueNameEntity  126         16384        16384         0.03
ImportEntity                           158         16384        0             0.02
ProxyEntity                            2252        1081344      131072        1.16
RawProxyEntity                         0           16384        16384         0.03
ReportRatingEntity                     219310      151240704    33603584      176.28
SequenceEntity                         24          1064         1024          0.00
StockEntity                            42203       9732096      6946816       15.91
StockExchangeEntity                    63          16384        16384         0.03
StockExchangeOpenEntity                192798      12075008     16302080      27.06
SummaryReportEntity                    75697       21037056     13287424      32.73
ToponymBackupEntity                    0           16384        16384         0.03
ToponymEntity                          8151885     796917760    191709184     942.83
TradingDayDividendEntity               235186      28606464     39288832      64.75
TradingDayEntity                       72674908    20045430784  9612148736    28283.67
TradingDayRatingEntity                 17646099    5224022016   2414919680    7285.06
TradingDaySplitEntity                  7173        1294336      3047424       4.14

Everything on end-of-day basis including company reports. It is about 30gb of data and updated regularly. This data is used for the stocks market stuff. On the other hand I use JForex as a backend for Forex strategies while not using the above data for that market.

I would say the approach you suppose is an entirely different strategy to trade with and it seems it is more discretionary, thus harder to automate. As far as I can guess from your post. The thesis describes a platform that makes automated strategy development easier and not discretionary trading. Though I think the thesis highlights this enough. If you have suggestions on how to make this distinction more clear, I welcome that.

-- About your comment: "From what I gather the system can be seen as a typical algo with an added input of fundamental analysis. Though you offer many ways to filter the fundamental part, it is full of holes and issues."
-> It can be seen like this, though this is just the trading system in the thesis which is explained with "holes and issues" because the referenced works explain it in more detail.

The thesis also consists of a strategy development process and a platform automating that to be able to implement various strategies. The fundamental data is only for the purpose of value investing in stocks and not forex as you describe. The examples in the thesis were chosen to be simple to serve the purpose of explaining the process and API.

Also the first halve of the thesis is the analysis part where existing knowledge is gathered and processed. The concept part describes the inventions. I guess the thesis only makes sense when you consider and know both IT (development stuff) and finance (trading, investing).
 
About TREE from SSS (Structured Software Systems):
I am using quotes from this document: http://www.strsoft.com/intro/

Another example: because a completed, executable trading system can be an element of a larger assemblage, the larger assemblage can control the execution of that trading system. In particular it can cause the trading system to be executed repeatedly over the same historical interval with different parameters. Thus, automatic parameter optimization can be done easily; so easily that it is not even a built-in system function, but something that one can assemble out of spare parts. A user-modifiable script determines which aspects of the tested system's performance are monitored, how the space of available parameter values is explored, and so forth. In most other trading software systems, parameter optimization is either not possible at all, or is a built-in function whose behavior is not open to user inspection and modification.
More...

This is the same idea I had with strategies that themselves optimize marketstrategies inside them. Thus an inner shell of backtests. Though I also offer an outer shell in my thesis to *optimize* rules before running a strategy which I call automated strategy design.

An example of a financial problem that doesn't fit the TS framework very well is searching through all available equities to select a stock portfolio meeting specified criteria. The passage of time isn't a critical aspect of this problem. Hence, while one could imagine developing a set of block types that are helpful for solving this class of problem, TS probably isn't the right communication protocol for them. A new domain specifically tailored for this class of problem would be a large project, but could be well worth the effort. (It's worth remarking that such a search domain might well be controlled by an outer TS schematic, which would re-execute the search domain subsystem at intervals reflecting how often the user wants to update the portfolio. The TS schematic would also keep track of trading profits and losses. In this way TS can be used to handle the things it's good at, without having to force all aspects of the system into the TS notation.)
More...

Portfolio selection is something I specifically make easy with my platform by not pressing the user into a graphical form of programming (with its inherent limitations). Instead the developer can define filters/signals that use any data or tools he wants to do this with (as long as the backtesting rules of never using future data is being followed). The downside is, that users will need to be programmers to use the platform.
 
Regarding BUILDING TRADING SYSTEMS USING AUTOMATIC CODE GENERATION by Michael R. Bryant, Ph.D.:

http://www.adaptrade.com/Articles/WhitePaper-GP-Content.htm

I already differentiate myself clearly in the thesis from genetic programming by comparing my platform concept with TradingSystemLab. Genetic Programming would also be usable inside my platform to generate individual rules, but I suspect generating code for complete strategies would not be a wise decision, even though possible by duplicating TradingSystemLabs functionality inside a strategy implementation in my platform. Though this would disable many of the benefits that the variability model approach yields and would degrade the platform to a mere copycat product. But who says it might not be interesting to allow this as an alternative to the variability model approach in the platform? (I don't think this is a goal that motivates me right now.)

Btw. I have added that whitepaper as a reference to my thesis. I was already familiar with genetic programming inside TradeStation to generate rules. But I forgot which book/paper that familiarity comes from... :/
 
ah yet another one i forgot all about

Visual Pattern Designer

a add on tool for Trade Station it allowed you to high-lite an area on a chart like a pattern etc. then it would automatically generate the code and a system for it based off vectors. kool as crap.

Mark
 
Quote from MarkBrown:

ah yet another one i forgot all about

Visual Pattern Designer

a add on tool for Trade Station it allowed you to high-lite and are on a chart like a pattern etc. then it would automatically generate the code and a system for it based off vectors. kool as crap.

Mark
More...

Thanks a lot! Looking at these is really fun and informative. I would love to get more recommendations (which you surely have I hope). :)

Also, If anyone is interested to see a sample report generated by my platform, see:
http://invesdwin.de/public/Report1Week.pdf
 
Regarding Visual Pattern Designer:

I am basing my impression on this old manual: http://www.scrigroup.com/limba/engleza/106/Visual-Pattern-Designer-Profes33917.php

The software must be really ancient judging from the looks. :D
I came into contact with pattern based trading about two years ago when reading some documents on forex trading.
I already knew Autocharist (http://www.autochartist.com/), but it seemed to be discretionary-only and very limited. Visual Pattern Designer on the other side looks very mighty.
I always wondered how pattern based trading could be automated and VPD now gives an answer.
Very interesting to let a neural network figure out how to recognize the patterns specified by the user. Quite the ideal use case for neural networks in fact. :D

To follow on with comparing it to my concept:
I think this fits right in to the AI-approach to generating individual rules. I even said using neural networks might be interesting for that. Visual Pattern Designer specializes itself to generate signals for entering trades. Combining this with other rules to create complete strategies that go beyond simply trading the patterns seems to be up to the developer. Also the walk-forward-analysis and automated strategy design by rule combinations is not subject of that specialization.

I will continue with looking at "Trading Blox" now.

Also I would still be very interested to get more concrete information about the "Ned Davis Technalyzer", which seems to run on cray computers.
 
I've added a reference to TREE from SSS to the thesis and state that the idea to implement the walk-forward-analysis inside AStrategy to optimize AMarketStrategy instances is similar to their idea a hierachical strategy where strategies wrap strategies to do optimization by running multiple backtests.

Also TREE has in common that rules can be reused between strategies. Thus this idea is indeed not so new in my thesis. Reusing indicators is common for platforms like JForex and Metatrader. TREE widens this to any rule by design, because they have a graphical programming language where this is required. The graphical programming in Glean from TradeLink (http://code.google.com/p/tradelink/) is a bit similar from this view. Though by doing graphical programming, it becomes hard/impossible to generate rule combinations. Using a textual programming language is required for this. I guess I could make this easier by implementing a domain specific language [DSL] that eases strategy and config class implementation. But I like the idea of simply using a language like java and adding this functionality in the form of a framework.
(And I am familiar with DSLs, because at work we do a lot of code generation via model driven architecture [MDA] based on UML with Eclipse EMF, Xpand, Xtext, ...)
 
Quote from subes:

Now you have gotten me interested even though I should go sleep. XD
What are those four theme components of which I considered 3/8? And why 3/8 when they are 4? Thanks in advance!
More...


I felt that you had begun on three of the four. Maybe just a word or two in a segment somewhere. These are small seeds. I felt that seeing the theme was half the battle so I gave you three portions (1/8th's of the whole)

For my convenience I tabbed one version of your thesis. See pages: 23, 26, 30 (3), 42, 45, 47 (2), 53, 61, and 69.

I continue to read your comments to others; whatever they say, you do explain that you have more depth than they surmise. It is also true that they are fairly estranged from the topics of your thesis.

Give a read to the ANSYS article in current Bloomberg/Business Week, pages 32 through 34.

the theme you are missing is subtly shown on the attached.
 

Attachments

About Trading Blox:

I read stuff on the website, watched videos and flew over the user guide: http://www.tradingblox.com/Manuals/TradingBlox User's Guide.pdf

This seems to have the same idea of splitting strategies into parts and developing those individually. It supports Walk-Forward-Analysis and could support automated design of rule combinations with the feature "Create new stepping parameters" (http://www.tradingblox.com/?page_id=70) offered in their highest priced edition "Trading Blox Builder".
Though their software is supposed to only optimize simple numerical parameter ranges. They also optimize one parameter after the other (as seen in their videos) without respecting interdependencies between those parameters. Thus rules with dependencies and exclusions on each other will not be possible to be used with the "Create new stepping parameters" feature to generate rule combinations. Here my concept is still unique with its application of variability models to represent these interdependencies. Also optimizing one rule parameter after the other is not possible, since these interdependencies might restrict decisions that can be made in other strategy blocks. Thus no global optimum could be achived using that approach, only a local one, which might not be an optimum at all depending on the order of rule parameter optimization.
Thus only when combining the rule parameters by multiplication of ranges instead of addition of ranges, it is possible to test a complete variability model.

Also Trading Blox is using a coarser granularity of strategy building blocks. It seems with the API of my concept a higher degree of detail is possible for reusing rules/blocks. This is because the size of the reusable rules can span among multiple strategy building block categorizations, or a tiny part of a single building block and anything in between.

Though Trading Blox has an approach to making strategy development easier by enabling graphical programming as it seems, which again limits its flexibility.

I would say Trading Blox comes the closest to what I describe in my thesis. It is also the best strategy development platform I have seen by now!

I could use some of their ideas for the testing reports to improve my reports. They have nice things there.

I also sent them my thesis to find out what they think about it. Maybe my ideas are compatible to their platform. :)
 
Quote from jack hershey:

I felt that you had begun on three of the four. Maybe just a word or two in a segment somewhere. These are small seeds. I felt that seeing the theme was half the battle so I gave you three portions (1/8th's of the whole)

For my convenience I tabbed one version of your thesis. See pages: 23, 26, 30 (3), 42, 45, 47 (2), 53, 61, and 69.

I continue to read your comments to others; whatever they say, you do explain that you have more depth than they surmise. It is also true that they are fairly estranged from the topics of your thesis.

Give a read to the ANSYS article in current Bloomberg/Business Week, pages 32 through 34.

the theme you are missing is subtly shown on the attached.
More...

I see, you mentioned Volume in a previous post if I remember correctly. In fact I kind of have not looked at using volume for indicators that much yet (even though the book Trading as a Business suggests it). Also it seems the numbers represent some strategy that operates on them...

It seems this is the article you mean: http://www.businessweek.com/articles/2013-03-07/ansys-aids-innovation-with-its-simulation-software
Is the online version complete or should I order the issue by mail?

I guess the message here is to try to be as good in model building (strategy development) as one is in creating real life things (actual trading). And sometimes old knowledge can lead to the key with which something new and successful can be created, even though the original knowledge did not purpose this.

Ok, so the pages tabbed represent the main contributions to the theme you got. :)
Though what do you mean with the numbers in braces 30 (3) and 47 (2)? Does this represent the numbering of the parts of four? Where is number one then? :D

Also thanks for the strenghening words again!
 
You must log in or register to reply here.
Back
Top