Concept for an Automated Strategy Development Process

[subes]

Quote from Nab:

What you are describing in your "thesis" is basically a primitive form of evolutionary/genetic programming/strategy design.

Indeed you could look at it like this. TradingSystemLab is then a sophisticated form of evolutionary/genetic programming/strategy design, which I compare in my thesis.
That platform gives the computer the whole responsibility about developing strategies.

On the other hand there is the classical way of developing strategies where the developer does all the work of experimenting and development of strategies.

Since I have not seen it yet, my idea was to define something in between of these two in order to have the benefits of both and maybe be able to reduce the drawbacks of both (which can only be validated after the platform is finished and experience in using it is gained). Thus my thesis is not about some new mathematical analysis, some fancy new strategy or a new technology, but in fact about a new approach to strategy development.

I think something new can be defined by looking at existing things and doing something new or unfamiliar with them.

If you want we can discuss this further, this is indeed an interesting way to look at this. If you disagree I would like to hear your reasons.

PS: Genetic programming was only the best guess on how to reduce the tests needed to find successfull strategy combinations (Rober Pardo also names in his book a few more ideas about how this can be achieved). In fact other approaches to this could work aswell. The interesting part is the gained insight the developer gets after analyzing the results to decide which strategy building blocks are the best ones for him. The developer decides in the end by changing the variability model and choosing which variants he wants to incorporate into the strategy as decision constants.

PPS: Think about a fund/trader/investor that works using a specific trading strategy and they want that particular strategy to be automated. Using TradingSystemLab with the genetic programming would not be able to use the rules that the fund/trader/investor wants, since the genetic programming effectively defines its own rules. Should the fund/trader/investor then go to a classical strategy developer? I think the cost for that classical strategy developer would be higher than hiring someone using the approach from the thesis, since there is a difference in efficiency during development. Though again this is just my hypothesis that can not yet be validated.

 

[Traveler]

You're going to have trouble backtesting on fundamental data unless it's the expensive "Point in Time" version that shows the info that was publicly available at the time. The cheap databases are full of numbers that got revised later and weren't currently available at the time.

 

[subes]

Quote from Traveler:

You're going to have trouble backtesting on fundamental data unless it's the expensive "Point in Time" version that shows the info that was publicly available at the time. The cheap databases are full of numbers that got revised later and weren't currently available at the time.

Yes, you are right. I stated this as a problem in chapter 2.4.4 and accepted it as a tradeoff. I did not know the correct data was called "Point in Time" version. Thank you for pointing that out, I added a footnote that states this now.
Do you know more specifics on where to get them and how much they cost to license?
I only see this on a quick search: http://thomsonreuters.com/products_...ent_overview/content_az/content_fundamentals/
Though reuters does not list prices and in the past did not tell me any prices for a different product when I asked them. :/
Maybe my approach in asking them is wrong...

 

[jack hershey]

Quote from subes:

Hi all,

this is my idea, which I have written my master thesis about (which is pretty much print-ready, but not final). I am currently in the process of implementing it and plan to use it primarily on the forex market, sometime later maybe also for stocks. The thesis talks about value investing because this was agreed upon with my teachers. Let me know what you think. This was started more than three years ago when my interest in automated trading began. I could think about working on this in a team, though not open sourcing it.

See: http://invesdwin.de/public/Thesis.pdf

Feedback might help me to further improve the text. I am open for discussion.

Best regards,
Edwin

Hi Edwin:

I read the first few pages and then switched to the references.

To make money in markets, you need to use current information. This means the fundamentals of stocks do put you a little behind the times. Quarterly reports which meet gornerment regulations are 1/4 year lagging.

Most people recognize this and are able to add what is needed to get current in portfolio management.

As you have read, the portfolio manangement of the Financial industry does not really work very well. Often they do get advice from academia. The record shows if and when this helps.

Academia is supported in its research by the government and that is almost humorous. You could skim through the grants of NSF in the US to get a reckoning on this.

For investing to grow capital a good standard is the "250 list". It is done by many reporters of the current scene. formally or informally. The list is a list based on projections at least three years out. The 250 stands for the annual return precentage expected during those three years. Most lists have at least 15 stocks.

Another schema is to use Sector Rotation. Within each sectro there are leading and lagging stocks. By knowing their relative offset within a sector, you can reduce the lagging inherent in your work so far. Enlarging from that, you can use the actual sector rotation to manage the portfolio to pick off when to go to cash and where to put new found cash into stocks that are going to have increasingly better and better fundamentals. In you work you did not deal with the dynamics of markets. Sectors can define 200 groups immediately and their relative performance immediately.

Lastly, there is a means to anticipate, the stellar performers of the future. This is done by grouping all stocks into categories by quality of performance. Your references mostly miss dealing with quality and its change over time. Look up EPS and RS for starters. You need to learn how filters work.

If you want to learn about investing and how capital growth occurs, that would be a good idea. First, examine the standards used by people who know how to make money.

The financial industry wnet though a switch from qualifying to quantifying. That is to say finacial businesses went from serving the customer to betting on how to make more fees and commissions.

This transition was really amusing It was so bad that GS used a stock formula for betting on their pairs type bond trading. When they got to losing 100 million a year, someone looked at the formula and saw it was misapplied.

wer I you I would become familiar with what makes things work in finance: criteria design, filter design, formulae rules to define and how operators are used to drefine things.

If you want to see a recent funny one look at SPS as some kind of miscalculation.

 

[subes]

Quote from Nab:

What you are describing in your "thesis" is basically a primitive form of evolutionary/genetic programming/strategy design.

I have added a new chapter "5.1 Categorization" and enhanced chapter "5.2 Benefits" to describe the differentiation among the classical, manual strategy development process, the automated strategy development process of trading system lab using genetic programming and the new approach of this thesis between these two extremes.

I also made a reference to the software development theory behind this which is called "software product line". This is in fact something that gave me the idea to this concept to some degree as I now realize afterwards.

@Nab: Thank you for giving me the idea to do this.

 

[subes]

Quote from jack hershey:

Hi Edwin:

I read the first few pages and then switched to the references.

To make money in markets, you need to use current information. This means the fundamentals of stocks do put you a little behind the times. Quarterly reports which meet gornerment regulations are 1/4 year lagging.

Most people recognize this and are able to add what is needed to get current in portfolio management.

As you have read, the portfolio manangement of the Financial industry does not really work very well. Often they do get advice from academia. The record shows if and when this helps.

Academia is supported in its research by the government and that is almost humorous. You could skim through the grants of NSF in the US to get a reckoning on this.

For investing to grow capital a good standard is the "250 list". It is done by many reporters of the current scene. formally or informally. The list is a list based on projections at least three years out. The 250 stands for the annual return precentage expected during those three years. Most lists have at least 15 stocks.

Another schema is to use Sector Rotation. Within each sectro there are leading and lagging stocks. By knowing their relative offset within a sector, you can reduce the lagging inherent in your work so far. Enlarging from that, you can use the actual sector rotation to manage the portfolio to pick off when to go to cash and where to put new found cash into stocks that are going to have increasingly better and better fundamentals. In you work you did not deal with the dynamics of markets. Sectors can define 200 groups immediately and their relative performance immediately.

Lastly, there is a means to anticipate, the stellar performers of the future. This is done by grouping all stocks into categories by quality of performance. Your references mostly miss dealing with quality and its change over time. Look up EPS and RS for starters. You need to learn how filters work.

If you want to learn about investing and how capital growth occurs, that would be a good idea. First, examine the standards used by people who know how to make money.

The financial industry wnet though a switch from qualifying to quantifying. That is to say finacial businesses went from serving the customer to betting on how to make more fees and commissions.

This transition was really amusing It was so bad that GS used a stock formula for betting on their pairs type bond trading. When they got to losing 100 million a year, someone looked at the formula and saw it was misapplied.

wer I you I would become familiar with what makes things work in finance: criteria design, filter design, formulae rules to define and how operators are used to drefine things.

If you want to see a recent funny one look at SPS as some kind of miscalculation.

Thank you for your aswer. Though by not reading the thesis you could not recognize that it actually does not focus on one specific trading style only, but also looks at strategies from a more abstract perspective in order to dissect them and define a new approach to developing them. Also I think benjaming graham, warren buffett and phil town who were/are using value investing also know how to make money on the markets. Though value investing was only used as an example to give the thesis more context to work with. I could also have chosen some gappers, scalper, <your suggestions>, ... strategy. Though I chose value investing because this is what I am most experienced with and thought it would be interesting to also analyze if an automation of that strategy is possible.

 

[Nab]

Quote from subes:


I think something new can be defined by looking at existing things and doing something new or unfamiliar with them.

You see, one thing that many in academic finance fail to release is that there are thousands of masters/PhDs/Profs in physics/mathematics/computer science employed outside academia to do nothing else the full day than finding edges in the market. This is not a recent phenomenon, but started as early as the 80-90s. Besides that they nowadays just outnumber the academic researchers, they are usually also intellectually playing in a different league. Hence, there has been a vast amount of "hidden" research being done over a long period of time. It is in the nature of the "field" that you don't spread the word of what you know, as it will ultimately kill your edge. Even to know what does not work is quite valuable. The literature you find is mostly either outdated our useless beyond being a general introduction into the thematic. The ones who write books are usually the ones who have failed and that's their last way to gain some $$$ and attention. Similar considerations apply to many of the "technology" providers. Don't get me wrong, there is some great technology available for $$$, however, there is also a lot of crap ... mostly targeted at the retail level traders.

 

[subes]

Quote from Nab:

You see, one thing that many in academic finance fail to release is that there are thousands of masters/PhDs/Profs in physics/mathematics/computer science employed outside academia to do nothing else the full day than finding edges in the market. This is not a recent phenomenon, but started as early as the 80-90s. Besides that they nowadays just outnumber the academic researchers, they are usually also intellectually playing in a different league. Hence, there has been a vast amount of "hidden" research being done over a long period of time. It is in the nature of the "field" that you don't spread the word of what you know, as it will ultimately kill your edge. Even to know what does not work is quite valuable. The literature you find is mostly either outdated our useless beyond being a general introduction into the thematic. The ones who write books are usually the ones who have failed and that's their last way to gain some $$$ and attention. Similar considerations apply to many of the "technology" providers. Don't get me wrong, there is some great technology available for $$$, however, there is also a lot of crap ... mostly targeted at the retail level traders.

You have quite a negative view on this. If I thought everything that I can think of has already been tested and exploited by somebody else somewhere hidden, I would not be motivated to try any of my own ideas. And what IF my ideas were in fact innovative and not yet tried? I would have thrown it away and not tried it because I thought too negative about this...

Unless I find someone who can tell me and prove that he already tried this and it failed and clearly shows me that I will fail too and won't even learn a lesson from it. I will consider abandoning something I worked many years for... Though I think even then I would rather not, but instead try to find a way to make it still work or adapt it somehow so that it works.

Even if somebody did this and it worked, I think I contribute by making this knowledge academically available (which I myself worked hard for to gain) and maybe I can get the same benefit from it as the person who already benefited from it before me and kept quiet. Anyway, what the thesis describes is a platform for developers to develop strategies on. The goal for a platform is to have developers to develop on it. If nobody knows about the platform, where should the developers come from? Or where should I get reviews on the idea to improve it and make the best out of it? I think keeping quiet about an idea is a way to failure of that idea, as Eben Pagan explained it nicely in his audio program "Self made wealth".

 

[Nab]

Quote from subes:

You have quite a negative view on this. If I thought everything that I can think of has already been tested and exploited by somebody else somewhere hidden, I would not be motivated to try any of my own ideas. And what IF my ideas were in fact innovative and not yet tried? I would have thrown it away and not tried it because I thought too negative about this...

You got me a bit wrong. Of course everyone in this field has to start somewhere in finding his alpha. However, one should not be as presumptuous to think that the idea one is working on is new, in particular if it is on such a basic level. This does not mean that basic ideas can not work, in contrary, in particular on longer time-frames there are a lot of simple things that do work to some extend, however, usually not well enough to impress the big fishes.

You should also keep in mind that in contrast to academia, this field is mainly not about helping others or being of use to others. Rather, it is about to have and keep a competitive advantage, since you have to pay bills at the end of the day. Your advantage is the intellectual property and technology your prey does not possess.

I have the impression that the long term effect of the nowadays more and more for free (free in the sense of the amount of own effort and talent needed to put into) publicly available information and technology will be that it will distort the food-chain in a way that there will be only a few very big predators left, and as well a huge amount of sheeps playing with themselves a null-sum game and feeding the left-over predators.

 

[jack hershey]

Quote from subes:

Thank you for your aswer. Though by not reading the thesis you could not recognize that it actually does not focus on one specific trading style only, but also looks at strategies from a more abstract perspective in order to dissect them and define a new approach to developing them. Also I think benjaming graham, warren buffett and phil town who were/are using value investing also know how to make money on the markets. Though value investing was only used as an example to give the thesis more context to work with. I could also have chosen some gappers, scalper, <your suggestions>, ... strategy. Though I chose value investing because this is what I am most experienced with and thought it would be interesting to also analyze if an automation of that strategy is possible.

Thank you for your reply.

Do you have a set of illustrations that parallel the text? It the reader supposed to fill in the gaps in the text with assumptions?

Do you have a device that tells you when the automation system has hit a stonewall and Why?

TIA.