Australian (ASX) Stock Market Forum

Amibroker as a fundamental analysis development tool - any developers?

Jump to new
O

ottg

Joined
31 January 2015
Posts
43
Reactions
0
The more I'm searching for a stock screener (for specific data manipulative requirements) the more it looks if it doesn't exist in the public domain. The quickest way to get something off the ground is to use AmiBroker which I read is a professional, powerful, comprehensive trading system development platform.

a. While it is mainly used for technical analysis can anyone indicate why it wont be suitable for fundamental analysis?
b. Any development or scripts need to be tested vigorously. It also appears from readings if Amibroker's cutting edge charting and graphics, and fast, flexible and powerful portfolio-level backtesting, optimization, and automated walk forward validation will be more than enough.
c. Do you know of any existing groups or forums that attempted this successfully and that we can perhaps join?
d. Will it be a good idea to use a development domain like SourceForge or Github to host artifacts, databasis, code and script management and central communications with feedback to this forum?
e. Anyone with the time, skills, tenacity and energy that will be interested in driving such efforts?
f. Who will be interested in active participation?

The objective is to develop a system based on a scientific approach, that will help long term investors identify profitable opportunities to buy and sell.
 
Hi ottg --

No other comments? I cannot resist. I have written extensively about the lack of value of fundamental data for trading. The issues revolve around granularity, timeliness, revision, agenda of reporter, etc.

I believe that the markets are strongly efficient with regard to fundamental data. That no person outside the board room of the reporting agency can profit from fundamental data.

That said, no one trading system development platform gives any advantage over any other platform with regard to use of fundamental data.

Best regards,
Howard
 
howardbandy said:
I have written extensively about the lack of value of fundamental data for trading. The issues revolve around granularity, timeliness, revision, agenda of reporter, etc.

I believe that the markets are strongly efficient with regard to fundamental data. That no person outside the board room of the reporting agency can profit from fundamental data.
Click to expand...

...Perhaps I should ask what you mean by "trading" (your word) before I comment. Could you elaborate on what you mean by trading? Perhaps you have a more nuanced meaning than what I understand the word to mean.
 
ottg said:
The more I'm searching for a stock screener (for specific data manipulative requirements) the more it looks if it doesn't exist in the public domain. The quickest way to get something off the ground is to use AmiBroker which I read is a professional, powerful, comprehensive trading system development platform.

a. While it is mainly used for technical analysis can anyone indicate why it wont be suitable for fundamental analysis?
b. Any development or scripts need to be tested vigorously. It also appears from readings if Amibroker's cutting edge charting and graphics, and fast, flexible and powerful portfolio-level backtesting, optimization, and automated walk forward validation will be more than enough.
c. Do you know of any existing groups or forums that attempted this successfully and that we can perhaps join?
d. Will it be a good idea to use a development domain like SourceForge or Github to host artifacts, databasis, code and script management and central communications with feedback to this forum?
e. Anyone with the time, skills, tenacity and energy that will be interested in driving such efforts?
f. Who will be interested in active participation?

The objective is to develop a system based on a scientific approach, that will help long term investors identify profitable opportunities to buy and sell.
Click to expand...

How much are you willing to pay for a fundamental-based system like that?

Will cost at least $1 million to hire developers, BA, DBA.

And at the end of it, don't think you could do back testing fundamental data like you could with charting or something similar.

A ratio or figure alone won't give you context... and while it's not impossible, might not be easy to code a smart enough system to take into account all the variables that may indicate a good opportunity when a few figures alone would indicate a sell.
 
Hi Systematic --

My categorization is expenditures are one of three categories --

Investing. When I take a position, I call my attorney and have my will modified. An island I intent to leave to my heirs. I have not yet called my attorney.

Trading. I expect to reverse my position sometime before I die, hopefully at a profit. All stocks, futures, mutual funds, houses.

Expenses. Things that will probably be worth less, even worthless, when I dispose of them. Food, clothes, cars.

Best,
Howard
 
Hey Howard,

Thanks for your reply, and clarification.

howardbandy said:
Trading. I expect to reverse my position sometime before I die, hopefully at a profit.
Click to expand...

...We're on the same page.


Can I also clarify (sorry, I forgot to add) whether you count 'price adjusted' or 'price relative' fundamental data (a common example would be dividend yield) when you say, "fundamental data" - or not?
 
Thank you for the replies so far. There is a thread I need to work thru before replying: Re: Backtesting based on fundamental data

Thanks for the heads up!
 
ottg said:
The more I'm searching for a stock screener (for specific data manipulative requirements) the more it looks if it doesn't exist in the public domain. The quickest way to get something off the ground is to use AmiBroker which I read is a professional, powerful, comprehensive trading system development platform.

a. While it is mainly used for technical analysis can anyone indicate why it wont be suitable for fundamental analysis?
b. Any development or scripts need to be tested vigorously. It also appears from readings if Amibroker's cutting edge charting and graphics, and fast, flexible and powerful portfolio-level backtesting, optimization, and automated walk forward validation will be more than enough.
c. Do you know of any existing groups or forums that attempted this successfully and that we can perhaps join?
d. Will it be a good idea to use a development domain like SourceForge or Github to host artifacts, databasis, code and script management and central communications with feedback to this forum?
e. Anyone with the time, skills, tenacity and energy that will be interested in driving such efforts?
f. Who will be interested in active participation?

The objective is to develop a system based on a scientific approach, that will help long term investors identify profitable opportunities to buy and sell.
Click to expand...

Hi ottg,

It is refreshing to see someone else interested in this approach.

I have already developed my own system that does most of what you have in mind. I am afraid I will have to decline an offer to participate for free in public, I know it's very selfish of me.

You've mentioned that you found a thread "Backtesting based on fundamental data". I've posted some of my thoughts on the subject there, enjoy.

Luutzu gave an estimate of at least $1m. In my experience, unless you find people to work for free, he is right. I think it will be well over a million.

I agree that there are no good fundamental based screeners/back testers/charts for a retail investor.
 
luutzu said:
So if fundamental economic data are not appropriate; company fundamentals also not appropriate... what are traders to use? Share price and trading volumes? the vibe?
Click to expand...

Greetings --

As my paper states, I believe that all of the inefficiency due to fundamental changes have been taken out of the market before small retail traders have an opportunity to participate.

What profit potential remains is short term -- a few days, maximum.

The primary data used to indicate buy and sell points in a data series is that data series itself -- as you say, price and volume. I do not know what the vibe is, or how it is measured, how it is used, how it is tested.

I have found that related data series can be useful. For examples --
The price data of the broad sector the issue belongs to.
And diffusion indexes created from issues that act the same as the primary data series.

I recommend generating signals no less frequently than daily and managing trades daily. At these frequencies, all technical indicators act the same. Choose your favorite -- stochastic, z-score, RSI, etc. Tuning whichever indicator you choose is more important than which indicator is chosen.

Then be very careful to follow good modeling and simulation technique. The sequence is think, train, test, trade. Do not trade until the out-of-sample tests show that the model does recognize reliable patterns that precede profitable trades -- not just noise in the training data.

Best regards,
Howard
 
KnowThePast said:
Hi ottg,

It is refreshing to see someone else interested in this approach.

I have already developed my own system that does most of what you have in mind. I am afraid I will have to decline an offer to participate for free in public, I know it's very selfish of me.

I agree that there are no good fundamental based screeners/back testers/charts for a retail investor.
Click to expand...

Thank you for your feedback and point taken on participation. For me the most important is the fact that I'm a long term value investor. This is due to the tax advantages for SMSF when stock is kept longer than 1-year. That said it doesn't mean that I want to sleep on the shares selected. I had a look at your video and while I can see conceptually how you perform filtering, it doesn't show (perhaps on purpose) what fundamental parameters you select on.

Are you prepared to have a chat about the fundamentals you use and why on this thread:
https://www.aussiestockforums.com/forums/showthread.php?t=29587&p=861760&viewfull=1#post861760
 
luutzu said:
How much are you willing to pay for a fundamental-based system like that?

Will cost at least $1 million to hire developers, BA, DBA.

And at the end of it, don't think you could do back testing fundamental data like you could with charting or something similar.

A ratio or figure alone won't give you context... and while it's not impossible, might not be easy to code a smart enough system to take into account all the variables that may indicate a good opportunity when a few figures alone would indicate a sell.
Click to expand...

Thank you for your well intended comments. Yes to have something professionally developed will be expensive that is why my 1st choice is a commercial off the shelve solution - but it doesn't exist....yet. Next possible option is open source but just look at Sourceforge how many uncompleted projects are there. Now if I had $1m available for development then it would become a business in it self and I believe I wouldn't have this discussion here :)

Back testing fundamentals is possible. Lets assume you have the filtering done based on value and quality and then grouped according to risk acceptance. (Each one off these factors is a separate discussion). Then besides long term dividend growth and long term prices growth how did the same shares perform during a down turn and then how quickly did they regain growth during the upturn again. As new fundamental data becomes available (eg from interim reports) the quality tests may show a change in the share performance levels which becomes an indicator of possible short term changes in price or dividend growth. However you cannot mix 6-months and 12-months data but you use it separately for decision making.

Because its all numerical values I believe quantitative analysis is possible as shown by KnowThePast's share screener. Now the question is can I do the same using Amibroker as a fundamental share screener. As I will be able to do that in Mathlab therefore I it should be possible in Amibrokers as its a mathematical development platform with a different purpose.

The key is clean primary company fundamental data as input and a well defined strategy.
Economic data I will rather use in discretionary decisions.
 
howardbandy said:
Greetings --

This is a link to a paper I wrote a few years ago outlining my thoughts on the use of fundamental data:

http://www.blueowlpress.com/Use of Fundamental Data in Active Investing.pdf

Best regards,
Howard
Click to expand...

Howard thank you for taking the time to reply. I read your "Introduction to Amibroker" quite some time ago and could quickly grasp the power as I'm familiar with Mathlab in digital signal processing. A great document thank you. The problem is that it tempted me to get my hands dirty - where I would rather purchase a solution. For the above document, thank you. I will give it the attention it deserves and comment later.
 
howardbandy said:
Greetings --

As my paper states, I believe that all of the inefficiency due to fundamental changes have been taken out of the market before small retail traders have an opportunity to participate.

What profit potential remains is short term -- a few days, maximum.

The primary data used to indicate buy and sell points in a data series is that data series itself -- as you say, price and volume. I do not know what the vibe is, or how it is measured, how it is used, how it is tested.

I have found that related data series can be useful. For examples --
The price data of the broad sector the issue belongs to.
And diffusion indexes created from issues that act the same as the primary data series.

I recommend generating signals no less frequently than daily and managing trades daily. At these frequencies, all technical indicators act the same. Choose your favorite -- stochastic, z-score, RSI, etc. Tuning whichever indicator you choose is more important than which indicator is chosen.

Then be very careful to follow good modeling and simulation technique. The sequence is think, train, test, trade. Do not trade until the out-of-sample tests show that the model does recognize reliable patterns that precede profitable trades -- not just noise in the training data.

Best regards,
Howard
Click to expand...

Hi Howard,

I disagree.

To keep it simple - how do use explain the Low-PE/PB strategies outperforming the market in almost all 3-5 year periods goind back almost 100 years? It's not a big outperformance, there are better strategies, but these are well proven and they contradict your findings.

Specifically:

howardbandy said:
What profit potential remains is short term -- a few days, maximum.
Click to expand...

I found it to be the exact opposite. Most fundamental based edges that I found would require 3+ years on average to show outperformance.

Conceptually, I agree with the reasons you cite for why it cannot work. But the data doesn't support that.
 
Top