Algorithmic Trading with MATLAB Webinar

[Hlm 10]

There seems to be only a few indicators.

I guess the real question is what indicators are you wanting that were not found in the toolbox, newsgroup, or file exchange? Also, the majority of indicators are fairly simple to write in a program such as Matlab. Put simply, if you are wanting a nice charting platform with a large amount of indicators then just use one of the many charting platforms out there. Matlab is geared more towards those that don't mind getting their hands dirty and doing some of the manual work in exchange for very few boundaries.

[darthtrader2.0 10]

I guess the real question is what indicators are you wanting that were not found in the toolbox, newsgroup, or file exchange?

 

Also, not to lead people astray with posting this webinar...

If you are not interested in whats listed in the financial toolbox, the various time series analysis tools, various data mining algorithms, ect...then Matlab might just be total overkill for the job.

If you just want to do crazy stuff with indicators you would probly be far better off with something like Neoticker or Ninja...

 

For the guys that are interested in Matlab, I would be interested to know what you are doing/looking to do with it that other retail trading software can not?

 

For me this is a 10 year+ project with the eventual goal of creating a fully bayesian auto trader..data mining high frequency data with SVM and bayesian networks and then using bayesian inference to place bets.

I decided this week that I want my setup to eventually evolve to Matlab, R, HDF5 and Reuters as the datafeed....I'll figure out a broker when the time comes. The big hurdle is understanding bayesian inference and SVM for which im just sticking for the next few years with R in order to learn. ("Bayesian Computation with R is a sick sick text)

[MidKnight 12]

I watched the first video you posted Darth (thank you) and except for the iso graphic, I didn't see anything revolutionary there....The iso graphic just looked like a fancy piece of eye candy over which the 3d textured heat map already shows. Except for that image, I didn't see anything in the video that I don't have already and more. Maybe I am missing the point with it....but surely anyone could do these math studies in many common languages, such as C/C++, C#, Delphi, VB etc...Several platforms allow these languages or others that would also be capable. Some food for thought maybe.

 

With kind regards,

MK

[Hlm 10]

Matlab is in between C++/C#/VB/etc and your charting software when it comes to ease of use and efficiency. If you do it in C++ then there is more manual labor involved. If you have worked with analyzing large amounts of data with both you will understand what I mean. It is many times used as a testing ground before programmed in another language (if needed). In Matlab you have less boundaries than most charting software in such things as advanced data mining, analysis, datafeeds, etc. If you have no intention of needing this then just stick with your current charting software. There are plenty of programs out there that do automatic trading and give you the ability to insert some of your own code. As with trading, ask first what you don't have and need, then go looking for it. Many people get in trouble because they put the cart before the horse.

[Hlm 10]

Another way to look at it is that Matlab could be an answer for those that...

 

• want their own charting software.
  
• want full control over their software.
  
• don't want to worry about software updates.
  
• don't want to worry about uncontrollable bugs.
  
• don't want to build completely from scratch.
  
• want the ability to work with large amounts of data efficiently.
  
• have no problem with getting a little dirty.
  

Those are some just off the top of my head. Again, it's not for everyone.

[Guest Tresor]

Hi Hlm,

 

Would you mind updating us on your advancing with Matlab? Maybe some screenshots of the charts?

 

Regards

[Flojomojo 10]

Hi everyone...I just saw the word "Matlab" on the TL front page and got curious. I also use Matlab for my data acquisition and analysis (although I don't generate such cool graphs like the Mathworks folks do in their webinar!).

Instead of the data feed toolbox and all the data subscriptions that cost a lot of money I use this: http://www.exchangeapi.com/

It costs $300 from the guy that programed it and it absolutely does the thing for me. It connects Matlab to the Interactive Brokers Api. I only use the data acquisition ability as a Forex data feed, but you can also automate orders etc. Heres the documentation: http://www.exchangeapi.com/Documentation.htm so you can check whether it will be sufficient for your purpose.

 

Happy developing,

Flojomojo

[Hlm 10]

Hi Hlm,

 

Would you mind updating us on your advancing with Matlab? Maybe some screenshots of the charts?

 

Regards

I will post some pics after I knock out a couple more stages which should be within the next month.

[Guest Tresor]

I will post some pics after I knock out a couple more stages which should be within the next month.

 

Many thanks

[Guest Tresor]

For me this is a 10 year+ project with the eventual goal of creating a fully bayesian auto trader..data mining high frequency data with SVM and bayesian networks and then using bayesian inference to place bets.

I decided this week that I want my setup to eventually evolve to Matlab, R, HDF5 and Reuters as the datafeed....I'll figure out a broker when the time comes. The big hurdle is understanding bayesian inference and SVM for which im just sticking for the next few years with R in order to learn. ("Bayesian Computation with R is a sick sick text)

 

Hello,

 

1. Which of the toolboxes offer data mining? E.g. to calculate the odds of filling different types of gaps, etc.?

 

2. Did someone try to compare the speed of Matlab's Optimization Toolbox with the speed of optimization tools built in specifically technical analysis software like Amibroker, MultiCharts, Tradestation?

 

I requested the trial of the following:

 

Matlab

Optimization Toolbox

Neural Network Toolbox

Fianancial Toolbox

Fianancial Derivatives Toolbox

Garch Toolbox

Data Feed Toolbox

Spreadsheet Link EX

Statistics Toolbox

Edited August 11, 2008 by Tresor

[sneo 10]

i use matlab for computing technical indicators and trading on them and i tot i had a good way of getting data from xls to matlab with matlab's xlsread but apparently, i think the suggestion on DDE is also a good way (which im going to explore).

 

matlab's financial toolbox and the related financial timesseries gui has limited technical indicators. for an extensive technical tools, i recommend TA-Lib which itself has various APIs (eg C++, Excel, Java, .Net, Perl/Python). for our purpose here, i believe its Matlab which can be downloaded from http://sourceforge.net/projects/mlmechtrade. Be sure to download 32bit version (for Windows 32bit operating systems).

 

I attempted using TA-Lib on Excel, but as the data gets alot and more than 3indicators are computed, the cpu and RAM requirements shoots up. i took around 5mins to run 10 products with 5 indicators on 3yrs daily data on a non-realtime update of prices into Excel. as i need to watch it realtime, matlab provides the computation power, less than 10seconds for the same data.

[Guest Tresor]

Here is a link to a free counterpart of Matlab: http://freemat.sourceforge.net/index.html

 

Obviously this software is not so feature-rich as Matlab. I e-mailed Freemat's developer with a link to Mathworks' algorithmic trading with Matlab and asked if Freemat can provide similar performance.

 

Did enyone of you try Freemat?

 

Regards

[sneo 10]

havenot hear of freemat until your post. have jus downloaded and briefly tested, seems to have the same syntax as matlab.

 

however, reading from the intro, understand that it still has yet to have ability to read .mex files. probably cant test out my stuff from matlab to freemat to give a performance guide. however, given it could import c++ codes, maybe i will test it out further in a month or two.

 

side note, i googled on freemat and found another software which close similarities with matlab called octave.

[Hlm 10]

I have used a few of the free ones (FreeMat, Scilab, Octave, and JMathLib) and have found Scilab to be one of the best clones out there if you are using Windows. However, Matlab has more powerful graphing and nothing beats its resources and community which in my opinion is a must if you are trying to tackle it yourself. If you are looking to go down this route I would suggest Matlab and try to pick up an educational version for a fraction of the price. However, if you are looking at just messing around and not wanting to spend any money, take a look at Scilab. If you want to run on Linux, Octave is probably a good choice.

[Guest Tresor]

I have used a few of the free ones (FreeMat, Scilab, Octave, and JMathLib) and have found Scilab to be one of the best clones out there if you are using Windows. However, Matlab has more powerful graphing and nothing beats its resources and community which in my opinion is a must if you are trying to tackle it yourself. If you are looking to go down this route I would suggest Matlab and try to pick up an educational version for a fraction of the price. However, if you are looking at just messing around and not wanting to spend any money, take a look at Scilab. If you want to run on Linux, Octave is probably a good choice.

 

Thanks Hlm,

 

My evaluation version of Matlab is on its way to me. I already arranged for reasonable prices in case I want to switch to Matlab.

 

I think you are right when it comes to Matlab's community (lots of trading scripts available), all math departments at universities are equipped with Matlab. One can hire math students for really cheap money to get one's feet comfortably wet with Matlab.

 

I heard some complaints about charting quality and indicators' calculation speed and real data handling speed. Can you confirm / deny?

 

Regards

[Hlm 10]

I heard some complaints about charting quality and indicators' calculation speed and real data handling speed. Can you confirm / deny?

These questions are pretty difficult to answer without more details in what the individual is looking for.

 

As for real data handling speed, it completely depends on where you are getting your data from. I have used Matlab to grab and display live data via both Excel (with SierraChart) and a database (with opentick) and have not had any problems...and I am talking down to single digit tick charts. For the most part Matlab is just raw speed as long as you don't do anything stupid in your programming. Now if you try and use the prebuilt data links then you may have issues...I have not tried them real time. I have also not tried any techniques where Matlab grabs data from the data provider directly. Put simply, if you can get data to Excel, a file, or a database...you shouldn't have any issue with speed from that source into Matlab. Many times (but not always) complaints come from people that aren't programming things right or effectively.

 

As for the indicators, again I don't know about the prebuilt ones on small time frames, etc. But for the most part Matlab is extremely quick with its calculations. If you program them efficiently they will most likely be faster than most charting platforms out there.

 

Now we get to the fun stuff...charting quality. If you are looking for all the bells and whistles out of the box with pretty menus and easy access, then don't waste your time in Matlab. In my opinion (for trading purposes) Matlab is for those that know what they want and are looking for absolute freedom when it comes to getting what they want. If you just want to experiment with the standard indicators out there then just stick to an actual trading platform. The ability to have a similar charting quality to other platforms is there. However that will take a decent amount of time and effort. In other words, I do believe that there is a trade off here between freedom/speed and beautiful charts.

 

If you can give me more specifics on those three items then maybe I can give you more specific answers. Also, if those comments/concerns that you talk about are online somewhere that may help me figure out what exactly they may be talking about.

[sneo 10]

real data handling: i believe its a matter of appropriate coding. i use live data on Excel and have yet to see a performance impact. i am talking about at least 7000 periods (min/hr/day) per product.

 

indicators: again, using TA-Lib on Matlab, its pretty well-written such that computation speed is fast (matter of millisecs). again, its how the arrays are being handled (and what matrix types used). the question will be how not to kill the computer by overloading it with a matrix of doubles when matrix of int8can make do for a simple binary signal generation.

 

charting: i presume if there is a need to use Matlab, probably coding a little more for charts won't hurt abit.

 

references can come from: http://en.literateprograms.org/Category:Programming_language:MATLAB

[Guest Tresor]

These questions are pretty difficult to answer without more details in what the individual is looking for.

 

As for real data handling speed, it completely depends on where you are getting your data from. I have used Matlab to grab and display live data via both Excel (with SierraChart) and a database (with opentick) and have not had any problems...and I am talking down to single digit tick charts. For the most part Matlab is just raw speed as long as you don't do anything stupid in your programming. Now if you try and use the prebuilt data links then you may have issues...I have not tried them real time. I have also not tried any techniques where Matlab grabs data from the data provider directly. Put simply, if you can get data to Excel, a file, or a database...you shouldn't have any issue with speed from that source into Matlab. Many times (but not always) complaints come from people that aren't programming things right or effectively.

 

As for the indicators, again I don't know about the prebuilt ones on small time frames, etc. But for the most part Matlab is extremely quick with its calculations. If you program them efficiently they will most likely be faster than most charting platforms out there.

 

Now we get to the fun stuff...charting quality. If you are looking for all the bells and whistles out of the box with pretty menus and easy access, then don't waste your time in Matlab. In my opinion (for trading purposes) Matlab is for those that know what they want and are looking for absolute freedom when it comes to getting what they want. If you just want to experiment with the standard indicators out there then just stick to an actual trading platform. The ability to have a similar charting quality to other platforms is there. However that will take a decent amount of time and effort. In other words, I do believe that there is a trade off here between freedom/speed and beautiful charts.

 

If you can give me more specifics on those three items then maybe I can give you more specific answers. Also, if those comments/concerns that you talk about are online somewhere that may help me figure out what exactly they may be talking about.

 

Hi Hlm,

 

The issues are just opinions that people express on different internet forums like elitetrader.com This is just a one example of such a thread: http://www.elitetrader.com/vb/showthread.php?s=&threadid=95273&perpage=6&highlight=matlab&pagenumber=1

 

My trial Matlab arrived to me today so I hope I will make my own opinion on this software.

 

I agree with you on the charting issue. The look of the charting is not important as it is not charting that make you profit from Matlab.

 

Regards

[Guest Tresor]

Hello Guys,

 

I arranged to meet with a Matlab user during this weekend so that he could show me what can be done in Matlab with regard to trading, but before this happens, I would like to play around with Matlab by myself.

 

I have a big favour to ask you to help me to write a proper command to import a csv file. I am trying to import the data of the instrument I trade into Matlab, but due to my poor knowledge of Matlab I can't do it.

 

I typed in the command ''csvread'' + the path to the file but Matab says:

''??? csvread(C:\Program Files\Notowania2PRO\PLIKI_CSV\fw20.csv)

|

Error: Unexpected MATLAB operator.''

 

Could you please advise me on the proper syntax / wording of this command. I attach 2 screenshots:

 

(i) CSV file - this shows how the columns and rows are structured (this file is constantly being updated on a tick basis. I would like Matlab to plot historical data and the new future data

 

(ii) data import in MATLAB - unfortunately I cannot create vectors from each column as my CSV file does not have column names

 

Thanks

[sneo 10]

1. define the matrix to import data from csv to: e.g dataFW20. this is to help to manipulate the vectors later on when you decide to do computations.

2. keep ensure that ur current directory is pointed to where the folder of the data is. this get pass the need to define the C:/blahblahblah.

 

try: dataFW20=csvread('fw20.csv'); <= add the quote (you may be getting the error if there aint no quote) and the semicolon (to hide the ans from command screen).

 

if you have doubts, the Help in Matlab is quite helpful, in fact, the one of the rare helpful "Help" i have seen from many softwares.

[Hlm 10]

Exactly sneo, the help command is very helpful and can usually give you great information without having to go to their site. Tresor, just type help csvread into the command window. I also love how it has "See also" commands and "Reference page" links.

 

Full command should probably be....

 

dataFW20=csvread('c:\Program Files\Notowania2PRO\PLIKI  _CSV\fw20.csv');

If you are having problems with text within the file you could just use xlsread instead...

 

dataFW20=xlsread('c:\Program Files\Notowania2PRO\PLIKI  _CSV\fw20.csv');

Edited August 22, 2008 by Hlm
Added last bit of code.

[bokresik 10]

Just thought I'd say hi and introduce myself. I'm an expert user of Matlab. I've been using it to analyze market data and develop algorithic trading systems for about a year. The flexibility is amazing and there is nothing you can't do with Matlab. I live in Los Angeles. If anyone is interested in collaborating on new research I would like to discuss more.

 

Brad

[Pava 10]

Hi..I live in Los Angeles too...I will be very interested in knowing..if it is possible to perform some staitstical or neural analysis on zigzag indicator to creat a prediction...attaching jpeg to illustrate my point

[bokresik 10]

Excellent! I am very interested in the zigzag indicator and have used it quite a bit. I have several ideas about how to improve its use and attempt to detect swing points. I don't use neutral nets, but I know they have merit. We could look at the histogram of time between highs and lows as a function of time of day and historical volatility. This would be an interesting study. I don't like use the word prediction because it is all probabilities. A good trader is wrong 40% of the time, so I think of it as trying to find a statistical bias. We should get together and talk about this further. I'll send you a PM.

[sneo 10]

wont it be easier to do pattern recognition with MATLAB instead of just zigzag indicator? it covers more grounds than zigzag and can identify the % likeliness against a current price pattern that is identified by user.