Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
The Dakota system is able to run multiple systems (strategies) at once, and instead of optimizing them based on historical curve fitting by parameters, it will optimize them based on real-time interaction with the marketplace -- ie how the entire portfolio of strategies interacts with the market, overtime forward-testing "out of sample" using what they call "adaptive flocking", and allows each unique strategy to communicate with each other as a whole so they can learn which areas of the market or timing of the market is most profitable.
These bots seem to support Visual Basic and C# (.NET).
Lately I've been on a kick of trying to work on non-correlating strategies built to be profitable on random walk data instead of curve fitted to historical data, in other words strategies that are adaptable and primarily focus on risk and money management instead of focusing on a particular "setup" to occur.
It seems like Dakota might be useful. I am soliciting feedback and opinions from others, has anyone used this stuff?
I am familiar with the Biocomp products, and have worked with them for many years.
I have multiple-instance licenses for Profit, Dakota, and Patterns.
I recommend them very highly for doing research, and for trading of moderate-frequency data if you can place orders manually without incurring too much latency. I doubt I could do "HFT" without a team of specialists to help, and a solid company backing the project even though Biocomp's products are quite scalable to large clusters of computers. I would be happy to help anybody who wants to try....
I have also worked with intraday bars even on the end of day versions, and other "bar types" as well on Dakota, with at least theoretical trading success. I still think people are much more intelligent at trading than computers. The main difference is that computer bots are much much faster, do not succumb to emotional failure, and sometimes have much much better endurance and attention span. They are also immune to intimidation by lawsuits and other coercions. Unfortunately that might lead to arrogance and negligence, if not criminality on the part of their masters...
When it comes to making "bots", I'll have to confess that I still have not figured out how to succeed with rigging up any automated trading. Dakota "bots" are instances of running adaptive strategy modules who combine their signals into a final output, not bots which interact directly with a broker. For me, auto-trade is still pie in the sky, an avid hobby with little real trading money in the game. Sometimes a friend tries some real money with strategies I write, but I cannot yet say I could start any (non fraudulent and profitable) commercial trading enterprise with my "robots". That does not discourage me from having a lot of fun and learning a lot about trading, but I have a hard time believing anybody who make claims of having found the "grail". (LOL)
I am even skeptical that a majority of the big firms are not kidding themselves and heading for a date or worse with the Black Swan. (If they haven't already....every trader has a counter-party in the trades, and a lot of "margin" is fake) It is much harder to pull off even small-potatoes RT trades (real time) successfully than it is to dream about it and run back tests on historical data, IMO, from experience with TradeStation. TradeStation data feed seems to be good data, but is interrupted over TV cable and subject to intermittent delays and "refreshes". That can be overcome for slower than HFTrading, but hooking Dakota, Profit and Patterns together AND to a trading platform and broker is a *****.
I am now using Multicharts as the integrating platform with TradeStation's data....moving forward but not arrived yet.
Biocomp will get hooked up. Can't say how long it is going to take....depends on how many people help me.
Where are the example "bots" template(s) for Multicharts to interface with your competing robots project?
I got rather irritated fighting with trying to program Ninja. Even C# is not an easy strategy prototyping language.
I get a headache when a software based strategy is much longer than one page of code. That is just one of the reasons I would like to use a "swarm" of simple strategies instead of one or even a few overwhelmingly complex ones.
One likely prototype scenario for me will be to put (several) Biocomp instances on modest server(s) in the cloud, and have it(them) deliver crunched numbers to my (and my partners') platform(s), also located in the cloud, and the delay and downtime, etc. etc. would be mostly to my (and my buddies') screen(s) and (overriding/trade monitoring) DOM-ladder manual trading setup(s). No "optimizing" on my trading/charting platform! Too many crashes.
Anybody else out there interested in Biocomp stuff?????.
Still, I do not think it is impossible, at least for a while until it all stops working.
Hi Seberbach and community -
I am a multicharts user and interested in Biocomp products and swarm optimization in general. Wondered if anyone has successfully integated Biocomp into Multicharts?
thanks
No I have not yet successfully integrated Multicharts with Biocomp products.
Give me a team of programmers, a large budget, and considerable time, I probably could guide such an effort to fruition. But the complexity is immense. Just think! Write a few APIs to put up graphics, icons and windows, call it Windows, integrate it with DOS, and we can be billionaires! No big deal.
20 years later: Finish the project? Never.
Not that I have not thought about this a lot for a decade....and I have lifetime licenses to both of your proposed integrands. I am an engineer by career, not a programmer. I know a lot about computers, but have not written programs full time for half a lifetime. Certainly I have not enough programming skills to just pop this out in a few weeks', or even months' work.
EasyLanguage has problems with reading text files, but can quite easily write text files. That is one barrier. Biocomp uses dot net and has scripting ability, but lacks ability to deal with order types such as limit, stop, stop limit, bracket, and calculating and ordering "inside bars" using tick data which is used to build the bar-type formats. Carl at Biocomp certainly has worked hard at studying ways of dealing with these issues, but it seems there is not enough interest in the software market to make that high speed tick data handling top priority.
I bought the dot net version of Multicharts, and find it a huge learning barrier to translate my EasyLanguage to dot net language, (documentation is improving though, given a couple of years to develop) and a formidable task to read and write files for the integrating. I did manage to integrate a spreadsheet called WaveWi$e to NGO and Profit to perform various different methods of model, optimise, select, test multistage-validation folds interleaved into patterns of in sample/out of sample instead of just dividing data sets into four sequential folds (MOST). It is feasible to use the end of day Biocomp products to work with intraday bars of any type, including types you invent yourself. All by writing sequential date, with any number of records per date, text files from the spreadsheet and reading into Biocomp. The problems arise when you want to export back to the trading execution platform for trading in real time, or wish to trade from Biocomp directly using other than just market orders on bar close or opening. That requires writing a new "equity engine" and/or sending the partially processed signals back to your Multicharts. With Solid State Disks and large RAM disks, file operations are getting pretty fast and can handle millions of ticks per file, but platforms lack communications ability due to their proprietary nature being unfriendly to other people's software being added in mash-ups.
I would be interested in corresponding further on this, but I certainly do not have any ready made solutions for these issues yet. It's a dream, like maybe someday climbing Everest.....
Thanks Steve for the contribution.
From what I understand Dakota and most neural nets are stop and reverse types of analysis. I realize Dakota is not neural net, but swarm optimization, and I like this idea because theoretically you would see the system that is "leading the pack" - correct?
My interest is more in discovering relationships that will identify a trending or mean reversion environment. From there I have systems that will enter on the execution side.
I believe that this is a concept that can be done with some optimization that does not need to be evolutionary, so I am considering products like Neuroshell's Chaoshunter, or even more a product called StrategyQuant.
I appreciate you relating your experience transitioning to .NET, it is something I have been considering more and more now that it seems that MC has improved the editor and documentation.
Since you use Biocomp products, or have life time license - would you mind talking about how you use them for trading - meaning the actual concept to execution process?
Also - any comments on their new product- Patterns?
Thanks -
I am just finishing a week or month long project of re- writing my old Dak 2.n bots and equity engines to the new Dak 3. V3 is much improved. The bots and other add ins can be written in VB or C#. V3 allows a better look under the hood and separates out the large functions of the original Dak 2 Equity engine code.
I was kicking and screaming about moving and rewriting but I am glad I did. That old Dak 2 script writing was horrible. In addition, Carl is always ready to help on his forums.
One quick note: I have found I need to run both Dak 3 and Visual Studio as admin to make sure all builds of dlls take place and debug works properly.
more later.....
Brian
