Thanks Carl. I need time to digest some of the stuff you said but you are good. I just hope this stuff won't cost me an arm... how about a discount for a first-time retail (poor) customer (and get it back on the upgrade path) I also saw a Neural Network product on your site. What happen to that? Are you phasing them out? Regards, KC
Here is the definition of Dakota swarm technology from your web site: Sounds like a textbook description of GA technology. Anyone reading this description would say this is a hill-climbing optimization algorithm of the Genetic algorithm type. If not how does it differ? BioComp has been around for years and you say this technology is great for trading financial markets. Why not post some performance of Dakota in real time? For example, train a system, post the trades for the next day and then at the end of 3 or 6 months you would have a great marketing tool to promote your product. How about next week? Do you have some real time performance with trades generated by any of your technologies? You can start a Journal here that I am sure would be widely followed. Looking forward to seeing it trade! Thank you
First, having been in the commercial GA business since about 1987 and personally being the author of the first GA library for Microsoft Windows, I can confidently say that Swarm Technology is not a GA. As an aside, GA's are not hill climbing optimization algorithms, and the fact that they are not is one of their fundamental strengths. They are combinatorial search algorithms which enables them to span discontinuities which confound most hill-climbing algorithms. Yes, because GA's are quite robust, they can be applied to hill-climbing problems, but that is not what they are designed for. There are much better algorithms for hill-climbing. Second, Swarm algorithms are applied in Dakota as adaptive technologies (go towards better) while GAs are optimization technologies (go jump on "best"). They are an optional adaptation algorithm in Dakota. We also have a hill-climber "Equity Controller" algorithm and you can write your own. I think you misunderstand Dakota. First, you are exaggerating what I said. I said they are helpful. Second, what you are asking for is like asking Tradestation or Equis to "train a system" and post results of Tradestation or Metastock. They too would likely look at you funny, because they are tools to create systems, not systems themselves. Dakota, like other trading software, is a tool to build trading systems, not a trading system itself. With Dakota, you build a swarm of adaptive trading systems using one of about a dozen provided trading bots (systems) or you can write your own. Which bot to use on what ticker is guided in part through a "Knowledge Base" feature whereby users voluntarily contribute their findings on what works and what doesn't. You mention "training". There is no training in Dakota. There are some video examples on our web site of swarms and their hypothetical performance, which is how this whole thread got started. In our newsletters I occasionally give status on how these trading systems are continuing to perform. At last check all but one were making additional hypothetical gains. I honestly hope this helps clarify your understanding. Thanks, Carl Cook http://www.BioCompSystems.com
I was surprised to see that this is the only thread on BioComp Dakota. In my humble opinion Dakota is one of the best, if not the best software application for the creation of market timing signals. I model the SP contract exclusively. You can see some screen images of my 2 main systems at: http://cid-ccc1c8d8b047979c.skydrive.live.com/browse.aspx/Dakota Systems Kind Regards, James
All the great posts from people with 1 or 2 previous posts describing how great BioComp Dakota is without any performance or followthru. Ok, so what are you trading tomorrow? Pick your very very best system and share the setup for tomorrow. Posting is free, why not call some trades? I bet there will be a great deal of interest if you only post results for 1 or 2 successful days before they happen. Good trading!
I am happy to post my systems trading signals for a couple of months. I will use Sp_z0020 alone to keep it simple. Tue April 1, 2008 Buy 1 ESM8 MOC Kind Regards, James
Signal for Wed April 2, 2008 MOC: Long Trading Instructions: None Open Position: Long 1 ESM8 @ 1371.00 Trading History: 4/1/2008 Bought 1 ESM8 @ 1371.00 Note: This is a hypothetical trading record. Kind Regards, James
James, I would be a lot more impressed had you have said buy MOO today (rather than MOC). Anyway, does your system use one of the canned bots they provide or did you write it yourself? I'd be interested to hear any details you would care to share. One of the problems I perceive with the Dakota product for a potential customer is that one is not sure what is it one really is getting? After all what is SWARM technology? Initially I thought it was just the application of Particle Swarm Optimization (if you're familiar with that) but subsequently I've come to believe it's that sophicated after viewing some of the videos. Are you using just the EOD product, or is the realtime version a better proposition? Richard
Hi Richard, "I would be a lot more impressed had you have said buy MOO today (rather than MOC)." I knew that comment was coming! "Anyway, does your system use one of the canned bots they provide or did you write it yourself?" I mostly use script bots that I have written myself. In brief, my system is composed of 39 Dakota systems that produce different, but historically profitable signals. The final market timing signal is generated by an ensemble signal trader script bot that dynamically uses subsets of the 40 signals based on equity performance walking forward. One of my main aims is to create diversity amongst the Dakota systems or subsystems. A bit like building a diverse portfolio. The modeling method that I use applies a number of tests for robustness including running the swarms on 5 years of data prior to the 'modeling' period and 2 years of data after the modeling period. I have only rejected 3 swarms based on out of sample performance! That's an acceptance rate of over 90%. "Initially I thought it was just the application of Particle Swarm Optimization" Dakota is better described as an application of Particle Swarm Adaptation. The Trading Bots adapt from bar to bar in the ever changing Bot Parameter / Equity space. There is more to Dakota than just the adaptation. If Dakota did not have the PSA I would still use it! Although the PSA does increase the level of performance. "Are you using just the EOD product, or is the realtime version a better proposition? " I am using the EOD product. Partly because I don't have time to monitor the market during the day (night for me I'm in Sydney, Australia) and the R&D work that I have been doing has kept me busy. It is probably a good idea to start with daily data and when you have made progress consider fine tuning with the realtime version. Daily data is easier to work with in my opinion. Ofcourse, it is possible to apply tradng rules ontop of the Dakota signal. Examples are stop loss, re-entry, and profit taking strategies. Infact, my trading partner went Long at 1344 by applying some position entry rules ontop of the signal. The Dakota signal can be used as the basis of a trading system. Kind Regards, James
04-02-08 03:57 PM Signal for Thu April 3, 2008 MOC: Long Trading Instructions: None Open Position: Long 1 ESM8 @ 1371.00 Trading History: 4/1/2008 Bought 1 ESM8 @ 1371.00 Note: This is a hypothetical trading record. Kind Regards, James