Sorry. I mean win rate. The strategy has a 30 pip daily max profit and a daily max loss. So specifically I mean that 70% of the days end with + pips and 30% with rest with - pips. I actually had some testing that showed a 100% weekly win rate. But further testing last week showed that was not valid. The weekly rate was only about 75% win rate. Wayne
you might want to check out this cluster for some ideas... http://labs.microsofthpc.net/ it would be great to have, but the electric bill would be a killer!
Not to irritate you, but clearly your experience is lacking in skills at least for C# C# is almost as fast as C++ and even faster when it comes to handling of small objects because of the garbage collector. Java is just a slow beast, but has very nice features though C# has very strong support for OpenSource take CodePlex for instance! Take a look here: http://www.nuclex.org/pages/csharp-vs-cxx-performance
greaterreturn -> Why even bother with DB40, imho it's slow Have you looked here: http://www.nanex.net/OPRADirect/OPRADirect.htm This is a strong compression routine, I'm sure it can make some nice performance. Have you thought about loading data into a list and then just use Indexed LINQ? IMHO the bottleneck must be in finding and reading data between 2 dates, maybe a single file structure could fix that. A file name consisting of date and symbol and then use Indexing Services to fast access Well... just some thoughts Cheers Janus Some old link I found in my bookmarks: http://kx.com/
Thanks Janus. FYI, everyone I turned of email notification on this forum temporarily since it distracts from my research. I don't know when I can get back to working on releasing TickZOOM. I 'll probably just start a new thread at that point. Wayne
Just to clarify. I said "forum" but I mean this thread. I still get notified of PMs and I'll check back every couple days here. I'll still let you know when the other site is ready and the forum on nabble here on this thread. But I'm not certain when I can get back to work on releasing the code. Wayne
It seems to me if your ultimate goal is to take it commercial, its a good idea to do research what is needed in the department of smaller institutions, startup/incubator funds who need to watch their expenses closely and cant commit to solutions like orc, portware or apama etc. - but still can actually pay for support/additional coding jobs (as opposed to selling lots of 'perpetual' licenses to private traders and support them on a best effort basis via forum - this is bound to be a disappointment for them - as can be witnessed on most of those ISVs open forums everyday). But it sounds like you are busy for the foreseeable future now anyways. Good luck, P.
Wayne, I have no familiarity with db40, but I'm quite familiar with MySQL, Eclipse and several other mainstream FOSS projects. My personal experience ranges from having my own business as an independent telecommunications software developer and, on the corporate side, as a developer and architect of large-scale messaging and network security infrastructures for several multinational Fortune 500 companies. I've also was heavily affected by and had to live through the disruptive time during the AT&T law suit against Berkeley Software Design who had commercialized BSD/OS. That said, none of the projects that I'm familiar with requires someone to surrender ownership of their contribution to the project. Take MySQL as an example, their CLA can be found at http://forge.mysql.com/contribute/cla.php It clearly states that joint-ownership of a contribution is preferred. But, if that's not possible, a permanent, non-revocable license be granted so that the contribution can be re-distributed. Look, I really think you're onto something that could do a lot of good for traders of every stripe, but I'm not so sure that an open-source license like the GPL is fully compatible with your ultimate goal for TickZOOM, especially when it comes to the rather sticky area of dual-licensing schemes. All I can suggest is that you run through what's been posted at the following web sites and get some legal advice from a lawyer who's familiar with open-source licensing so that you can review your plans and make sure that you have a firm legal footing for the project. http://www.oss-watch.ac.uk/ http://www.opensource.org/ http://producingoss.com/ http://www.softwarefreedom.org/
Pippi, Thanks for this most wise suggestion. That's makes total sense. I will add that some traders are "professional". What TSGannGalt means by that is that they come from the trading industry or work as traders who already make money trading at big trading firms, prop shops, etc. They have money and almost qualify as your small "firm" below. They understand the need for a solid commercial platform that's reliable, redundant and super fast because they have or had access to those type of tools you mentioned. Are you one of those? Most in the public never heard about those. I'm going to eventually have to do some reconnaissance on those system just to at least understand what's out there in the industry better. By the way, Pippi, I just found out by someone in PM that I have to get the multi instrument and higher performance features working in TickZOOM right away even for myself. Why? They reminded me that's the route to more consistent profits, diversifying strategies. I originally set out with that as my goal with this platform for myself but never finished that last stretch yet. Anyway, let's not give up on releasing TickZOOM yet. Something, makes me think that a solution will be forthcoming to keep this project going. In the meantime, I'm working on things that are still in the direction of releasing the code (and for my own trading). For example, I spent last night and today, so far, getting all the automated tests up-to-date which continuously test all the calculations for accuracy. I also optimized data loading with the next step being eliminating the memory limit you referred to in NeoTicker. I already changed the double's back to int's which brought the processing performance somewhat back. But I realized some other features I added recently have slowed it down so I'm going to test how to optimize those. Optimizing for speed makes it about 3 times more difficult to program than normal. Notice, that as soon as I try to run my test against multiple instruments, I'll immediately get sick of the long run time again for all that data on a single CPU since I have a quad core. In that case, I'll have to solve the parrallelizing of the back test. Of course, it already does that for optimizing. But I don't do much optimizing any more. I find I can look at charts and pick out good parameters nowadays. But I do want this thing fast enough to run optimizations for research purposes at least. Let's not give up hope. I originally said I could get this out after the new year but got over excited by all the response. Some people PM me, say "take your time" "don't burn out" ,etc. That's very kind of all of you. And much appreciated. Sincerely, Wayne