Launched High Frequency Automated Trading Blackbox
- Thread starter 2cents
- Start date
Agree.
It's not something to be discussed in a public forum. If someone is really interested, it's better if you leave a PM with your personal E-mail on it.
In terms of legal issues, some countries don't like it when a hedge fund markets their product without the proper registeration... I don't know how the internet works with marketing... but if you haven't realized, most hedge fund's website has a very long disclaimer you have to agree on before you can get to the pages with the details...
A reminder is CTA, CPO, US-based funds, and other US products are regulated under US law... It becomes really tricky with international stuff...
Again, I'm not sure how ET (internet) works... I'm not a lawyer...
TS, your broadly correct... there are proper venues to make figures available, ET's not one of them
coming back to the whole business of building, mechanically backtesting & launching a blackbox itself, i know you have a fair bit of experience in that field, therefore your inputs are always appreciated mate... at my end i'd say, in terms of lessons learnt, essentially, and for a HF blackbox:
- language & dev environment:
like most i kinda agonized b/w VB/Excel, C++ & Java. someone on ET suggested i shld look at netbeans, therefore i'll give back, that was great advice, otherwise i was going with eclipse, much heavier, clumsier etc. and then we had to make additional choices etc, which really accelerated everything
- telecoms:
i can't justify a BT Radianz type connection at this point. nor a cage at a hosting center in NY. therefore using public optical fiber & broadband as back-up, but from Tokyo to NY (180ms ping perf) => loads of time-out / session down type incidents on a daily basis. obviously happens in the midst of execution / confo message exchange every once in a while... actually many times a day... thats obliged us to work fairly hard to have a v.robust & smart auto-reconcile auto-restart process in place, to avoid the need to be glued to the screen just for manual restarts...
- backtest:
requires clean tick data. i'd say next time whether i pay for the data or not, i won't take anything on faith... i'll just upload it chunk by chunk, resort the whole lot on timestamp basis, print the resulting charts and pay a monkey to eyeball them against bloomie or any other reasonable mkt data source and find me any 10pips or more discrepancies... so i don't have to do it myself...
- API:
we played with 2 non-FIX broker APIs and the differences are, in fact, ginormous, depending on how you actually want to place orders, retrieve executions, rejects, cancels etc etc... loads of time wasted retooling parts of the execution engine just because of "minor" discrepancies between the two... if i'd had the choice at this particular point (but i don't), i wld have gone with a FIX broker no question...
- flexibility:
need to build multi-instruments, multi-accts, multi-process / thread from the start... which we did... and am so glad we did...
- confidentiality of the algo:
don't trust anyone. to sleep well, i had to break my system down into, broadly, excel-based command lines, that remain totally unclear to my IT partner, and separately the execution+ engine, the development of which i felt safe to outsource for the most part.
in addition, even if my IT partner had access to a sample excel command spreadsheet, and tried to use it just like that, it would be guaranteed roadkill...
any inputs from people with experience much appreciated, as well as any questions from traders who are seriously considering building & launching a HF blackbox...
cheers
Someone teach him something...
If you ever get to the stage where your programs and systems become truly object oriented, and at the same time require the utmost in performance without going to bare metal C, you are going to find that you wasted lots of time with Java.
Java support for Generics is poor at best. Using the power of Generics, hard core milti-threading knowledge, and understanding Patters and how they are used, are probably what seperates a beginning programmer from a programmer that has designed lots of realtime trading systems and knows how important each is. C#'s support for generics is superior to java, and the C# language will continue to move to bring more and more powerful features that you will not want to touch in other languages.
I implemented a very object oriented Core Framework in C# for building strategies, one in particular. I got an idea for another strategy, where the strategy has many complex legs to it. The framework I built ate it up and I was up in two weeks with the new model. If I had not written the framework with Generic components, correct multi-threading, and correct Patterns, it would never have been able to deal with the complexity. Most core framework trading systems are built to do one thing well, but are not flexible enough to deal with new strategies. Being able to have a flexible system that can handle new strategies is huge. It can mean tons of money because you get to trade first.
C# is going to rule the world. IMO it is imperative that C# is wrestled away from MSFT so that you don't have to use windows to use C#. Alternatively, MSFT development tools supports Linux. Miguel De Icasa (originator of Mono project) has proven this can be done. It is that important for Computer Science.
nitro
this is what is said a couple of weeks ago that god the automated Quants are taking in on the chin and the vol is back Goldman got screwed, when there is a lot of volatility the models can not determine what is value and what is not so the markets get real exaggerated moves
08-01-07 07:58 AM
I think the SPU selling off more than a 100 handles in a week is more than a shakeout...I think the "Black boxes" are tweaking out bad and another Amaranth is going to be in news soon.
08-01-07 07:58 AM
I think the SPU selling off more than a 100 handles in a week is more than a shakeout...I think the "Black boxes" are tweaking out bad and another Amaranth is going to be in news soon.