How Long Before A successful AT system

i would think that people talk about very different things
here. from tick data pattern recognition to long term
trend following. i find the battle against fitting extremely
difficult and it is an ongoing struggle. the more complex
your system, the higher the danger. our first intraday
trend follower took about three months. but we have
several systems now on papertrading, each of them
tested with quite some in- and out of sample time, and
we still find it hard to tell which one will do good in
advance. annoying ...
 
My first 1-minute AT system took about three month to design rules & write AT codes from scratch. After that, I coded several exisitng sytems into AT bots, that is much faster. One week for each additional bot.
 
Thanks so much for fantastic answers. This has been extremely helpful and encouraging. There seems to be, if not consensus, some common ideas here. I'm still reading if anyone wants to add more. I'm working on my first system right now through tradestation.

I have ANOTHER question... instead of having a computer at my own home running a program like tradestation, are there companies that provide an off-site service that I could upload my finished program to (and updated versions as well) which would trade my fully automated system off-site?
 
Quote from August:

I have ANOTHER question... instead of having a computer at my own home running a program like tradestation, are there companies that provide an off-site service that I could upload my finished program to (and updated versions as well) which would trade my fully automated system off-site?
More...

You might also want to consider buying your own server and colocating it in a datacenter. The most generic kind are those catering to web-servers (do a google on "server colocation"). For about $200 a month you can get a 1U rackspace for your server, uninterrupted power, 100 MBps internet connectivity (usually to multiple redundant backbones) and several hundred Gigs worth data transfer a month.

Some brokers allow for direct connectivity to their servers without going through the internet, e.g. via leased lines or virtual leased-lines such as Radianz, but that costs a lot more $$$.
 
I've been quiet for a while because I've been doing mass studying on trading systems based on forum recommendations.

I have a question though. Neural Networks. Is this a valuable tool for the people who have successful trading or is it only good in theory? I don't hear a lot about them, though there are plenty of applications which offer them.

Thoughts? (the material I'm reading hardly mentions them)
 
Is it a common phenomenon to have a harder time programming on the shortside?

Also - I'm working in tradestation right now and someone said that you can't short with them.... now, people say a lot of things, but I figured I'd ask if they really were super limited with shorting before I found out by surprise.
 
Quote from August:

I've been discussing my efforts for building a trade system with a couple people more than others and it dawned on me that both people say that it took about five years before their program worked so that they were very happy. It was about 3 years before it was useable.

..
More...
That is about right. If you start out with an edge and all you are doing is coding it, it could take two weeks. If you are trying to find quantifyable edges, it is a never ending research. _Even_ if the edge is built in, like if you are a market maker with close to zero commissions, you are always monitoring the market for a shift that requires your "system" to adjust as well. Even for people that have a built in edge, this is a must.

For most of us retail traders, predictive trading is the only course, and it is worse than finding a needle in a haystack, and the haystack is constantly getting more hay poured on it while you are searching...

I would say there are probably three people on ET with an audited ATS that work year in and year out after comissions. I believe I have one, _before_ comissions. I cannot currently overcome them. Giving up the B/A spread and comissions is just too tough...

nitro
 
With 3-4 years under my belt trying to program reliable trading systems( profitably but not knowing if it's due to my systems or just a distribution anomaly ), I would say that watching a trading system developper with 8000 + posts saying that he is unable to be profitable after commissions is devastating.

After watching on Wilmott's forum an option trader with 1500 + posts saying that it is not possible to "beat the market", I don't know what to do.

Perhaps I have to take my unlikely gains and look for an other business. Please Nitro, I've read numerous posts from you and metooxx and I was thinking you were the few to make a living of it... but perhaps you're saying that to preserve your edge... which is great.
 
Been at it 4 years part time and 6 months full time and have 3 markets that are now tradable and can generate new ideas in 2 to 4 weeks after the framework has been laid down.

Best advise is to work and hire traders and programmers that you can trust. Make them partners or they will "borrow" your ideas and you just wasted 6 months to 1 year developing a system for them.
 
Quote from nitro:

Giving up the B/A spread and comissions is just too tough...

nitro
More...


I bet Nitro is talking about intraday systems. I'm in the same boat intraday.

I think the answer to this scenario is to get into a higher timeframe. Solid entries with limited risk and bigger reward.


But, we all know Nitro, he won't settle for spam, he's got to have his sushi! :p
 
You must log in or register to reply here.
Back
Top