Not sure what I've built.......

Quote from boskop:

Interesting. I'm working on a similar project for over 2 years next to my day job. I only use C/C++. The overal output of the system consists of buy/sell signals as well. I do not have any results that I feel comfortable to write here, except for the many concepts that proved to be useful or not in the overal chain of decisions that eventually result in such buy/sell signals. It's a long and exciting journey through ideas and experiments.

If you see consistent results and small drawdowns, I'd go live with a small account and see how it performs in the wilderness.

Do you simulate slippage?
More...

Cool. I had hoped others were working on similar projects. I hate to ask cause its a tough audience, but can you define slippage?

Also curious what others are using for a data feeds. Are you storing intra-day price & volume data? I'm sniffing packets of a data stream and then inserting to a DB.
 
Quote from dreturns:

Cool. I had hoped others were working on similar projects. I hate to ask cause its a tough audience, but can you define slippage?

Also curious what others are using for a data feeds. Are you storing intra-day price & volume data? I'm sniffing packets of a data stream and then inserting to a DB.
More...
By slippage I mean the difference between the expected price of a trade, and the price the trade actually executes at.

The historical data is a topic I have not yet spent much time on. I use the tickdata available at http://ratedata.gaincapital.com/. Only EURUSD bid/ask for now. After correcting a few errors it seems useful enough. If I ever find a good model, I assume it has to work well on any source of historical data. For the model generation I use various filters that should be able to cope with strange input data (including news events). I wish I had data from a real feed though...
 
Quote from dreturns:

I have been working on this hobby project for years. I have had a live feed since 2004 for 30 stocks, approx 200 million data points saved. I have built a process around a Linux&OracleDB server that spits out recommendations of when to buy / sell. I also have it back test to adjust variables over time.

All results are tweeted in real time so I can watch remotely while I work the day job.

When I use it, I have made 30 - 40% returns / yr. Which is not enough to quit the IT day job.

I want to ask, does anyone think this thing has any value to anyone else other than me?

Has anyone else here had a similar project?

Thanks in advance.


Link to historical results: www.dailyreturns.com
Twitter account: dreturns
More...

dRetrend;


Well nice work during a good market,MSFT....ORCL.
30% isnt enough to get a gold house, sell it,, but its enough to trade yourself:cool: Wisdom is profitable to direct.

Dont worry [generally speaking LOL]on slippage on those liquid stocks, they bounce around enough to limit in order nicely,usually:cool:
 
Quote from boskop:

By slippage I mean the difference between the expected price of a trade, and the price the trade actually executes at.

The historical data is a topic I have not yet spent much time on. I use the tickdata available at http://ratedata.gaincapital.com/. Only EURUSD bid/ask for now. After correcting a few errors it seems useful enough. If I ever find a good model, I assume it has to work well on any source of historical data. For the model generation I use various filters that should be able to cope with strange input data (including news events). I wish I had data from a real feed though...
More...

For my purpose, its just a recommendation. I don’t worry if there is slippage. What I find however is the price typically goes lower than the models recommended buy price. I attempt to cover that with stats on time to reach the min price after the buy rec and before the sell rec.

The historical data density matters a lot I would think. If you use any second derivatives (calculus rate of change not financial) and or slopes. The number of data points will affect those values greatly. You would have to invent constants of some kind to compensate.

Many here and in messages say put up some money. I have. It seems to work. But, like everyone else with a wife, house, kids, and a basement full of servers the trading budget is for lack of a better term, limited.
 
This has been your hobby for years, your basement is full of severs, right?
Just decide how much your project is worth, and pull the plug when you have lost that amount of cash.
 
Quote from dreturns:
Many here and in messages say put up some money. I have. It seems to work. But, like everyone else with a wife, house, kids, and a basement full of servers the trading budget is for lack of a better term, limited. [/B]
More...

Have you considered signing up with a prop firm?

I started developing a simulation framework recently, in R and C++ and using historical data from WRDS plus live feeds from IB. It's taking a while due to work and school and family commitments. I don't have a basement full of servers yet; in fact I'm considering deploying my strategies on the cloud eg AWS when I'm ready to scale up. But once my systems are self-sustaining, my plan is to get an account at one of the prop firms out there, then build up a decent track record, then eventually pitch to one of the quant funds out there which are open to new strategies. It seems that a prop firm would be the logical next step for you too, given your circumstances.
 
Quote from dazzwater:

Have you considered signing up with a prop firm?

I started developing a simulation framework recently, in R and C++ and using historical data from WRDS plus live feeds from IB. It's taking a while due to work and school and family commitments. I don't have a basement full of servers yet; in fact I'm considering deploying my strategies on the cloud eg AWS when I'm ready to scale up. But once my systems are self-sustaining, my plan is to get an account at one of the prop firms out there, then build up a decent track record, then eventually pitch to one of the quant funds out there which are open to new strategies. It seems that a prop firm would be the logical next step for you too, given your circumstances.
More...

Yes as far as prop firms. I did try it for like 6 months. The issue there was making my models fit the same day buy sell. I tried but could not rope my models to same day buy sells. Overnight @ 4 to 1 wasn’t helping much. Then came the series 56 test.

As for servers in the basement. System, network, and database administration. Not to mention the feed source. Code the logic for the models. Then holy crap you need an interface. Then you need to tune it so it can scale. If you are technical enough, F the cloud, its not that great of a savings and there is no control. Servers on ebay are cheap.
 
It is not what you do not know, that will hurt you. It's what you do know and isn't so, that will hurt you. I would educate myself real good about trading in real life before 'believing' any computer model developed by a non trader person...
 
You must log in or register to reply here.
Back
Top