Return Styles: Pseud0ch, Terminal, Valhalla, NES, Geocities, Blue Moon.

Pages: 1-

Graph Analysis

Name: Anonymous 2011-06-07 22:20

So why don't more of you /prog/rammers write software that trades stocks or currency? I've decided I'm going to try trading currency, and I set up a scraper to put exchange rates into a data base every 5 minutes.

Maybe I'll find out otherwise when I start the software in a few days when I have actual data, but isn't it just graph analysis?

And if it isn't, couldn't you just set a genetic algorithm on the market or something?

Name: Anonymous 2011-06-07 22:26

MORE LIKE
JEWS ALREADY DID IT
AMIRITE LOLLLLLLLLLLLLLLLLLLLLLLLZzz!!11oNE!!1ONE1!

Name: Anonymous 2011-06-07 23:36

>>1
Markets don't follow a set pattern. You need to model supply and demand, which is determined by human action and motivation (have fun subjecting that to analysis).

There are far too many variables in a modern economy to simulate it effectively.

Name: Anonymous 2011-06-08 1:25

>>3
But the guy in Inception came up with the perfect formula.

It must exist!

Name: Anonymous 2011-06-08 1:29

>>4
Your dreaming.

Name: Anonymous 2011-06-08 2:00

Let me get this straight op. You want to buy stocks or whatever and then when your stocks or whatever increases marginally you sell it and make money off the difference and repeat?

Is this your strategy?

Name: chris 2011-06-08 2:03

Inception != Limitless

Name: Anonymous 2011-06-08 2:35

PI

Name: Anonymous 2011-06-08 2:44

Protip: Bitcoins are hot.

Name: Anonymous 2011-06-08 2:47

Bitcoins. The payment of evil!

The only reason to use Bitcoins is to avoid taxes and such shit.

Name: Anonymous 2011-06-08 13:35

Sounds like an interesting project. I have also had the pleasure of penetrating the anus of financial ``data'' using a FIOC web scraper.

What framework if any do you use? (i use pig-disgusting code based on beautifulsoup)

Where do you source your data from?

Name: Anonymous 2011-06-08 16:24

>>1
Data sources are important. Google/Yahoo Finance are not realtime/accurate/etc... You'll want to try directly source Forex data. Good luck paying for that!

Also, Forex data is Intraday - which means you have a lot of data on your hands... Trading at once a minute during a trading day(10am -> 4pm) you're going to have 420(7hours by 60minutes)table rows per stock per trading day. If you were to keep a month's worth of data (12,600 table rows [420 x 30]) for 18 currency pairs, you'll end up with 226800 Rows. This is best case scenario, Forex Data can update as often as every second.

Not that this is a massive amount, but designing a system where you can comfortably manipulate at least this amount of rows is maybe a little daunting for less experienced programmers.

How are you looking to do pattern matching? DNA based? Or just change up and down? Will you compare past performance to news? Will you compare to other pairs, etc?

What sort of software will this be? Perl, Python, .NET, etc?
What sort of DB do you have in mind?

Not trying to piss on your chips or steal your ideas -> I do this sort of stuff for a living so I'd like to help a wee bit if I can...

Name: Anonymous 2011-06-08 16:29

>>1

Google ELLIOT WAVE


Also, TIME CUBE

Name: Anonymous 2011-06-08 16:34

>>12
What sort of software will this be? Perl, Python, .NET, etc?
What sort of DB do you have in mind?
What the dicks does that matter? Those aren't the things that will stunt any form of development of this kind of stuff.

Name: Anonymous 2011-06-08 16:41

>>14
You're quite correct.

I'm only asking out of curiosity. If he's using some platform etc I know about, I'd be happy to recommend tips/packages/books etc...

Name: Anonymous 2011-06-08 17:16

>>15
Stop being a normal person on /prog/

Name: Anonymous 2011-06-08 19:14

>>16
You're not yet angry enough

Name: sage 2011-06-09 0:42

Erlang

Name: Anonymous 2011-06-09 2:08

OP here.
Wow, I didn't actually expect this thread to get many responses

>>6
That's exactly the strategy I had in mind if I used graph analysis rather than a genetic algorithm. It would calculate very minute changes and perform trades very quickly, rather than looking at long term prediction.

>>11 and >>12
I'm using C# and .NET if that's what you're asking. The scraper uses a package called the HTML Agility Pack. I think somebody asked about that somewhere... I'm using .NET mostly because the linux partition of my computer is pretty fucked and I don't have a connection good enough to get a recent disk image and install dev tools.

To acquire data, I scrape rates from a page from a company called Oanda. The rates are updated every 5 minutes, so my scraper checks every 2 1/2 to make sure it has accurate and up-to-date data. It's keeping all the useful data in a comma separated values file. I plan on using this data to run simulations, but I'm not sure it's accurate enough to actually use to trade currency.

>>12
I'm glad you're interested, because that last bit there is where I'm still trying to figure out what to do. I don't have any experience in AI or genetic algs, so I'm using my spare time to write one that plays tetris... What do you mean by pattern matching, etc? Although, I'm still writing tetris itself so I haven't started that quite yet. I should be done by tomorrow.

And the only question I've yet to answer is the database one that someone asked. I suppose I should switch to using a relational database instead of a text file at some point, heh.

Name: Anonymous 2011-06-09 4:56

>>19
If you want to retain huge amounts of information use a neural network.

Name: Anonymous 2011-06-09 8:30

>>19

Lasts posts I read about Oanda said it is a swindler company.
Enjoy being fucked by them.

Don't change these.
Name: Email:
Entire Thread Thread List