[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/biz/ - Business & Finance

View post   
File: 175 KB, 456x305, Algorithmictradingsystem.png [View same] [iqdb] [saucenao] [google]
933954 No.933954 [Reply] [Original]

Anyone know algo trading?

I'm thinking about getting it in. I'm just wondering if reading a few investment books and implementing an algorithm on their teachings will be enough to get me started, or if it's harder than that.

Also:
>What is the minimum capital to start?
I have 3k liquid and 500 that I grew from 400 riding off HEAR bullshit
>how different is simulated trading from live trading?
>is this a jew scam?

>> No.933955

>>933954
Also, I'm a computer engineer so I should have no problems getting an algorithm working and debugging/tinkering it.

>> No.933963

Well what you need is a statistical trading edge, and if you have that you can design a program around it. The nice thing about algo trading is you can backtest your strategy, and find out objectively it is a true edge, which is a lot better than most of the chucklefucks on this board who think their "buy the breakout" strategy is hot shit.

>> No.933973

>>933954
I have a question are you trying to Get a program for HFT? I mean Programming a basic algorithim to trade for you wouldnt take any longer than an hour in a high end langauge like python. But I would think you are retarded to shit out a 1 hour script and throw money into it right? I really wouldnt read fuck all books when it comes to any trading, they are all generally useless unleess they start with "I cant teach you to be a good trader but heres some advice blah blah blah"
Not sure what the exact point of your thread is other than trading in general itself so ill answer the questions i know from experience
>What is the minimum capital to start?
Depends on 2 variables. You and your Broker if they have a minimum deposit. but with your numbers you should be just fine just about anywhere
>how different is simulated trading from live trading?
Quite diffrent with non comercial spending, Like when you invest 3k in a simulation It usually makes the market go up, Not the case in real life. If you want a REAL hands on experience to get used to how the markets really are id reccomend finding some forex shit with a 1:1 leverage
>is this a jew scam?
The books are the scam, The Algo trading is real. all big firms usually have a computer doing trades for them.
Why? a computer can do 10 trades per second where as a Human can do like 1 or 2 a minute.

I may be slightly confused at what you want to know so if you got any questions ill stick around for a bit.

>> No.933985

>>933973
Of course I would be backtesting and everything before I deploy. If you don't recommend books, what would you recommend for learning how to analyze stocks?
3.5k is enough to get me anywhere? It seems like a low capital to start with. I'm just scared that I'll need a higher capital to start (although I don't plan on starting now and when I do deploy a live algorithm I'll probably have around 5k to use).

>>933963
An edge is an algorithm that makes more on average than losses, right? like 50% on a .25 profit and 10% on .4 loss come out to an edge of .085?

>> No.933990

>>933985
Nah generally youre totally fine with anything depending on what your trading that is, Most people reccomend for starting on day trading manually anywhere from 250$-1000$
> If you don't recommend books, what would you recommend for learning how to analyze stocks?
What I Did before i could afford to even step into the market was I learned my indicators. I would download a platform set them all out and try to gauge where i thought the stock was going to go. and when i was Right more times than i was wrong I started to buy in.

Only time i created algo trading was with coinbase.com because i thought i would just leave my computer on with a script running and it would generate income for me.

>> No.934059

>>933973
>Quite diffrent with non comercial spending, Like when you invest 3k in a simulation It usually makes the market go up, Not the case in real life.

that isn't the difference - the sim should show the prices of the live market and your trade should have no effect at all in a sim

main difference is you don't get an accurate simulation of the order book so you'll end up getting filled orders when you wouldn't have been filled and vice versa

>> No.934060

>>933973
>Why? a computer can do 10 trades per second where as a Human can do like 1 or 2 a minute.

a human can be in an out within a second or two if they want...

a computer can do rather more than 10 a second

>> No.934067

>>934060
was just laying down an example. Normally this doesnt happen though as buying at market carries heavy fees on small gainers. Where as limit buying does not.

>> No.934071

I think OP misunderstood this idea. Large hedge funds an proper algo trading companies have super computers and their internet is linked to the stock exchange centre. So there is minimum pip slippage etc. Thus, 3k isn't enough for stocks at all...

But there is one thing to start with ---- python machine learning. Have fun.

>> No.934072

i do want to know more about algo trading. i do well with options but it's obvious 99% of the buy/sell orders are algos dicking around for scalps. i need to learn how it's done because it clearly affects my trading, but i don't have much time

>> No.934075

Moreover, there are certain basic concepts that you need to have full understanding of. Like indicators and buying/selling strategies. Which means that the preparation in this project of building an algo robot is just long...

>> No.934076

>>934071
>their internet is linked to the stock exchange centre

no - they don't use the internet, they collocate their servers in the same data center as the exchange's servers and have a direct connection to them

if the strategy requires instruments traded on more than one exchange then again the internet isn't used if they need speed - fastest route is usually via privately owned microwave towers

>> No.934077

The real algo requires supercomputers.... For Barclays, they have the entire floor stacking up computers of pizza box size and there are 50,000 of them. You sure your algo programme can outperform theirs?

Btw, is OP new to trading?

>> No.934081

his algo doesn't have to outperform theirs

presumably he isn't going to be trying marketmaking in highly liquid instruments

there are plenty of potential edges out there some of them don't scale well - no need to compete with the big boys

>> No.934085

>>934077
I'm not looking to cuck the big guys. I just want to make a small profit whole learning both machine learning and the market

>> No.934097

>>934085
Best way to learn the market is like i said go 1:1 trade and loose and win. Its a low risk low reward .
1$=1$

Trading you have to treat like a business you will loose some you will gain some.
Trust no one, Not even me. Its something you have to experience for yourself.

>> No.934108
File: 488 KB, 600x450, 1442990517210.png [View same] [iqdb] [saucenao] [google]
934108

>>933954
Yeah been doing it for a little over five years now.

>>minimum?
Your only option for a broker with an API is Interactive Brokers. Their account minimum is 10k and the minimum for Day Trading is 25k. So, practically speaking, the minimum to algo trade is around 30k.

>>how different is simulation
Much different. You'll want to create your own simulations that live up to your own standards of realism.

>Merchant schemes.
Most of the books on the subject are garbage, but this is a very real thing for those who want to figure it out bad enough.

>> No.934174

>>934108
What do you algo trade? (Forex, Commodities, Stocks, etc)

Most people seem to talk about Forex with this stuff, but isn't Forex a pretty crowded market in terms of algos?

>> No.934179

>>934174
I only trade stocks for now. I don't know anything about ForEx.

>> No.934181

>>934179
its time u learn u bastard.

>> No.934197

>>934181
Yeah, I know. It's on the todo list.

>> No.934204

>>934197
America?

>> No.934208

>>934179
Yeah, I've been interested in algo trading recently but I'm more interested in using it to trade stocks as well.

What platform do you use? I've been looking around, but it seems like the main platforms out there aren't especially great. It seems like they only offer simple back testing and no walk-forward stuff that would be important to actually understanding real time performance.

>> No.934264

>>934108
do you use interactive brokers to algo trade? how would i go from good-at-excel learning to writing something that can buy and sell on its own?

>> No.934615

>>934204
I don't know. I've never been outside.

>>934208
I couldn't find a platform I liked either, so I built one from scratch.

>>934264
Yeah, IB. Learn some programming.

>> No.934780

>>934615
I see, that's a bit disheartening. I have mild programming experience (I wouldn't be worried about programming any scripts), but I'm not sure I'd have the expertise to build a whole platform. Thanks for the info, something to think about.

What's your process for coming up with new strategies?

>> No.934981
File: 61 KB, 800x533, strategy development cycles.png [View same] [iqdb] [saucenao] [google]
934981

>>934780
If you're not willing to put years of your life into it then algo trading isn't for you.
Trust me when I say that building a trading platform is the easiest part.
Discovering your first few reliable strategies is where most people give up.

> my strategy discovery process
Observe some ideal stocks very carefully.
Identify a particular behavior or set of behaviors that you want to exploit.
Build the math to exploit it.
Test it, tweak it, deploy it or trash it.
Rinse, Repeat.

>> No.935009

>>934615
>Yeah, IB. Learn some programming.
i need to be spoonfed more. am i supposed to study C++, Perl? i've done FORTRAN years ago

i just want to be able to get a algo running that will react to data from my IB practice account and open/close trades on it's own

>> No.935033

>>933954
you're 20 years too late son.
the big boys have been on it for a while.
there is no (mathematical) edge to play on anymore.

>> No.935066

>>934780
you don't have to build an entire platform yourself - for a starters there are various open source projects out there, plenty of code examples for interactive brokers API etc... just adapt stuff to what you need

obviously you're not going to be trying anything that relies on low latency with this sort of approach but there are lots of approaches you can take and not everything has to be ultra high frequency

find an edge that doesn't scale well perhaps

or maybe just look at something like trend following which doesn't necessarily require low latency or special hardware/infrastructure

but regardless - no one is going to hold your hand for you and it is still competitive, don't expect a free lunch

>> No.935122

>>934981
Thanks for all the advice, it's been helpful.

>>935066
Yep, I'll just need to do more research on platforms. I'm not opposed to taking the time to figure out how to build/modify something for my needs.

Not really interested in competing with big dog HFT guys, it definitely seems like there's plenty of non-HFT room for now.

You're right, this stuff is competitive and complex, definitely don't expect a free lunch.