/ag/ algotrading general

I'm just curious to hear some opinions on why people think that python is "slow". Is it even really that much of a deal? What causes slowness? Compilers?

Also, why is speed such a big issue, unless of course you're performing high frequency trading, in which case the language you use might be of consideration, but even then, hardware issues, the efficiency of an algorithm are, and location distance from exchanges are factors that would matter far more.

Also, If people could perhaps gives some good tips for technical indicators/ quantitative formulas they use, that'd be cool.

File: 1494504733873.jpg (116KB, 781x767px) Image search: [Google]

116KB, 781x767px

Fuck off none of us will give you any advice. Figure it out like the rest.
>>2966564
>>2966611

>>2966616
It's well within your interest to have others do the same thing you do. It makes what you do more effective.

>>2966629
(excluding arbitrage)

I am going to implement an LSTM model for trading small cap stocks that institutions don't bother with. It will probably not work.

>>2966611
I can't really see the type of computations being done having their performance impacted so much between a lower level language and Python, also, like you said if you're doing anything like HFT the latency between you and whatever server would be way more important than whatever language you're doing. If you're on the west coast and your exchange's server is in the east coast that time delay is going to be measured in milliseconds verse the nanosecond difference of what language you're doing.

I don't think quants use HFT, I think HFT is mostly something that was used to just get a speed edge on people and kind of "see the future" before others.

I think the most basic quant shit is probably standard deviations of a price and just being statistically more right than you are wrong with it.
Eg if you see a something is trading below it's -1 std deviation you can make a guess that'll probably rebound back within it just due to the fact most action is going to take place within 1SD, sometimes it'll be wrong, but it should be correct most of the time.

>>2966611

Python is slow?? Fuck??? I use VB5 it's twenty fucking years old.

But it works, and I don't.

It's set up to trade on the ASX (I'm in Australia), we don't have fast highly liquid markets here other than CBA and BHP so don't need a HiFreak megacock 8 screen triple anus setup.

Is the 5 day MA > than the 10 day MA?
Is the RSI above 60?
Has the 50 day crossed above or below the 200 day?

How the fuck should I know? Well look at the chart numballs! o wait, I don't, it's algo'd.

The mysteries of HFT, big computers, fast connections, quant math guys, fancy formulas, the hype, the fear, it's out of YOUR league little guy, fuck that shit, you got a computer! you got an ass to sit on! use it!

Everyone laughed at me when I was starting out.

>>2967209
I disagree about python being slow, like mentioned here >>2966842 the stuff in trading isn't really going to be overly intensive (as far as I know). And yeah, hft is beyond what the average trader can compete with. Htfers do compete with each other, however.

"Python is slow" is a meme parroted by people who can't explain how that'll seriously effect them.

Many of the expensive operations in quantitative work SHOULD be handled by numpy/pandas libraries which are highly optimised and run fast enough.

Python is fine. If you're being bottle necked by pythons speed, you're doing something wrong or you shouldn't be posting here.

>>2966611
>Compilers?

Woops meant to say 'Interpreters', must've been thinking about Java at the time.