[ 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.

/sci/ - Science & Math


View post   

File: 26 KB, 474x379, th-1136037158.jpg [View same] [iqdb] [saucenao] [google]
14713528 No.14713528 [Reply] [Original]

I trained an AI to trade stocks and crypto and got a nice 50% accuracy.
What would your approach be? I used LSTM cells and the data from the past 160 minutes with 1 minute intervals to predict 20 minutes into the future.
Also leave some research done on this topic below.
Skip the articles from mainstream media and medium. Those guys are retarted and only copy each other's shitty code.

>> No.14713532

>>14713528
>a nice 50% accuracy.
Save some energy and do a coin flip?

>> No.14713536

>>14713532
No shit. You'd think anons on /sci/ wouldn't be this slow

>> No.14713880

>>14713528
The shorter the timespan the more difficult predictions are .

>> No.14713993

>>14713528
I find it so hard to believe this "Quantitative Trading Hedge funds" like RenTech and Two Sigma can be as successful as they are for as long as they have been around. It has to be insider trading with the quant shit being smoke and mirrors right? Or am I missing something here?

>> No.14714021
File: 48 KB, 190x263, groo.jpg [View same] [iqdb] [saucenao] [google]
14714021

>>14713528
>/sci/ plz help me get money for free
>no i will not offer you a percentage of my gainz

>> No.14714036

>>14713528
>I trained an AI to trad-
stop right there
you are assuming the market follows anything.
the last year has been irrational random shit.
spikes on bad news, dips on good news, and the opposite as well.
you can never predict this shit because it doesnt make sense and is manipulated at times

>> No.14714040

the post above me is the best one itt /thread

>> No.14714063

>>14713528
i'm assuming you're using ai to predict price
so my approach would be not to use ai to predict price, instead use it to predict whether your strategy will do well given current market conditions
every hedge fund is using ml to predict prices and others know that so they'll find ways to mess with it

>> No.14714284

>>14713993
they have massive amounts of assets so when they do trade they can easily make money
the algorithms are good enough but they aren't a silver bullet
the companies are likely more asset shells for the holders in them
I know some students who intern with companies like that and they are just rich person shells so their science kids can have assets transferred to them for doing nothing
Someone else might have a better idea on what goes on with them

>> No.14714307

>>14714036
A sufficiently trained AI will still be able to see the pattern. They're actually quite good at sussing out hidden variables that we have no fucking clue about.