Is AI trading effective?

[Waylen]

People are frequently talking about AI trading. What’s your view on AI trading as I don’t have trust in this sort of trading.

 

[ktm]

AI is only as good as the folks coding it.

 

[Q.E.D.]

AI has been around for > 30 years, and to one extent or another, used by Wall Street houses for many years.

It is only "Generative" AI that is but a few years old.

 

[Pekelo]

Effective? Yes.

Profitable? No.

Joking aside, if AI was watching Roaring Kitty's tweets, it may have bought CHWY. If you a smart guy was watching the rally, you may have bought a bunch of puts, because they never last.

 

[d08]

Effective? Yes.

Profitable? No.

Joking aside, if AI was watching Roaring Kitty's tweets, it may have bought CHWY. If you a smart guy was watching the rally, you may have bought a bunch of puts, because they never last.

AI or rather ML is trained on features and results. Not just the one side.

 

[Pekelo]

AI or rather ML is trained on features and results.

What I wonder is, if AI is smart enough to recognize a bullshit rally and can short it?

 

[d08]

What I wonder is, if AI is smart enough to recognize a bullshit rally and can short it?

It depends on the features, the inputs you give it and if they have predictive value. So if a human can notice a pattern, the AI can most probably do the same.

 

[Pekelo]

So if a human can notice a pattern, the AI can most probably do the same.

I understand. But can an AI realize that a posted meme is not a fundamental (or good enough) reason for a stock to rally? Maybe so, thus the price moved back in a few hours.

On the other hand TSLA'a price moves on any BS Musk posts, and doesn't always retract.

 

[Businessman]

People are frequently talking about AI trading. What’s your view on AI trading as I don’t have trust in this sort of trading.

Yes AI is highly effective for trade prediction, i have developed a neural net model that can predict prices with 95% accuracy.

I give it the last 30 bars and the model will tell me if the next bar will be up or down with 95% accuracy

 

[Snuskpelle]

What data you feed into your model is generally far more interesting than the model itself.

You *can* solve the problem of mapping input data to trading behaviors with "AI" (super wide term), among other tools, but because AI is the cool word right now plenty of people are likely throwing it around for no reason. Or the opposite - specifically because they're using it to market crap - when something has AI in it it must be good, right?