AI & Machine Learning

[Joe]

Artificial Intelligence and Machine Learning are hot topics these days. They are going to fundamentally change our lives in many ways in the future. Somebody asked the question on Quora "Is it possible to use machine learning/artificial intelligence to win at roulette?" and there were some interesting replies.

link:s://:.quora.com/Is-it-possible-to-use-machine-learning-artificial-intelligence-to-win-at-roulette

The consensus seems to be that it can be done. AI doesn't seem to have been explored much here, I suppose because it's quite a technical subject, but if anyone's prepared to put the effort in I think the gold is there.

[Steve]

AI would eventually determine how variables correlate to spin outcomes, then use related data to make predictions. Thats what cross referencing is. I have it already working.

Ive done the same thing with a trading bot, for stocks and cryptocurrency. It looks at all the main technical indicators, and some of my own indicators, then simulates with diferent variables to see what algorithms are most profitable. But without modelling cause and effect, its just curve fitting. It already works, but im still refining it.

I would post a screenshot but im not at my pc.

[Steve]

Its actually simple in principle... cause and effect. The variables are the cause, the effect is the winning number.

Once it's understood, people might understand how the typical roulette systems use bet selection related to thin air. These are the same people that think logic, cause and effect (AP) is the nonsense.

Money talks, bullshit walks. Ive been doing it a long time and know what works rather well.

[Joe]

It looks at all the main technical indicators, and some of my own indicators, then simulates with diferent variables to see what algorithms are most profitable. But without modelling cause and effect, its just curve fitting. It already works, but im still refining it.

But you can't tell me that you know the cause of a stock going up or down. Technical indicators don't tell you what the causes are, but traders use them all the time.  Machine learning is really just sophisticated curve-fitting.

link:s://diginomica.com/2018/06/04/ai-curve-fitting-not-intelligence/

Correlation is not causation, but in order to make predictions you don't always need to know the causes, you just need to know how variables correlate, which is often easier than identifying the true causes.

link:s://statisticalhorizons.com/prediction-vs-causation-in-regression-analysis

[luckyfella]

I wrote about this long time. Instead of exploring it ppl go bla, bla, bla......usual bs

[luckyfella]

If you happen to be inspired enough to go build your own predictive regression model be prepared to waste 10years of your life playing with big data and mostly end up with nothing much.

[Joe]

Regression is just an example, and it's a very basic form of ML, some wouldn't even call it ML because it's been around for over a century. There are dozens more algorithms to try, including deep learning. And yes, it could all be for nothing, but nothing ventured, nothing gained. It's a different approach, a more sophisticated way of curve fitting than most systems here. 

The trick is to find the right variables. BTW, did you write about this on the forum? If so, do you have a link?

Currently there's a challenge on kaggle.com to find a way to predict random numbers.

link:s://:.kaggle.com/c/random-number-grand-challenge

[Joe]

Currently there's a challenge on kaggle.com to find a way to predict random numbers.

Correction, the challenge is finished. See? already been done. ;-)

[precogmiles]

Correction, the challenge is finished. See? already been done. ;-)

That was an Aprils fools joke. Read the rules

Start Date: 3/31/2014 11:59 PM UTC
End Date: 4/1/2014 11:59 PM UTC


You can not use AI to predict absolute random.

[Joe]

That was an Aprils fools joke. Read the rules

Really?

link:s://:.kaggle.com/c/random-number-grand-challenge/discussion/7586

Pseudorandom number generators are not absolute random. And what does that mean anyway?
Obviously if you define random as unpredictable then it's a waste of time trying to predict random numbers, but that's only one view of what random numbers are. IMO the term is meaningless.

[precogmiles]

Really?

link:s://:.kaggle.com/c/random-number-grand-challenge/discussion/7586

Pseudorandom number generators are not absolute random. And what does that mean anyway?
Obviously if you define random as unpredictable then it's a waste of time trying to predict random numbers, but that's only one view of what random numbers are. IMO the term is meaningless.

OK, believe what you will. I hope you succeed in using AI to predict absolute random. Good luck and best of wishes.

[Joe]

I can't see the logic of not believing AI & ML can be of use for winning at roulette and yet believing that precognition can. Shouldn't it be the other way around?

Anyway, best of luck to you too with the whole "effects can happen before their cause" thing. 

[precogmiles]

I can't see the logic of not believing AI & ML can be of use for winning at roulette and yet believing that precognition can. Shouldn't it be the other way around?

Anyway, best of luck to you too with the whole "effects can happen before their cause" thing. 

AI & ML can definitely beat roulette. It just can not predict absolute random as the kaggle competition was attempting to do. That is my opinion, I was not trying to be sarcastic, but I wish you good luck. Stranger things have happened, like developing precog abilities.

[Steve]

But you can't tell me that you know the cause of a stock going up or down. Technical indicators don't tell you what the causes are, but traders use them all the time.  Machine learning is really just sophisticated curve-fitting.

Partially true. Most traders follow the indicators, which is their cause for buying or selling. So indicators are directly related to cause.

Another cause we have access to are breaking news, but it needs to be very positive news.

Probably the most direct cause is the intentions of bull and bear whales. You can see when they're trying to influence the price because the market floods with lots of small orders through the price spread. It's market manipulation and should be illegal, but it happens constantly. If you watch for this manipulation you can act on it and profit.

but in order to make predictions you don't always need to know the causes, you just need to know how variables correlate, which is often easier than identifying the true causes.

In roulette we do have access to many direct variables which are the cause. Understanding them makes analysis much more efficient. That's why the permutations approach takes much longer.

I can't see the logic of not believing AI & ML can be of use for winning at roulette and yet believing that precognition can. Shouldn't it be the other way around?
Anyway, best of luck to you too with the whole "effects can happen before their cause" thing.

It's a bit of a guess of how precognition works. I believe it's similar to considering the beginning of a wave in an ocean. The start of a wave gives clues to what happens next. At least its a simplistic example of what may happen on a much larger scale. Alternatively, time is an illusion, and everything actually happens in the present, so on some level there is no separation between timing of events. This is a common theory often discussed in quantum mechanics. But ultimately they are theories. If statistical results dont support the viability of precognition, it probably doesnt have merit. But data and my own experiences do support it. So its not just a theory alone to me. It's more like likely fact, although the mechanics of it all are not understood yet. I'm a reasonable person and dont believe in hocus pocus. Again I believe there's clearly enough evidence to support it being viable. There are countless things we dont know about the universe.

[luckyfella]

Work of guys who have done it in the trading world. The same can be done for gambling games.
Ps. Less talk more real output pls