AI prediction/forecasting for Bitcoin and crypto trends

[nites]

Hi,

I reserach and extensively use AI and machine learning for trends prediction in bitcoin and other crypto's as well.
The model is built using cutting edge algorithms and it is capable of predicting 5 days ahead with more than 90% accuracy.
On top of prediction, almost all trading strategies and risk managment would be improved significantly.

Would you in the sea of dishonest vendors selling renamed RSI, be interested in buying real AI predictions?
Here's photo of testing chart for the last few months:



EDIT:
One more photo where model is trained for daytrading, predicting next day in advance:

[1715479127]

1715479127

[1715479127]

1715479127

[1715479127]

1715479127

[1715479127]

1715479127

[1715479127]

1715479127

[palle11]

The chart on Imgur only looked like RSI or MA and therefore it won't be easy to trace to a particular thing you are trying to explain, no arrow or other indicators to direct to your focus. Well I clicked because it is Imgur but you can as well put out the chart here instead of people clicking on the link because there is no proper information there.

Edited: Oh you right on your rank for now. 

[nites]

Hi palle11,

I've tried to append images to the post, using correct bb code, but it doesn't work. Seems there is some restriction for new accounts.

Let me clarify. This is not an indicator, like moving average, RSI or any other lagging indicator.
What you see on chart is 'prediction' for last 100 days.
Model was trained, and last 100 days were used as test for verification of performance - orange line is prediction and blue one is actual price.

It is predictive model that outputs forecast for x amount of days into the future, it doesn't smooth and recalculate price like moving average, etc.

[Bitcoin_Arena]

Interesting.

You know what would be more believable?

A sneak peek of the forecast for the Bitcoin trends in the coming days, and then we wait and watch how the market performs maybe fore the next hours or days.

Right now, it's hard for one to believe this. You know hoe easy it is to draw lines based on the current trend.

Well I clicked because it is Imgur but you can as well put out the chart here instead or people clicking on the link.

He has a newbie rank. He can't post images until her ranks up

I've tried to append images to the post, using correct bb code, but it doesn't work. Seems there is some restriction for new accounts.

True

[nites]

A sneak peek of the forecast for the Bitcoin trends in the coming days, and then we wait and watch how the market performs maybe fore the next hours or days.

Currently I'm running training to obtain prediction for the next 5 days. It takes a lot of resources and it will be finished in approx. 1 hour.
Will post predictions here after it is done. Currently, there is next day prediction and 5 days ahead.

To get predictions for hours or minutes, it takes 50x amount of time as there is more data for training and processing. I will have to update my GPU to multiple TITAN RTX GPUs to do that.

Right now, it's hard for one to believe this. You know hoe easy it is to draw lines based on the current trend.

Perfectly understandable. One of the reasons I started thread.
It just baffles me that people are renaming and reselling free indicators for thousands of dollars, not to mention selling phony signals created using strategy from 1st Google page.

[sheenshane]

It catches my attention and wanted to know how accurate your AI is for prediction.

We know the price was still predictable or what cause the trend.  So I'd like you to create a separate thread and give a regular update and we will see the result if it's consistent at 90% accurate result.

Possible I will believe if you will have show more proof on it.

[nites]

Finally done,

here's the forecast for the next 4 days.

19937.01 19381.303 20040.582 20372.076

So basically, price will drop today(depending in what time zone you are), and recover a little bit in the last two days of prediction.
Weekend is coming, so there will not be much volatility, next good forecast will be done on Sunday for the following week.

[Bitcoin_Arena]

It just baffles me that people are renaming and reselling free indicators for thousands of dollars, not to mention selling phony signals created using strategy from 1st Google page.

True, telegram is like a den for all sorts of trading signal scammers. I am glad I took the hard route and learnt how to trade for myself.
I just get curious sometimes to see if there are any trading strategies out there that are actually so good... perhaps with at least 70% win rate.

Finally done,

here's the forecast for the next 4 days.

19937.01
19381.303
20040.582
20372.076

So basically, price will drop today and tommorrow(depending in what time zone you are), and recover a little bit in the last two days of prediction.
Weekend is coming, so there will not be much volatility, next good forecast will be done on Sunday for the following week.

Thanks,

So the prices above. Are they like the average trading price for BTC per day or there's a specific time the Bitcoin trading price is expected to be approximately close to the stated price?

For example today, starting from the time you posted, Bitcoin price has been roaming between $20,380(lowest so far) and$ 21,200 (highest so far)

[Tytanowy Janusz]

The model is built using cutting edge algorithms and it is capable of predicting 5 days ahead with more than 90% accuracy.

With such accuracy you dont need to sell your product to become very rich very fast. In fact selling this product seams irractional and raises suspicions. I coded HFT market maker by myself that was giving me decent profits. last think i would do was selling it.

here's the forecast for the next 4 days.

19937.01 19381.303 20040.582 20372.076

I doubt that this is the 10% of failed predictions because it failed hard. Market, so far, went in oposit direction.

[Husires]

If you succeed in developing such a tool, you will be rich in a short time. All you have to do is take a loan and then buy and sell, and since you will know the best time to buy and sell during the next four days, day trading for you will be very profitable.

• As for trying to sell such a service, you will find it difficult to do so in the forum.
• Post constant price updates prediction and gain trust, and then you may raise money.

[CryptoHeadlineNews]

If you succeed in developing such a tool, you will be rich in a short time. All you have to do is take a loan and then buy and sell, and since you will know the best time to buy and sell during the next four days, day trading for you will be very profitable.

That will actually be a brilliant idea. Taking a loan, buy and sell to make lots of profits before the loan pay back day.
But If actually this A.I has a success percentage rate of 40% to 60% then I will love to ask, please where have you been since last month from 10th to 16th June 2022 when the price of Bitcoin starting falling from $30k to $20k, because I think if this A.I tool was close to being accurate, that would have been the best time for you to have shot your shut, because am sure by now people from all around the world would have been looking for you for your A.I tool, but however, a prediction sometimes may not be accurate and that's why it's call prediction, but moreover I wish you the best as you work more on your project and come to give us update anytime soon

[hyudien]

Finally done,

here's the forecast for the next 4 days.

19937.01 19381.303 20040.582 20372.076

If you refer to the date you posted and now it's been 3 days after that, what I see is all prices have been breached and you can say yes it worked. But have you made a profit or tried to predict it again?

I admit the sophistication of AI makes it easier for us to do things beyond the reach of market algorithm movements. AI can source resources in a short period of time and the results can be compared and then come up with an almost correct price prediction. But believe me, a natural market can change predictions that may not be on target.

[nites]

Thanks everyone for questions and replies, if I miss to answer something, let me know.

I've made a mistake with predictions I posted. I've found there is some error in the code and function for future prediction is not working properly.
Basically what happened, model predicted the test data, not the future, so technically it was not a prediction.
Currently, I'm able just to post next day forecast, until I fix for the multi days function.

Take a look, predictions I posted and real price for the 4 predicted days; 100% correlation.

19937.01                              
19381.30
20040.58
20372.07

19970.56
19323.91
20212.07
20569.92

So the prices above. Are they like the average trading price for BTC per day or there's a specific time the Bitcoin trading price is expected to be approximately close to the stated price?

After training and predictions are made, it is converted back to prices for easier reading and interpretation. By no means it is predicting exact price, it will sometimes be very close, sometimes not.
What we are interested as traders is 'trend', hence I'm looking if the price will up or down.

With such accuracy you dont need to sell your product to become very rich very fast. In fact selling this product seams irractional and raises suspicions. I coded HFT market maker by myself that was giving me decent profits. last think i would do was selling it.

Well, yes and no. To become rich with trading you need capital, either your personal or seeded one from investors. I'm trading my own money right now, as I have no guts to trade other people money.
Frankly, I would never sell the system; code and logic behind it, but information in sort of forecasts - why not?

If you succeed in developing such a tool, you will be rich in a short time. All you have to do is take a loan and then buy and sell, and since you will know the best time to buy and sell during the next four days, day trading for you will be very profitable.

That sounds very simple, but is far away from reality. Even if the forecasts are good, there is yet to build trading strategy and risk management framework for very profitable trading.
What do you mean by loan, raising money from investors or taking the loan from the banks?

As for trying to sell such a service, you will find it difficult to do so in the forum.
Post constant price updates prediction and gain trust, and then you may raise money.

I'm not selling anything, yet. I started thread to see opinions and sentiment on the topic - if there is good interest and I manage to provide good forecasts, then it is mutual benefit.
Yes, I'll post regular forecasts, when I finish fixing sequence prediction = more than one day into the future.

That will actually be a brilliant idea. Taking a loan, buy and sell to make lots of profits before the loan pay back day.
But If actually this A.I has a success percentage rate of 40% to 60% then I will love to ask, please where have you been since last month from 10th to 16th June 2022 when the price of Bitcoin starting falling from $30k to $20k, because I think if this A.I tool was close to being accurate, that would have been the best time for you to have shot your shut, because am sure by now people from all around the world would have been looking for you for your A.I tool, but however, a prediction sometimes may not be accurate and that's why it's call prediction, but moreover I wish you the best as you work more on your project and come to give us update anytime soon

Loan from the crooked banks is a big no. Raising money from investors, maybe.
Well, sadly I was developing the system at the time of the fall. Would also like to test the performance in that period.

[Danydee72]

For me this sounds to be an interesting project. But I have one problem with the numbers. Before I can rely on numbers or formulas, I need to have exact values and know what they stand for. No cleaning, rounding or whatever people do to make them "easier" for reading. So my question still remains: These bitcoin-rates you mention, are this averages or just the values taken at a certain time? Or do they stand for a trend?
For me a correlation cannot be made without a little bit more information about these numbers. One may follow the principle: "Don't trust any statistics you haven't falsified yourself".

[nites]

For me this sounds to be an interesting project. But I have one problem with the numbers. Before I can rely on numbers or formulas, I need to have exact values and know what they stand for. No cleaning, rounding or whatever people do to make them "easier" for reading. So my question still remains: These bitcoin-rates you mention, are this averages or just the values taken at a certain time? Or do they stand for a trend?
For me a correlation cannot be made without a little bit more information about these numbers. One may follow the principle: "Don't trust any statistics you haven't falsified yourself".

Hi, thanks for the interest. Yes, when the predictions are done, they are converted back to prices. They may be close to actual values, but they stand as the trend.
So for the 100% correlation, I meant trend, or directional accuracy prediction, not exact price values.

Like that quote.

[Husires]

Hi, thanks for the interest. Yes, when the predictions are done, they are converted back to prices. They may be close to actual values, but they stand as the trend.
So for the 100% correlation, I meant trend, or directional accuracy prediction, not exact price values.

Like that quote.

I am interested in artificial intelligence/image processing, can you post more information about how you work? Where do you get the data and how is the comparison made? Are you comparing prices? blueprints? Pictures, graphs, or in short how to get the results.
Do you intend to create a site that displays forecast data (provided that the data is not modified after its time "Hashing") and then, after two or three months, we can compare the extent of the error.

[nites]

I am interested in artificial intelligence/image processing, can you post more information about how you work? Where do you get the data and how is the comparison made? Are you comparing prices? blueprints? Pictures, graphs, or in short how to get the results. Do you intend to create a site that displays forecast data (provided that the data is not modified after its time "Hashing") and then, after two or three months, we can compare the extent of the error.

I get standard OHLCV data from brokers or data vendors. In most cases it's free and clean.
I can't tell you much details, but image processing and in particular speech recognition problems are quite similar to financial markets.
If you are going to try to develop AI system for markets, take a look at speech recogntion algorithms, I'm quite sure that's the way to follow.

Not sure about creating site, but will post forecasts here, so we can evaluate performance.

[crwth]

Upon checking your post, it seems like there's no prediction there. I'm expecting something like this for your results or something.



See the yellow line? I drew that as somewhat a prediction of the price or something. If you can show that and it works, that's definitely something. I know an indicator that does that too, the Ichimoku^[1]

It looks like this



It predicts the future or something like that. Maybe you could do that as well or something?


Reference:
[1] - https://www.investopedia.com/terms/i/ichimoku-cloud.asp

[nites]

Yes, that is just point of visualisation. Currently I can only do it for 1 day in advance until I fix recursive function for forecasting multiple days into the future.

For example, one day prediction for today:

[nites]

Okay let's test next day forecast for some time, I'll post daily predictions for the next day and make table to track the results. Will post in form of screenshot, so it is not editable after the fact.
Results will probably be worse than multi day prediction, but let's test it in real time.

Prediction for tomorrow 19.07.2022 is upwards.