Topic: AI prediction/forecasting for Bitcoin and crypto trends (Read 843 times)

I'm getting quite sure that it's not possible to get it accepted here, because most of the people here don't understand and they can't judge what is good or what is not. I glanced throught threads in this trading discussion and it's astonishing that most people on this forum think you can beat market consistently with indicators and price action if you practice/learn enough. Total ignorance.  
There is a reason why 95% people in trading lose money, and it will always be the same.

Doubt is healthy, I'm also doubtful and suspiscious about everything, that significantly influenced me and lead to creating such a system. Otherwise, I would still trade indicators and supp and resist nonsense.
Trading bots are basically sets of rules with classic indicators, and they will obviously fail.

There is difference between predictions and trading signals. Yes, you can maybe make few good trades/signals in a row, but predicting correctly daily close direction is much harder.
Currently, system is running next day prediction, and I'm working right now to make it predict e.g. 5 days as you suggested.

I have to completely disagree that you need to observe it for a year to evaluate it's performance. Especially when the system is built on strong mathematics fundaments.
Not to mention you are basically losing money for that year observing it.

How I personally trade predictions: I buy or sell half a contract if prediction is upwards or downwards, and if price pullbacks in opposite direction I add half contract more.
It is highly effective and profitable, yet very simple, but that's only because predictions are most of the time correct.

I doubt it but maybe you did an amazing invention. In order to prove that your AI is effective, we need to observe it for about a year maybe. Because short-term movements are not that hard to predict. It's shown in your images that the AI is not perfectly accurate but its trend is correct. What is the time frame of your chart mate? Can you show us the movement prediction for the next five days? Also, it will be easier to understand if you can make its movements into a candlestick. Maybe some of us are willing to make trade positions out of it if you are also willing to share the next 5-day movement.

I got some friends last year that made thousands of dollars daily using bot trading but it didn't last long. So maybe my friend's setting was doing well in that bullish period but it became ineffective when bitcoin started its correction.

It took me 2 years to build, but for research and learning how to do it, more than 5 years. It was not meant to impress anyone.
Most of people misunderstood chart on the first post, believeing it is some indicator or whatnot, but it is prediction test.

I have already showed it works better than anything seen on this forum. Predicting 14 days in a row correctly is extremely low chance event. It can't be luck. Probability of this happening is 0.00006104.
Rather in fact, if someone tried to recreate it again, you would have 0.0061% chance of success in doing so.

Giving it for free is just not doable. Not because it took me lot of time to get here, but because running the system everyday takes enormous computing resources.
And of course, I need money to develop it further with more features and invest in hardware which costs hell of a lot.

Maybe he thinks that saying he built the project for 2 years is already enough to impress anyone in the room and then won't question it's credibility plus he drop some screenies there on the first page. Usually screenshots are the most used evidence when someone propose something interesting here in the forum.

If let say his invention really works and it can predict better more than the others, well I don't think he will just use it and gave prediction here for free because he work hard for it and also other people will now depend on it without them experimenting and risking money on their own. That's why he is trying to sell it for some cash.

Nites, Let the nay sayers and critics opinion go like “water off a ducks back”.
Keep in mind, especially when trying to share something and get it accepted on a forum, “Winners never quit and Quitters never win”.
Next prediction, please.

But what evidence? Please, clarify what is credible evidence? Where is legitimate review? Only credible evidence is whether predictions work or not, and I'm showing that in most transparent way.

There is no more simpler way to post predictions, prediction is upwards or downwards, price can go upwards or downwards, same for evaulating results, true or false.

I don't want to give up, but people like stompix blatantly lying and twisting words, without any knowledge in basic maths and probability is loudest mouth on the forum.
It's insulting to people that use brains in different ways than just writing nonsense on forum, in this case; creating AI prediction system.

you said you worked hard for 2 years so you arrive on this forum and don't show more credible evidence and at the first legitimate review you gave up and left? imagine that you arrive at a company to show your project, do you think that the company will not have questions and criticisms? how will you deal with it? giving up or bringing credible evidence?

just post your predictions in a way that everyone can understand and analyze

Thanks to @stompix, I decided to lock the thread. As there is no point explaining or providing something smart and sophisticated to people that don't believe it or remotely understand it.
I had sincere intentions to help and make something good, but once again in this world, ignorance wins. Kudos to people that were open minded and were trying to lead intelligent discussion.

Well, farewell and good luck, you will need it more than anything.

Why would anyone take a loan and then trade it, it's complete nonsense. I don't want to take loan from people here or raise money.

I wanted to sell predictions which can help thousands of people become profitable and in return I would earn small fees from broker. No one ever posted predictions in advance and no one was ever close to get 14 predictions in a row.

But seems that people will rather believe in hamster trader or technical and fundamental junk and trust people like @stompix that don't know anything except to write nonsense on forum.

From my perspective i will not encourage to obtain loan and accomplish his project, rather i will advice that taking loans is not the major concerns, because it might taken a loan and after developing the program or delivered the project and it fails to function the way anticipated. So the best thing to do is to develop it with your capital irrespective how long it might taken, but develop with your fund without debiting yourself.

I'm again baffled with your comment and your ability to blatantly lie and ignore facts.

You are making yourself extremely ignorant with this comments, I would ask you why, but you might lie and pretend again.

900 out of 1000 just to be lucky following a trend? What on earth are your process of reasoning, I will suggest you learn some probability before talking about numbers like complete ignorant fool.

- you probably don't know, as that is expected from you from this point; but most successful hedge funds like RenTec employs speech recognition to predict price, their CEO literally said that in quote, Google a bit maybe.
-try to build any system with TA, AI or regular one and u will not predict direction of closing candle 3 times in a row, and it is literally impossible to do it 14 in a row like I did
-again you showing extreme levels of ignorance about numbers and probability, I made thousands, but not starting with 10$. If you knew basic math, you would know that's impossible.

Again and again, you lie. I never faked anything, only one doing it here is you.
You have date and time of predictions I posted, and you have last edit time, so you can see predictions are posted and not edited.

Yes, no one is interested seems like, or they don't post amount of posts like you do.

But that speaks more about you and people here. It speaks very loudly that you are completely ignorant, rude lier with enormous amount of low importance complexity. And somehow people believe in your ability to cope with your problems. It is one more confirmation that people don't know anything and they thing stuff you write on this forum makes sense and you know something, not realising you are completely ignorant.

Here's small advice, go learn some basic maths and statistics, it will help you lie more efficiently

Cheers.

You will need 900 out of 1000 to make me think you're extremely lucky at following a trend when the price isn't moving at all because of external factors, I've seen guys managing to get lucky shoots with less than 250$ deviance over the price 3 months in advance three times in a row with pure guesses while we had competitions here with no machine learning, just typing a random number.
But even that won't mean I trust a single word about:
- you running a system based on speech recognition that predicts the price
- you have an actual system at all and it's not just basic TA that will go terrible wrong
- your need of money when you could have already made tens of thousands starting with 10$

After the data faking in the first post I don't trust your results with down up posted at random hours at all without a double checking, but since nobody seems interested in your obvious fake system there is no need for it.

Another one, 14 out of 14 predictions correct in a row.

@stompix

How many correct predictions you need, to believe it's possible to predict price direction with ~90% accuracy?

I removed prediction for 07.08.2022; run system again, and it returned same price as yesterday, meaning that direction for today's price is highly uncertain.

Prediction for 06.08.2022 is downwards.

This prediction is also correct, 13/13 in a row.

No prediction for tomorrow, because system returned same price, interpreting it as uncertain whether will price go up or down.

Another one, 12 correct predictions in a row!

Prediction for 04.08.2022 is downwards.

This one was also correct, 11 in a row.

Prediction for 03.08.2022 is downwards.

Let's continue with predictions.

Prediction for 02.08.2022 is downwards.

Recap:
System predicted 10 closing candles in a row.
To put things in a perspective of probability; probability of that happening is 0.0009766 and chance of success is 0.097%.
You might not think that 10 is a lot, but doing it in a row is indeed remarkable.

Hi davidvictorson,

It took me more than two years to research and later to develop the model. Yes, you need to add more features and processing to make models work, simply plugging price, indicators or news into neural network will give you nothing. As I mentioned on the thread, speech recognition problems are very similiar to markets, taking ideas from that field is quite useful.

What do you mean about long term predictions - predicting longer horizon e.g. 10 days into the future, or the performance will not be good long term?

Currently I have 2080ti GPU, and it takes me ~50 minutes to compute the prediction I post here regularly.
Well, there's plan to make server with multiple Nvidia Tesla GPUs, so I can run predictions for more instruments, coins, but I will need to raise some serious money to be able to set up something like that.