Trading strategies to tackle uncertainty in crypto markets and traders' psychology

The main hurdle with crypto trading is dealing with two types of uncertainty. First, the market is always fluctuating unpredictably, and second, this unpredictability makes it tough for traders to make firm decisions about when to buy or sell. One way to approach this is by using the concept of quantum superposition to explore all the potential outcomes in both the market and traders' thoughts. Then, we can apply an evolutionary algorithm to find the best trading strategy, which is basically a sequence of buy or sell actions. In a recent case study, three AI agents worked together to analyze about a month’s worth of historical Bitcoin price data. They came up with a trading strategy for the week ahead that had success rates above 60%. If you want to read more about this, check it out here: www.xinvisionq.com/subpages/caseStudies.html. And if you're curious about our algorithm, there's a paper with more info available here: www.xinvisionq.com/pdf/QxEAI.pdf.

Thats a huge result of 60% but can we really trust AI to compete with human in regards to predictive movement of the market? I mean yes trading is hard and has historical results and thats also our basis as trader, does it mean we can rely fully on them maybe if not now more than ever in the future.

You are not the first entity to introduce a new kind of methodology for trading to this community. We have seen 1000s of similar algo trading introductions in recent past but nothing had brought any differences in trader's final result. These days AI and quantum are real clickbaits, hence even you might have come up with a real thing, I am sorry, you are falling into same category just because of saving my time and energy. This is the reason, long back itself I have concluded that trading is not a way to make money but to lose. I know, people who are yet to make similar conclusion are your targets but there are people here to listen rather than rushing up which is what exactly I look for, out of my experience. Still, only 60% ? Quantum principles are also not enough to beat these markets? Very sad, because there technical analysts who claim about more than 90% accuracy with their manual signals. Better improve your "evolutionary algorithms".

In comparison to humans, AI agents don't "think" all they do is compute, but that isn't a bad thing necessarily because they don't have any emotions. We humans tend to overthink things too much, basically we're rational and irrational, and this doesn't play out to our advantage especially for trading which is just being able to make decisions constantly under incomplete information. No matter how hard we try, nobody can accurately predict the absolute price of crypto at any given time, so what's more important is finding the market's overall trend of the recent past, and somehow when the multiple AI agents cooperate and learn the historical trading data they are able to come up with a trading strategy to produce decent short horizon forecasts. The emphasis here is studying more recent historical data, around 30 data points to then predict the foreseeable near future of about 5-7 data points. Basically, the longer the forecast horizon the odds will merge closer to 50-50, which is exactly consistent with the well-known random walk of the market. If the market is truly in a random walk then indeed no one can beat it since there's no best model to predict the future, and you're right in the long run everyone loses, not just crypto, for stocks, futures, forex, etc. as well. However if the more recent past of the market has some trend and the AI agent might be able to learn the historical data and find it (because they don't "think" they just compute) to predict the near future of about 5-6 data points. That is of course we assume that "nature doesn't jump", if a black swan lands suddenly from nowhere then everybody loses anyway, don't matter humans or AI. Basically think of it this way, our tool whatever you want to call it, AI quantum evolutionary whatever, is just your very own personal assistant decision maker - just plug in the data and it'll analyze it for you and give you a strategy that you can use as a reference just like you would from reading the news, talking to other Bitcoiners here on this forum, or listening to those technical analysts - but at the end of the day it's up to you to make the final decision of whether to buy or sell, and just FYI we never claimed that our tool can make you rich overnight. [moderator's note: consecutive posts merged]

This is the first challenge a trader can face when he don't know what to do, how to start or which pattern to even use, how will such develop a reason for making nany decision while he trades, at the end we are going to discover them having much of losses on trades, we can't do without considering for learning nat first because this is the most secured guide every trader may need before a start, then we must have the understanding of the market being a volatile one and the associate risk involved before starting.

As you make progress in the market and acquire knowledge and experience from observation and practice, the decision of when to buy and when to sell becomes easier. Because of the experience you have acquired and the knowledge you have, you will be able to see patterns easily and better than when you just started trading. There is still uncertainty, of course, but not to as much as when knowledge is still very much below average.

You have a point but this point has never been new. If it's easy to predict the market with 100% accuracy or closer, won't everyone be rich by now? And if that happens, there will be no liquidity provider because all of them must have gone bankrupt and big guys and banks/institutions (market makers) wouldn't dare to risk their money in it again. So this uncertainty in it makes it survive. Besides, I rather call the market dynamic because it is readable, and if you are good at your analysis, you will be able to know the right direction most times.

Those trader should know when they can buy or sell so that is why they must analyze before they trade. That will eliminate them to buy or sell in a wrong time because they can determine the time they can do action. I don't rely on AI agents to let it analyze but I always analyze by myself so I can see what I can do related to the market situation. If the market say that the situation is not good, I will skip to trade and just wait for more. I don't have any reason to force myself to keep trade because that can make me in a risk situation.

I'm curious about your approach. What is the "random walk" that you are talking about? What's the context of this? It's not quantifiable and would be hard to analyze if that's the case. Because if I understand correctly, this will help you decide whether you would short or long, right? Do you also have the data showing how much historical data the AI analyzed? I'm curious because that's a good result in some way. Maybe a lot more testing would be best when this is applied continuously now.

All the important "meat and potatoes" information is in the current closing price at a given time, so basically no matter how much knowledge you gain about the current price of Bitcoin now you still won't be able to predict the future prices and that's exactly what the efficient market hypothesis states and that is also where the term the market is in a random walk all the time comes about, random meaning that there is no pattern that can be recognized easily. At the end of the day the best a trader can do is to guess and guess "emotionally"; but it's these emotions of greed and fear that will eventually cost traders' big time in the long run, so a "good" trader isn't the one who has a breadth of knowledge about the market's historical past, but one who can "control" their emotions effectively. Thus when it comes to AI, AI is non-emotional and only computes, and so AI is somehow able to find some form of trend from very complex data in ways which we humans don't seem to be very good at.

I never tested how much accurate prediction can give AI agents, so I don't want to trust with AI agents, I am trading with my own strategies, So for that need to research myself. Yes when we see market condition is very bad when definitely we will never entry in the market, but I don't know how AI agents decide it. And when we see good time to start trading then I will buy coins.

So first, yes when the market trend is not very clear, you can wait & see and don’t trade; just like how people say I’ll just buy when the trend is fairly clear that the price will increase, this indeed is a good strategy, and actually this exact strategy is employed by many trading firms. However, in the long run how much profit can you actually make by adhering to this strategy only? And second, how many people can stick hard to trading rationally, without any emotions and determinedly trade with their strategies they’ve come up with? Third, the market is called an efficient market for a reason, thus there isn’t going to be a very obvious trend all the time, most times it’s volatile, and that is why it’s called a random walk. It’s important to keep in mind that the price of bitcoin is determined by the “collective actions” of all the traders’ involved (some people buy and other sell which then causes the price of bitcoin to fluctuate), it’s not if you, as an individual, thinks that it’ll increase or decrease that actually causes the price to do so. At the end of the day, no matter how much research you’ve done, you’ll still just have to guess the movement of the market.

AI agents can be only pretty trendy and get the info from lots of sources in one place - and that's in a good case. At worst - he will just hallucinate a bit and won't give any proper advice whatsoever.

I love this part of the conclusion,   Really! Four weeks, lol. Quite a large sample size hahaha. This paper does nothing but highlight the importance of data which everyone already knows. There is no specific use case for the 'superposition principle' outlined explicitly in the paper, this should be a philosophical paper .

That will more better for you so you can improve your skill in trading and not just rely on to others. Maybe AI agents could help us to gather the data useful for us in analyzing. But we must analyze and research furthermore by ourselves so we can know what happen to the market exactly. By analyzing the market, we know what we must do and not just enter the market without any data. We will be careful decide because we want to responsible with our capital.

AI agents are only tools to help you do something mundane. They are not gurus, at all. And shouldn't be considered as such.

Of course, four weeks of data may not seem like much, but let’s be blunt, what does the price of Bitcoin or some stock from 10 years ago have to do with predicting the future price of Bitcoin for the following week or for the next 10 years? Looking back that long in the past or forward towards the future doesn’t really have much of an effect for forecasting what’s happening at hand now, it doesn’t mean the larger the dataset the more accurate the prediction you’ll get – quality not quantity. Yes, for some mathematical or statistical analysis it might be of some use, but if we’re just looking at the forecast horizon of the foreseeable future, studying the more recent past is more efficient, well at least in our opinion we believe that you don’t need a tremendously large dataset to forecast a short horizon, so we only used four weeks of data in our case study. As to the superposition principle, we clearly stated in the paper it is utilized to model the challenge of dual uncertainty; the uncertain market and the traders’ actions, where at any given time the market could go up or down and the traders’ can either buy or sell, and this is, to the best of our knowledge, an approach that has not yet been attempted.

So price of bitcoin older than 4 weeks has no impact on current price but principle of superposition does. You are using  the term superposition like some spiritual guru is using by giving the example of Schrödinger's cat. Again I am not trying to be someone who is just passing some unnecessary sceptical comments. I have gone through your paper and this is my honest opinion. You started your paper with the absolute importance of data in order to make prediction models and when I pointed out that the data sample is too short for claiming 80% success rate, you started new argument about the relevance of previous data. I did not say 10 years, I said not enough sample size which will be pointed out by anyone who has given your paper a serious glance. This is the reason I said this is more of a philosophical paper.

The current price of bitcoin is determined by the collective actions of all the traders’ participating, whether they buy or sell is what causes the price of bitcoin to fluctuate, no amount of historical data, whether from years ago and days ago determines the current price of bitcoin, and definitely the superposition principle does not determine the price of bitcoin, don’t know where you got that impression. As to your opinion about four weeks of data being too small of a dataset to claim 80% accuracy, we particularly used a smaller dataset to forecast the following week because for finding the trend of the short forecast horizon the more recent historical data gives more insight as opposed to data from a year or two ago. Essentially, for statistical analysis, indeed using large amounts of data might be of better use, but for forecasting it’s simply not just the bigger the better. And lastly our methodology can be utilized to time series forecasting generically, stocks and bitcoin are just one case study, our algorithm can forecast total sales data; retail demand, QSAR of drug design, scientific discovery, etc…, basically any form of time series forecast data. Our algorithm has been tested on many real-world datasets, so we don’t get your point of why you keep persistently mention that our paper is a philosophical one.

You are going against your hypothesis. Also, tell me what would be the basis of prediction model if not the previous data?  Are you kidding me? You have mentioned it yourself in the paper, in fact, that is the core of your paper.  But you jumped straight to price prediction, I wonder why? Proof? Let the community examine.