source: http://www.jacobstead.com/Research-World
If you could make better predictions about future prices, investing in cryptocurrencies wouldn’t be such a risky business, right? Luckily, new alternatives are coming your way to let you do just that.
In this article we are going to look into the three prediction techniques that rely on people’s judgment and technology:
- Expert Questionnaires (the “Delphi” Method)
- AI-based solutions
- Prediction markets
We will highlight the essential information for each techniques and try to come to a conclusion about what would have the highest success rate when predicting cryptocurrency prices. Let’s get started!
Method 1: Rely on the Judgment of Experts — The “Delphi” Method
The Delphi Method (named after the Oracle of Delphi from Greek mythology) is a forecasting technique based on the results of several rounds of questionnaires completed by several panels of experts. (https://www.investopedia.com/terms/d/delphi-method.asp) Typically, there are five to twenty experts who answer the questionnaires in 2 to 3 rounds. The experts are chosen by a coordinator who finds professionals dealing with this topic, from academics, senior professionals, or other subject matter experts. The coordinator contacts them and suggests participation in the study, usually for some remuneration. The responses are anonymous and each expert can comment on the answers given by others.
Experts give average estimates with regards to the questions posed as well as their justifications. Their replies are weighted in order to get an aggregate result. With each new round experts can correct their responses based on the feedback given by other experts in the previous round.
Domains of application
Initially, the Delphi Method (developed by the RAND Corporation in the 1950s) was used to forecast major technological and scientific developments, such as labor automation and the space program. As the method proved to be highly efficient, people in the subsequent years started to apply it to a broader range of domains. It is frequently used in business and is especially popular for public policy decision-making. For example, researchers have used the Delphi Method in a study regarding “happiness” in order to uncover the best strategies to improve people’s overall satisfaction.
Success Rate
Success rates differ from one Delphi study to another, and in some cases it is very accurate. Its accuracy for sales forecasts can be as high as 96%-97%, and in future technology forecasting it reduces uncertainty from 100% to 20%-30%, meaning the forecasts are 70%-80% accurate.
The Delphi Method is quite accurate for long-term studies, however, its implementation with short-term goals, such as predicting cryptocurrency prices, is problematic. Gathering experts and iterations of several rounds of questionnaires takes time. As a result, it is very difficult to use Delphi to predict immediate fluctuations in any market.
Real World Examples:
eLAC Action Plans is an initiative that relies on the use of information technology as the main pillar for the socio-economic development of Latin America. In 2005, the Delphi Method was used to identify the major challenges for the initiative, and more than 1,400 experts participated in the five rounds of Delphi questionnaires. The case demonstrated the potential of Delphi methods for participatory public decision-making.
TechCast is an online project by George Washington University. It uses the Delphi Method to forecast major technological developments relying on the opinions of more than 150 experts from all over the world. The project gathers expert responses to predict the approximate date of the adoption of new technologies. In a sample forecast, for instance, they projected that smart grids will be widely adopted by 2026. Each forest is accompanied by the experts’ justifications for their opinions.
This is another online project relying on the Delphi Method for technology-related forecasting. It is administered by the New Media Consortium and forecasts major trends in technology adoption in the education industry. Based on the data gathered from experts using the Delphi Method, Horizon publishes yearly reports on technological trends for schools, higher education, libraries, and museums.
Method 2: Rely on Technology — AI-based solutions
Technology-aided prediction techniques rely on machine learning and large datasets. Usually there is a training dataset based on a computer “learning” the major patterns that are used for making predictions. For example, a training dataset for the prediction of the likelihood of the development of certain diseases based on one’s physical features would include the characteristics that are used for prediction, such as data about age, height, weight, etc. After the computer learns the patterns in the training dataset, an algorithm can be used to predict whether a person is ill based on the given set of physical traits. This can then be compared to real information about the accuracy of those predictions.
AI-based solutions are advancing at an extremely fast rate, and they are becoming highly accurate. However, unlike the Delphi Method and other solutions that rely on people’s judgment, AI-based prediction techniques make the justifications for the forecasts ambiguous. The patterns identified by computers are not explicit and scientists can struggle to understand the underlying logic behind the machine’s predictions.
Domains of application
AI-aided solutions are currently used in a broad range of industries from healthcare to finance. The algorithms are especially successful when working with numbers, texts, or images.
How successful it is
The accuracy rate of AI-based prediction algorithms is different for each case. It might be affected by the amount and quality of available data, as well as by how well a particular machine-learning technique suits the data and problem in question.
Overall, AI-based solutions achieve outstanding accuracy for some topics. For example, it can predict a Supreme Court decision with 83% accuracy and the death date of a person who is seriously ill based on their age, medical records, and drug use history with a 90% accuracy rate.
However, it does not work this well for all domains. It shows especially poor results with regards to stock price predictions, having an average accuracy rate of 50%-60%. It happens because the algorithms do not take into account market corrections, human sentiment, and unforeseen events.
Real World Examples:
Google has created an algorithm that can predict with 70% accuracy whether a person has heart disease by simply “looking” at their eyes. The machine analyzes people’s eye scans with a specific focus on the fundus — the interior rear of an eye — which contains many blood vessels that reflect a person’s age, blood pressure, and other characteristics.
- Predicting Oscar winners
One company has used AI to predict who will win an Oscar in 2018. Their predictions for the major categories — best picture, actor, and actress — turned out to be correct. The AI relied on the analysis of Oscar-related public sentiments.
- Trading
Though AI solutions do not work very well for predicting stock prices, they are nonetheless used in the financial industry. For instance, AI-based prediction mechanisms are implemented to allow automatic trading. Some of these algorithms, for instance, rely on social media data to predict immediate fluctuations in stock prices and make automatic trading decisions.
Method 3: Rely on Both People and Tech — Prediction Markets
The core of prediction market-based forecasting is “the wisdom of the crowds.” Prediction markets aggregate the public opinion on the problems in question based on the bets made by users.
On prediction market platforms, “bets” are created on the outcome of events — for example, “Who will become the next US president?” Users then “bet” against one another on the outcomes they deem to be most likely. Money serves as proof-of-stake on prediction markets. With advancements of technology, there is now a distinction between traditional prediction markets and decentralized prediction markets.
In traditional prediction markets, the price of a single bet is typically between $0 and $1, and the current price reflects the probability of the outcome based on the aggregate bets. For example, if the price is currently set at $0.36, it means that only 36% percent of users believe this outcome will happen. The price is dynamic and changes along with the public opinion voiced through the bets on the platform.
In a decentralized prediction market, the “odds” of a choice winning are dynamic as users “stake” the outcome they believe will be the true outcome. Users can risk however much they want and the “odds” are set dynamically. For example, if one hundred users bet $1 each, and a single user bets against them with $100, then the odds would remain 1:1. Decentralized prediction markets also have the benefit of allowing users to create bets on any topic they want, so with the development of these platforms the prediction markets’ application domains will inevitably become much broader. They can potentially cover anything from political issues and sports bets to the stock market or cryptocurrency prices.
Domains of application
Historically, prediction markets have mainly been used in political forecasting. Currently, major prediction markets are used mainly for scientific purposes, and the number of participants on these platforms that the “bets” can cover is limited.
However, the situation is changing now as decentralized prediction market platforms, such as the Bodhi Network, are starting to come into play. We have covered several of them in our previous article about prediction markets.
How successful they are
Traditional prediction market platforms, such as the Iowa Electronic Markets, have proven to be highly efficient when predicting the outcomes of political events. For instance, they were more accurate than the polls when forecasting the results of presidential elections in the US 74% of the time. And prediction markets are far more accurate in the long run as they outperform polls all the time when forecasting more than 100 days in advance.
As decentralized prediction markets are just now emerging, there is no data yet available on how accurate their forecasts are. Still, there are reasons to believe that they will be more accurate than traditional prediction markets. They will not have any limitations with regard to the number of participants or the total sum of bets. Furthermore, the decentralized prediction markets may not have any restrictions on the users’ place of residence, making them truly global. These factors will likely result in increased forecast accuracy. What’s more is that decentralized prediction markets can be fraud-proof, once again making them potentially more efficient than the traditional solutions.
Real World Examples:
The oldest online traditional prediction market. It is operated by the University of Iowa and is used primarily for scientific purposes. The bets on the market are created by its owners, and they deal predominantly with US politics. The maximum betting amounts are restricted to $500 in order to comply with the US regulations.
A decentralized prediction market platform that works using cryptocurrency. The number of participants and maximum bets are non-restricted. The bets can be created by the users and the topics are not limited. One will be able to create bets on presidential election outcomes, sporting events, cryptocurrency prices, and other matters.
Conclusion: What Works Best for Cryptocurrency Price Prediction?
Based on the three techniques presented above, we believe that prediction markets will work well for the prediction of cryptocurrency prices for the following reasons:
The Delphi Method is no good for cryptocurrency price prediction because it takes too much time to make a prediction
It might be used to make a prediction for the price of, say, Bitcoin in the long-run, but it is no good for making predictions about the current changes in price. One simply will not have time to gather up a panel of experts and have several rounds of discussion between them to predict the immediate fluctuations in an accurate way.
AI-based solutions might not work well with cryptocurrency price predictions either
As we’ve noted above, AI forecasting is especially bad at making stock price predictions. Given that the cryptocurrency market functions similarly to stock market, there are reasons to believe that AI will not work very well in this respect as well. Of course, it still has a 50%-60% success rate, and unlike Delphi it can be used to predict the immediate fluctuations in prices. But relatively low success rates still don’t make AI the best choice when you have a lot of money at stake.
Decentralized prediction markets seem to be the most suitable option for the prediction of cryptocurrency prices
Traditional prediction markets won’t work for crypto price predictions simply because they do not cover this area. But on decentralized prediction markets users will be able to create any bets that are of interest to them, including the ones on cryptocurrency prices. Thereby, one will be able to bet on the prices of any cryptocurrency both in the long run and in the short term. If there is no open bet about the currency you are interested in, just create it!
Given that decentralized prediction markets like Bodhi can have a high number of participants from all over the world, the predictions are likely to be highly accurate — at least as accurate as the ones on the traditional prediction market platforms. On Bodhi the prediction result is calculated by a whole network of participants, increasing the accuracy and lowering the costs.
Monetary stimulus will motivate people to share their informed predictions, and by using Bodhi users can win even more than on other prediction market platforms since the network uses its own unique calculation method.
Users of these platforms will likely be blockchain and crypto enthusiasts who are typically well-informed about the cryptocurrency sphere and possibly even about trading cryptocurrencies. An aggregated forecast on market prices from potential and actual crypto-market participants would potentially yield more accurate results.
Bearing in mind everything said above, we find that among the techniques covered here decentralized prediction markets have the highest potential for the prediction of crypto market prices. Interested? Go on and read our previous article about why you should try decentralized prediction markets right now!
About the author:
Kirill Shilov — Founder of Geekforge.io and Howtotoken.com. Interviewing the top 10,000 worldwide experts who reveal the biggest issues on the way to technological singularity. Join my #10kqachallenge: GeekForge Formula.
Decentralization - Hacker Noon_Read writing about Decentralization in Hacker Noon. how hackers start their afternoons._hackernoon.com