It’s been more than a year since the crypto-currency market “shot” for the second time, and dozens, and maybe hundreds of digital assets, showed an increase of 1,000 or more percent. For the year of 2017, a three-month correction followed, and many crypto investors have pondered how to improve their investment strategy. A powerful pullback of market capitalization demonstrated that the “buy and hold” method, which is relevant in the second half of 2017, can now play against the investor.
In this connection, experts and analysts began to experiment even more actively with various methods of analysis. Classical technical and fundamental analysis, although considered working methods suitable for any market, is not satisfied by every trader and investor. As an alternative, in 2018, the analysis of objective statistical data for each crypto currency (number of users, subscribers, listing in crypto-exchanges, the dynamics of the previous price change, etc.) begins to gain popularity.
Based on these data, systematized in the table, correlation models are constructed, and the dependence of the price (capitalization) of the cryptoactivity on various parameters is determined.
Simplified, the system works as follows. A specific parameter is analyzed (for example, the number of subscribers of the top-level Telegram channel of the crypto currency). The more crypto currency is involved in the comparison, the more objective data will be obtained.
For example, in the case of N of the crypto currency, with the capitalization of $ 100 billion, 50,000 subscribers in Telegrams, and in M of Cyrillic currency, with the capitalization of $ 200 billion 90,000 subscribers.
Further, the indicators are averaged, a general capitalization parameter is output, for example, for 1000 subscribers. After receiving this parameter, one can predict the growth or decline in the capitalization of each specific crypto currency, knowing the number of subscribers of its Telegram channel at the moment, and the discrepancy of the ratio of this quantity to the price with the average.
However, the number of subscribers in Telegram, Reddit or Twitter is just one of the possible parameters, and it is by no means a fact that its analysis gives the most objective information. By betting only on one parameter, the expert falls in dependence on him. If in fact these data will be insignificant for the market, then the entire forecast will be useless. That is why the most effective technique for constructing trading algorithms based on statistics is multicriteria analysis. It involves taking into account dozens of parameters when determining dependencies.
Daniel Chen, the founder of the OpenToken project and an expert from Andreessen Horowitz, started developing his own algorithm for analyzing statistical data to reveal their correlation with the market capitalization of crypto currency.
Algorithm based on CRV Crypto Research
Having developed his own code, Chen demonstrated his work on the basis of the CRV Crypto Research table, which presents constantly updated statistics on 51 of the world’s largest crypto currency.
It looks like this:
It is worth noting that Chen himself considers this table not the best source of information due to the ratio of the number of rows to the number of columns (51 to 21). According to Chen, this ratio should be at least 10: 1.
Next, for each parameter, the correlation value with market capitalization is determined.
The values on the right are the coefficients of determination, or the correlation coefficients, squared. As a rule, it is the determination that is considered as the main indicator.
The square of R in the first line is 0.138249, on the graph this will look like this:
This parameter is far from optimal, since the best correlation coefficient is 1. The closest to this value is the parameter “Number of users of Reddit and Market capitalization”, equal to 0.81.
In order to reduce the heterogeneity of the readings, you can order the data on a logarithmic scale. After that, the parameters will look like this.
Now the data looks closer and more orderly, but the logarithmic influences some indicators. In particular, the square R for “Number of Reddit Users and Market Capitalization” decreased from 0.81 to 0.36.
In addition, Chen proposes to test statistical hypotheses. In order not to go any further into specific formulations and calculations, we immediately derive the obtained value and determine that they demonstrate.
A general assessment of the coefficients shows that the capitalization of crypto-currencies depends most on the popularity of the digital asset. However, these data are obtained from already collected statistics. It is not entirely clear how to base this on forecasting.
To assess the dynamics of changes in capitalization, it was decided to conduct the latest express test of changing the ratio of capitalization and the number of subscribers for a certain period. After the sharply deviating values were removed, we obtained the following relationship:
This means that for a certain period, as the number of subscribers of the crypto currency page in Reddit increased, its capitalization fell!
In this case, part of the evaluation period fell on the winter of 2018, when the entire crypto-currency market entered a deep correction. However, our statistical system must take into account all factors, and not be influenced by them – otherwise it will not be considered universal.
The data obtained from this small experiment make it possible to understand the following:
- The system, built on the analysis of statistics, does not take into account external factors, and can give out incorrect or contradictory forecasts in periods of high volatility.
- Often, the data that seem most important, in practice, does not have any effect on the change in capitalization.
- The search for correlation between individual parameters is only a small part of the quantitative analysis. To obtain the most objective data, huge amounts of statistical data are needed.