A great inflow of spectators and investors has flooded the bitcoin “business”. This is due to the price explosion in the past month.
December 8th of the past year witnessed the market capitalization of BTC soar above that of VISA INC at more than USD300billion.
Nakomoto, the original creator of BTC defined the of Block creation(i.e. mining) to be adjusted every 2016 blocks with a two week adjustment period with which the amount of BTC created per block decreases by 50% per 210000 blocks. Thus, the number of bitcoin wouldn’t expect more than 21 million in existence.
Even with this model for the supply of BTC the intrinsic value of BTC is hard to determine as result of cryptocurrency inability to generate interest making it impractical to adopt the existing valuing formulas. The price of Bitcoin is determined by mostly the chaotic process of continuous double auction mechanism (CDA). Fundstrat has shown that most of the variation of BTC price can be explained using two blockchain stats
- The number of unique addresses
- The estimated transaction value per address
However, Metcalfe has similar model wherein the value of telecommunications network is as similar as the square of the number of the connected nodes.
This short paper illustrates that the variation of the BTC price can be more accurately explained by additionally introducing a market sentiment volume about BTC that is positively scored on Twitter and StockTwits. We use sentiment data of the cryptocurrency market aggregated by Decryptz. This model shows an improved accuracy in both in-sample and out-of-sample regressions.
According to Metcalfe’s Law Robert Metcalfe, co-inventor of the Ethernet network technology, formulated in the early 1980s that the value of a telecommunications network is proportional to the square of the number of connected nodes of the network. Given n nodes, the total number of connections that can potentially be established in the network is the triangular number n(n-1)/2, which is asymptotically n2. The value of the network, therefore, grows on the order of O(n2) while the cost of the network increases linearly with n.
Metcalfe has later modified the law using n×log(n) instead of n2.  In the context of the BTC blockchain network, each node for Metcalfe’s law can be represented by a unique “address” – i.e., an identifier of 26-35 (usually 33) alphanumeric characters representing a destination for a bitcoin payment.
However, this is a compromise since each BTC user can have multiple BTC addresses. In fact, by design, it is an unintended practice to use the same BTC address for multiple transactions. For this reason, the number of unique addresses is not the precise representation of the total number of unique users. It is impossible to know precisely at least from all publicly available information.
Since BTC does not make any cash flows the price of BTC is determined predominantly by the dynamic demand-supply balance of buyers and sellers. Also, the trading and investment behavior of participants are largely frisking at the speculations in the market, properly examination of which will absolutely help explain the BTC price.
In order to measure the sentiment quantitatively, a sentiment dataset aggregated by Decryptz from Twitter and StockTwits is put into the picture.
Their natural language processing (NLP) engine ingests those social networks’ data streams every day with a filter identifying and collecting the volume of conversations for each cryptocurrency based on cashtags (“$”), hashtags (“#”) and keywords. The engine subsequently analyzes the language in each conversation to determine the sentiment as well as its intensity level. The engine assigns an intensity score to each conversation in a range of -4 (extreme negative sentiment) to +4 (extreme positive sentiment). Further, the system will determine its sentiment and intensity level by analyzing the language in the conversation. Depending on the level, it assigns an intensity score to the conversation level.
Models & Results
We build two ordinary least squares (OLS) models to regress the price of BTC. The first model: SMA5(n)×log(SMA5(n))
Where n is the number of unique BTC addresses obtained], and
SMA5(v) divided by SMA5
v = the estimated BTC transaction value in US dollar.
t = the number of confirmed BTC transactions per day.
In addition to the aforementioned two variables. The second model uses the positively-scored market sentiment volume about BTC on Twitter and StockTwits.
Fundstrat uses n2 instead of SMA5(n)×log(SMA5(n)), and
SMA5(v) divided by SMA5(n) instead of SMA5(v) divided by SMA5(t).
However, we find that the latter in both cases produce a better result. This can be attributed to the fact that n×log(n) better explains Metcalfe’s law than n2. There are unused BTC addresses; v, therefore, better captures the BTC network activity than n. We use observations from September 1st, 2014 to September 19th, 2017 as in-sample (IS) period, and observations from September 20th, 2017 to December 1st, 2017 as out-of-sample (OOS) period.
A simple model that accurately explains the price of BTC was presented. The BTC price can be accurately explained using a market sentiment volume about BTC, positively scored on social networks. The assumption is that the price of BTC is predominantly a result of the demand-supply balance of buyers and sellers. This is greatly influenced by market sentiment.
It is likely that market sentiment is also highly affected by the price itself. In such a relationship, cause and effect influence one another and together create a feedback loop. As a result, asset values are not driven only by the economic fundamentals. It is also often driven by biased assumptions and actions of participants spreading across the market. This leads to markets having boom-and-bust disequilibrium, which the price of BTC seems to be experiencing.