| Abstract [eng] |
According to the efficient market hypothesis, arbitrage events should not exist or are eliminated very quickly, since the asset price reflects all available information and it is impossible to beat the market, i.e., profit from risk-free trades by taking advantage of price differences between markets. The emerging field of behavioral finance challenged the long-established efficient market hypothesis by arguing that investors are not always rational and that arbitrage opportunities may exist. It is unlikely that all markets are an efficient and it is worth distancing yourself from the efficient market hypothesis. Still fairly new, with little regulation and still developing, the cryptocurrency market is seen as an inefficient market that often has large and repetitive arbitrage opportunities between exchanges. However, “catching” them is not so easy. It is not clear to the investor in which exchange to wait for arbitrage opportunities and where it is worth keeping his funds. Trades must be completed at approximately the same time before they can be judged to be profitable. The purpose of this final thesis of master's studies is to evaluate arbitrage as one of the investment strategies in the cryptocurrency market and to create a mathematical model of the waiting time between two bitcoin arbitrages. For this thesis, bitcoin was chosen to be examined, as it is still the most popular cryptocurrency and has the largest share of the cryptocurrency market capitalization, as well as two years of bitcoin arbitrage events, with sales on “Bitstamp”, “BitBay” and “CEX.IO” crypto exchanges and purchases on 19 other exchanges. In the thesis, arbitrage investment strategy was examined and a mathematical model of the waiting time between two arbitrage opportunities was created. It was done by estimating the time distribution between two arbitrages and creating a Markov chain model. It has been shown that the most suitable waiting time distribution is described by mixtures of two gamma distributions after separately estimating the probability that the waiting time is equal to zero. However, it is difficult to select a single distribution or a mixture of distributions when evaluating the time between arbitrage events in a single exchange. It is worth examining the distribution of waiting times between arbitrage events for each pair separately. Markov chain model with two states: arbitrage and no arbitrage showed that we have an arbitrage event in average 2–4 minutes and it will repeat again every minute. The resulting stationary probabilities showed that there is a higher probability of no arbitrage, which is a slight indication of an efficient market. |
