Is Bitcoin High-Frequency Trading Only a Noise in the Market? Evidence from the Analysis of Optimal Sampling Frequency

The main aim of the paper is to study and determine the optimal sampling frequency for modelling cryptoasset volatility. For that purpose, the noise-to-signal ratio is estimated for six selected cryptoassets, separately for each month of existance on Kraken exchange. The minimal noise-to-signal ratio among sixteen frequencies implies the optimal sampling frequency. It is hypothesized that during the periods of increased activity in the market related with increased investor attention, the optimal sampling frequency increases. Moreover, according to heterogeneous market hypothesis, such frequency is hypothesized to differ across assets and over time. The results showcase that optimal sampling frequency is above 1-hour for most months in the case of most assets. It is particularly interesting in the case of Bitcoin, as the optimal sampling frequency increases above 60-minute only 5 times over 9 years. This implies that signals from larger moves at lower frequencies are more valuable than signals from high-frequency trading, except of few specific periods, such as bubbles in 2017-2018 and 2021-2022 or COVID outbreak in the first half of 2020