Predicting price trends of digital products using various forecasting techniques

: on the example of the steam community market

Abstract

The gaming industry has experienced steady growth over the years, contributing to its increasing commercialisation. One factor adding to this trend is the growing popularity of online marketplaces for in­game items, some of which are traded using real­world currencies and considered by a growing amount of young people as some new asset class. This study addresses the question of price predictability for these items, as the efficient market hypothesis posits that it is impossible to consistently predict future prices based on past prices. While this topic has been extensively discussed in the literature for classical financial time series forecasting, it has not yet been explored in the context of in­game item marketplaces.This study used data from the Steam Community Market to investigate the predictability of ingame item prices in the context of online marketplaces. Multiple linear and non­linear forecasting models are applied to the data. This study shows that the price is predictable to some degree for many items, although the improvement is small compared to the naïve benchmark.Specially, linear models showed auspicious results for stationary data and short­term predictions, while non­linear models rarely delivered a strong performance. These findings suggest that forecasting digital items may be as challenging as forecasting traditional assets.

Date of Award	3 Feb 2023
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Nicolò Bertani (Supervisor)