Prediktivní síla sentimentu na sociálních sítích ve sportovním sázení

Podzemský, Lukáš

In the modern society social media platforms have become integral part of everyday life. Similarly, the sports betting market has emerged as a global phenomenon in recent years. This thesis combines these two worlds and ex- plores the predictive capabilities of social media sentiment, specifically from Instagram, in predicting outcomes in the English Premier League and assess- ing its value in sports betting markets. By collecting over 368,000 Instagram comments posted before 190 matches in the 2024/2025 season, the study em- ploys sentiment analysis using three models: VADER, BERT, and a hybrid model combining the two. These sentiment scores are used to derive match outcome probabilities, which are then compared to probabilities implied by bookmaker odds. Through logistic regression and performance evaluation met- rics, the research investigates whether probabilities derived from sentiment offer a statistically significant edge over traditional odds. The findings suggest that sentiment-derived probabilities could marginally enhance forecasting perfor- mance in a few isolated cases. However, these improvements were modest and inconsistent. Moreover, none of the sentiment variables were statistically sig- nificant when included alongside odds in the regression models, indicating that bookmaker odds...

guest :: login Digital Repository
		Search		Submit		Help		About