A regression analysis of factors affecting total points of SHL (Swedish Hockey League) teams in a season

Sammanfattning

This report explores which team-level performance metrics best explain the total points accumulated by SHL teams over a full season. This is done by performing a multiple linear regression analysis using Python. Three different variable-selection strategies were compared: a model based on our intuition and knowledge about hockey , a two-variable model designed to minimize multicollinearity, and a larger model (4–8 predictors) chosen to maximize adjusted R^2 value while keeping pairwise correlations under 0.4. Residual plots were used to check that the model assumptions were reasonably satisfied. The best-performing model with a adjusted R^2 of 0.944 includes shooting efficiency (SEFF), power-play shots on goal (PPSOG), shots on goal against (SOGA), save percentage (SVS%), unblocked shots (FF), and blocked-shot percentage (Blocked%).Based on these results, we derive on-ice implications and propose some strategies.