We developed a model for predicting the final rankings of competitors in alpine skiing using a two-stage approach. The first stage focuses on predicting the standard deviation of times in an individual race, which corresponds to an estimate of the race's difficulty and the variability of results. The second stage of the model predicts the deviations of individual competitors' times from the standardized average, enabling the prediction of individual results. Together, these predictions form a normal distribution, from which the predicted time for each competitor is sampled. This approach allows for the generation of a wide range of possible rankings and the calculation of probabilities for various scenarios, such as podium finishes or exact predictions of final rankings. Experimental evaluation conducted on data from three seasons (2020/21–2022/23) demonstrated a correlation between the predicted and actual results, with the model often correctly predicting which skier will outperform another, even if it does not always determine the exact order.
|
