20.500.12556/RUL-173264 Modeliranje vpliva smrti igralca na rezultat igre League of Legends Analysis of the impact of player deaths on game outcomes in League of Legends League of Legends (LoL) se je uveljavil kot eden izmed najbolj gledanih ter igranih e-športov na svetu. Naše delo preučuje vpliv smrti posameznega igralca na verjetnost zmage v igri League of Legends. Z uradnim aplikacijskim programskim vmesnikom podjetja Riot in jav- nimi viri zberemo velik nabor tekem diamantne lige in iz vsake tekme izluščimo minutno predstavitev stanja. Na podlagi teh značilk zgradimo in kalibriramo model verjetnosti zmage. V nadaljevanju opredelimo pojem okna smrti - in- terval med stanjem tik pred smrtjo in prvim stanjem po vrnitvi igralca v igro. Spremembo ekipne verjetnosti zmage v tem oknu porazdelimo med igralce z metriko Performance Score, ki je odvisna od vloge igralca v igri in uteˇzena z globalnimi SHAP ocenami. Empirično analiziramo 150 472 oken smrti. V povprečju se verjetnost zmage ekipe mrtvega igralca zniža za 2,03 %, pri čemer v 58 % oken upade, v 42 % pa naraste. Ugotovimo, da se pozitivni prispevki igralca, ki je umrl pogosteje pojavijo, kadar igralec tik pred smrtjo tudi ubije nasprotnika oz. dobi zlato. Negativni prispevki pa se pojavijo v pričakovano več primerih in so rezultat smrti, ki ne koristijo niti igralcu, niti ekipi. Predstavljeni pristop omogoča hitro prepoznavanje najvplivnejših smrti in učnih priložnosti. League of Legends (LoL) has established itself as one of the most watched and played e-sports in the world. Our work examines the impact of the death of an individual player on the probability of victory in League of Legends. Using the official Riot Games application programming interface and pub- lic sources, we compile a large dataset of Diamond-rank matches and derive a minute-by-minute representation of each game state. Based on these fea- tures, we construct and calibrate a win probability model. We then define the concept of a death window —the interval between the state immediately before a death and the first state after the player returns to the game. The change in team win probability within this window is attributed to individual players using a Performance Score, which depends on the role of the player and by the global SHAP values. Our empirical analysis covers 150,472 death windows. On average, the probability of victory for the team of the deceased player decreases by 2.03%; the probability falls in 58% of windows and rises in 42%. Our results show that positive contributions from the deceased player occur more frequently when the team secures objectives or achieves favorable trades during the absence. The proposed approach enables rapid identification of the most impactful deaths and learning opportunities. League of Legends strojno učenje analiza podatkov vpliv igralcev League of Legends machine learning data analysis player impact true false false Slovenski jezik Angleški jezik Diplomsko delo/naloga 2025-09-15 11:05:14 2025-09-15 11:05:18 2025-11-07 03:51:13 0000-00-00 00:00:00 2025 0 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 38152 250548995 Ferreira_Ruben_-_Modeliranje_vpliva_smrti_igralca_na_rezultat_igre_League_of_Legends.pdf Ferreira_Ruben_-_Modeliranje_vpliva_smrti_igralca_na_rezultat_igre_League_of_Legends.pdf 1 005B0A90C48F307E1DDE2350983CC662 09e8a2d870a4ddcae74b4ca8ac49a9b6ded74f1c1a307e8139a84b0736962c5b ee63b1de-9212-11f0-9328-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=217480 Fakulteta za računalništvo in informatiko 0 0 0