Discriminant Analysis of Match Outcomes in the Dota 2 International Championships (2018–2024): Performance Indicators of Winning and Losing Teams

Abstract

Aim: The purpose of this study is to examine performance differences between winning and losing teams based on matches from the Dota 2 International Championship held between 2018 and 2024, using a discriminant analysis approach. The study aims to identify the key performance indicators that determine success in professional-level matches. Method: The study utilised data from professional matches played during the qualifying stages of The International tournaments. The 2020 tournament, which was not held due to COVID-19, was excluded from the scope of the study. The variables analysed included kills, deaths, assists, last hits, denies, gold per minute (GPM), experience per minute (XPM) and net worth. Differences between winning and losing teams were examined using an independent samples t-test, effect sizes were calculated, and discriminant analysis was applied to identify the variables distinguishing the groups. Results: Whilst statistically significant differences in favour of the winning teams were identified for the variables ‘kills’, ‘assists’, ‘GPM’, ‘XPM’ and ‘net worth’ across all years, ‘death’ values were found to be lower for the winning teams. However, no significant differences were found in the ‘last hit’ and ‘denies’ variables in most years. Discriminant analysis showed that in-game performance indicators distinguished between winning and losing teams with a high degree of accuracy, with classification success ranging from 88.3-98.0% percent. Conclusion: This study contributes to a better understanding of the key performance dynamics that influence match outcomes in the professional Dota 2 scene. The findings, obtained through discriminant analysis, provide a quantitative perspective to the literature on esports analytics; at the same time, they serve as a guide for professional teams, coaches and performance analysts in their strategic decision-making processes.

Keywords

• Discriminant Analysis
• Dota2
• Esports
• Match Analysis

Dota 2 Uluslararası Şampiyonası'nda Maç Sonuçlarının Ayırıcı Analizi (2018–2024): Kazanan ve Kaybeden Takımların Performans Göstergeleri

Abstract

Amaç: Bu çalışmanın amacı, 2018–2024 yılları arasında düzenlenen Dota 2 Uluslararası Şampiyonası maçlarını temel alarak, kazanan ve kaybeden takımlar arasındaki performans farklılıklarını ayırıcı analiz yöntemiyle incelemektir. Çalışma, profesyonel düzeyde oynanan maçlarda başarıyı belirleyen temel performans göstergelerini ortaya koymayı hedeflemektedir. Yöntem: Araştırmada, The International turnuvalarının eleme aşamalarında oynanan profesyonel maçlardan elde edilen veriler kullanılmıştır. COVID-19 nedeniyle düzenlenmeyen 2020 turnuvası çalışma kapsamı dışında bırakılmıştır. Kill, death, assist, last hit, denies, gold per minute (GPM), experience per minute (XPM) ve net worth değişkenleri değerlendirilmiştir. Kazanan ve kaybeden takımlar arasındaki farklar bağımsız örneklemler t testi ile incelenmiş, etki büyüklükleri hesaplanmış ve grupları ayırt eden değişkenlerin belirlenmesi amacıyla diskriminant analizi uygulanmıştır. Bulgular: Kill, assist, GPM, XPM ve net worth değişkenlerinde tüm yıllarda kazanan takımlar lehine anlamlı farklılıklar belirlenirken, death değerleri kazanan takımlarda daha düşük bulunmuştur. Last hit ve denies değişkenlerinde ise çoğu yılda anlamlı farklılık saptanmamıştır. Diskriminant analizi, oyun içi performans göstergelerinin kazanan ve kaybeden takımları yüksek doğrulukla ayırt ettiğini ve sınıflandırma başarısının %88,3–98,0 arasında değiştiğini göstermiştir. Sonuç: Bu çalışma, Dota 2 profesyonel sahnesinde maç sonuçlarını etkileyen temel performans dinamiklerinin daha iyi anlaşılmasına katkı sağlamaktadır. Ayırıcı analiz yöntemiyle elde edilen sonuçlar, espor analitiği literatürüne nicel bir perspektif sunarken; aynı zamanda profesyonel takımlar, koçlar ve performans analistleri için stratejik karar alma süreçlerinde yol gösterici nitelik taşımaktadır.

Keywords

• Ayırma Analizi
• Dota2
• Espor
• Maç Analizi

References

1. Agarwala, A., & Pearce, M. (2014). Learning Dota 2 team compositions. Unpublished manuscript.
2. Büyüköztürk, Ş., Kılıç-Çakmak, E., Akgün, Ö. E., ve ark. (2008). Bilimsel Araştırma Yöntemleri. Pegem.
3. Demediuk, S., York, P., Drachen, A., & Walker, J. A. (2019). Role identification for accurate analysis in Dota 2. In Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment (pp. 130–138). AAAI Press.
4. eSports Earnings (2018). Top games awarding prize money. https://www.esportsearnings.com/games
5. Font, J. M., & Mahlmann, T. (2018). Dota 2 bot competition. IEEE Transactions on Games, 11(3), 285–289. https://doi.org/10.1109/TG.2018.2855734
6. Habaraduwa, U., Blom, P. M., & Wiltshire, T. J. (2024). Implicit Coordination Dynamics: A Synchrony-based Study on Team Positioning and Performance in Competitive Dota 2. IEEE Transactions on Games, 17(2), 442-450.
7. Hall, J., Stickler, U., Herodotou, C., & Iacovides, I. (2020). Player conceptualizations of creativity in digital entertainment games. Convergence, 26(5–6), 1226–1247. https://doi.org/10.1177/1354856518778477
8. Jennrich, R. I. (1977). Stepwise discriminant analysis. Statistical methods for digital conputers, 76.

9. Katona, A., Kiss, G., & Szűcs, G. (2019). Time to die: Death prediction in Dota 2 using deep learning. In 2019 IEEE Conference on Games (CoG) (pp. 1–8). IEEE. https://doi.org/10.1109/CIG.2019.8848028
10. Malyshev, Y. 2019. Dota 2 positions & roles guide – how to take the most of your position https://gosu.ai/blog/dota2/positions-and-roles-guide/
11. McClarty, K. L., Orr, A., Frey, P. M., Dolan, R. P., Vassileva, V., & McVay, A. (2012). A literature review of gaming in education. Pearson Research Report.
12. Nascimento Junior, F. F. D., Melo, A. S. D. C., Da Costa, I. B., & Marinho, L. B. (2017). Profiling successful team behaviors in League of Legends. In Proceedings of the 23rd Brazilian Symposium on Multimedia and the Web (pp. 261–268). ACM. https://doi.org/10.1145/3126858.3126885
13. Perez-Liebana, D., Samothrakis, S., Togelius, J., Lucas, S. M., & Schaul, T. (2016). General video game AI: Competition, challenges and opportunities. In Proceedings of the 30th AAAI Conference on Artificial Intelligence (pp. 4335–4337). AAAI Press.
14. Pradhan, S., & Abdourazakou, Y. (2020). “Power ranking” professional circuit eSports teams using multi-criteria decision-making (MCDM). Journal of Sports Analytics, 6(1), 61-73. https://doi.org/10.3233/JSA-190420
15. Qian, M., & Clark, K. R. (2016). Game-based learning and 21st-century skills: A review of recent research. Computers in Human Behavior, 63, 50–58. https://doi.org/10.1016/j.chb.2016.05.023
16. Samala, A. D., Bojić, L., Vergara-Rodríguez, D., Klimova, B., & Ranuharja, F. (2024). Exploring the impact of gamification on 21st-century skills: Insights from DOTA 2. International Journal of Interactive Mobile Technologies, 17(18), 33–54. https://doi.org/10.3991/ijim.v17i18.42963
17. ShiaoPi, JDC214, OmniEulogy, & Jul-must. (2019). The Dota 2 1-to-5 system: An introduction to farm priority. LiquidDota. https://www.liquiddota.com/forum/dota-2-strategy/545994-the-dota-2-1-to-5-system-an-introduction-to-farm-priority
18. Thompson, B. (1984). Canonical correlation analysis: Uses and interpretation (Vol. 47). Sage.
19. Todorov, V. (2007). Robust selection of variables in linear discriminant analysis. Computational Statistics, 15(3), 395–407. https://doi.org/10.1007/s10260-006-0032-6
20. Tyack, A., Wyeth, P., & Johnson, D. (2016). The appeal of MOBA games: What makes people start, stay, and stop. In Proceedings of the Annual Symposium on Computer-Human Interaction.
21. Uddin, J., Fahmida, I., Moyeen, S. I., & Hasan, M. M. (2022). DOTA2 winner team prediction based on stacked bidirectional LSTM network. In 2022 4th International Conference on Electrical, Computer & Telecommunication Engineering (ICECTE) (pp. 1–5). IEEE. https://doi.org/10.1109/ICECTE55303.2022.10007319
22. Valve Corporation. (2017). The International Dota 2 Championships. https://www.dota2.com/international/overview
23. Winn, C. (2015). The well-played MOBA: How Dota 2 and League of Legends use dramatic dynamics. In Proceedings of the DiGRA 2015 Conference.
24. Yangibaev, S., Mattiev, J., & Mokwena, S. (2025). DotA 2 Match Outcome Prediction System Using Decision Tree Ensemble Algorithms. Big Data and Cognitive Computing, 9(12), 302. https://doi.org/10.3390/bdcc9120302