• Goal: Predict match outcomes using player performance metrics
• Approach: Logistic regression + visualization

• Simulated dataset (100 players)

• Variables:

  • Kills Deaths Assists (KDA)
  • Gold per Minute (GPM)
  • Damage per Minute (DPM)
  • Vision Score
  • Win/Loss

ggplot(data, aes(x = KDA, y = Win)) +
  geom_point(alpha = 0.6) +
  geom_smooth(method = "lm", color = "blue") +
  labs(title = "KDA vs Win Probability")

ggplot(data, aes(x = GPM, y = DPM, color = factor(Win))) +
  geom_point() +
  labs(title = "GPM vs DPM by Match Outcome",
       color = "Win")

\[ P(Y=1) = \frac{1}{1 + e^{-(\beta_0 + \beta_1 KDA + \beta_2 GPM + \beta_3 DPM + \beta_4 Vision)}} \]

• Models probability of winning
  
• Uses multiple performance metrics

model <- glm(Win ~ KDA + GPM + DPM + Vision, 
             data = data, family = "binomial")
summary(model)

\[ \log\left(\frac{P(Y=1)}{1 - P(Y=1)}\right) = \beta_0 + \beta_1 KDA + \beta_2 GPM + \beta_3 DPM + \beta_4 Vision \]

• Transforms probability into log-odds
  
• Linear relationship with predictors

• Positive coefficients = Increase in win probability
• GPM and DPM typically strongest predictors
• KDA contributes but less dominant. Likely a byproduct.

• Higher gold and damage strongly linked to winning
• Vision has moderate impact
• Model estimates are unlikely to be predictive, since more variables are dynamic throughout the game

• Statistical modeling helps evaluate player performance
• Logistic regression accurate with extreme hypothetical stats
• Can extend to real esports datasets (e.g., pro matches)