Player Projections

🏀 Select Player

🔧 Projection Feature Categories

⚖️ Weighting Strategies

R Code Equivalent

# Player projection model
library(dplyr)

# Feature engineering
player_features <- player_games %>%
  group_by(player_id) %>%
  mutate(
    # Rolling averages
    season_avg = mean(points),
    last5_avg = zoo::rollmean(points, k = 5, fill = NA, align = "right"),
    
    # Recency weighting (exponential)
    recency_weight = 0.9 ^ row_number(),
    weighted_avg = sum(points * recency_weight) / sum(recency_weight)
  ) %>%
  ungroup()

# Build projection
build_projection <- function(player, game_context) { 
  proj <- 0.4 * player$season_avg +
          0.4 * player$last5_avg +
          0.2 * (player$season_avg + game_context$matchup_adj)
  
  # Apply adjustments
  proj <- proj + game_context$minutes_adj * 0.8
  proj <- proj + (game_context$vegas_total - 220) / 20 * 1.5
  
  return(proj)
}

# Simulate distribution
simulate_player <- function(projection, std_dev, n = 10000) { 
  sims <- rnorm(n, mean = projection, sd = std_dev)
  over_prob <- mean(sims > player$line)
  return(list(projection = projection, over_prob = over_prob, sims = sims))
}