Update app.R

app.R

@@ -48,7 +48,7 @@ download_private_csv <- function(repo_id, filename) {
 t <- download_private_csv("TimStats/Passwords", "demo.csv")
 pass <- t[1,1]
 
-#model <- xgb.load('TimStuff2.model')
+model <- xgb.load('TimStuff2.model')
 
 # Helper functions
 download_and_process_image <- function(url) {
@@ -408,58 +408,21 @@ transform_statcast_data <- function(df,
   return(df)
 }
 
-calculatetimstuffall <- function(game) {
-  #game <- transform_statcast_data(game) %>% mutate(SADiff = calculate_SADiff(pfxx, pfxz, SpinAxis))
-  
-  # Filter and handle fastball calculations with error checking
-  # Get primary pitch info directly within the function
-  primary_pitch1 <- game %>% 
-    filter(pitch_name == "Sinker" | pitch_name == "Fastball" | pitch_name == "Cutter" | 
-             pitch_name == "FourSeamFastBall" | pitch_name == "OneSeamFastBall" | pitch_name == "TwoSeamFastBall") %>% 
-    group_by(`Pitcher Name`) %>% 
-    count(pitch_name) %>% 
-    summarise(max = max(n))
-  
-  primary_pitch2 <- game %>% 
-    filter(pitch_name == "Sinker" | pitch_name == "Fastball" | pitch_name == "Cutter" | 
-             pitch_name == "FourSeamFastBall" | pitch_name == "OneSeamFastBall" | pitch_name == "TwoSeamFastBall") %>% 
-    group_by(`Pitcher Name`, pitch_name) %>% 
-    count(pitch_name)
-  
-  primary_pitch <- left_join(primary_pitch1, primary_pitch2, by = c('Pitcher Name')) %>% 
-    filter(max == n) %>% 
-    select(`Pitcher Name`, pitch_name) %>% 
-    filter(pitch_name != "Undefined" & pitch_name != "Other") %>% 
-    mutate(Primary = pitch_name) %>% 
-    select(-pitch_name)
-  
-  game <- left_join(game, primary_pitch, by = c('Pitcher Name'))
-  
-  # Calculate primary pitch averages
-  primary_avgs <- game %>% 
-    filter(pitch_name == Primary) %>% 
-    group_by(`Pitcher Name`) %>% 
-    summarise(
-      Primary_Velo = mean(start_speed, na.rm = TRUE),
-      Primary_IVB = mean(IVB, na.rm = TRUE),
-      Primary_HB = mean(HB, na.rm = TRUE)
-    )
-  
-  # Join averages and calculate differences
+calculate_timstuff <- function(game) {
   game <- game %>%
-    left_join(primary_avgs, by = c('Pitcher Name')) %>%
     mutate(
-      velo_diff = round(ifelse(is.na(Primary_Velo), 0, start_speed - Primary_Velo), 1),
-      IVB_diff = round(ifelse(is.na(Primary_IVB), 0, IVB - Primary_IVB), 1),
-      HB_diff = round(ifelse(is.na(Primary_HB), 0, 
-                             ifelse(phand == "Left", -HB - Primary_HB, HB - Primary_HB)), 1)
+      VAA = calculate_VAA(vz0, ay, az, vy0, 50),
+      EAA = calculate_EAA(extension),
+      SADiff = calculate_SADiff(pfxX, pfxZ, spinDirection),
+      Pitch = pitch_name,
+      ishandL = ifelse(phand == "L", 1, 0)
     )
-  print(colSums(is.na(game)))
-  feature_vars <- c("start_speed", "IVB", "HB", "spin_rate",
-                    "z0", "x0", "extension", "velo_diff", "IVB_diff", "HB_diff")
+  
+  feature_vars <- c("ishandL", "start_speed", "IVB", "HB", "EAA", "x0", "z0", "spin_rate", "SADiff")
   complete_rows <- complete.cases(game[, feature_vars])
   game_complete <- game[complete_rows, ]
   game_na <- game[!complete_rows, ]
+
   game_na <- if(any(!complete_rows)) {
     na_rows <- game[!complete_rows, ]
     na_rows$TimStuff <- NA
@@ -471,156 +434,23 @@ calculatetimstuffall <- function(game) {
     empty_df
   }  
   
-  # Round the features to 1 decimal point
-  game_complete$start_speed <- round(game_complete$start_speed, 1)
-  game_complete$IVB <- round(game_complete$IVB, 1)
-  game_complete$HB <- abs(round(game_complete$HB, 1))
-  game_complete$spin_rate <- round(game_complete$spin_rate, 1)
-  game_complete$z0 <- round(game_complete$z0, 1)
-  game_complete$x0 <- abs(round(game_complete$x0, 1))
-  game_complete$extension <- round(game_complete$extension, 1)
-  game_complete$velo_diff <- round(game_complete$velo_diff, 1)
-  game_complete$IVB_diff <- round(game_complete$IVB_diff, 1)
-  game_complete$HB_diff <- round(game_complete$HB_diff, 1)
-  
-  # Split right-handed and left-handed pitchers
-  rhp <- game_complete %>% filter(phand == "R")
-  lhp <- game_complete %>% filter(phand == "L")
-  
-  # Process right-handed pitchers by pitch type
-  if(nrow(rhp) > 0) {
-    # Split by pitch type
-    rhp_ff <- rhp %>% filter(pitch_name %in% c("Fastball","FourSeamFastBall"))
-    rhp_si <- rhp %>% filter(pitch_name %in% c("OneSeamFastBall","Sinker","TwoSeamFastBall"))
-    rhp_ct <- rhp %>% filter(pitch_name == "Cutter")
-    rhp_sl <- rhp %>% filter(pitch_name %in% c("Slider","Sweeper"))
-    rhp_cb <- rhp %>% filter(pitch_name == "Curveball")
-    rhp_ch_spl <- rhp %>% filter(pitch_name %in% c("Changeup", "ChangeUp","Splitter","Knuckleball"))
-    
-    # Apply models for each pitch type with appropriate scaling
-    if(nrow(rhp_ff) > 0) {
-      rhp_ff$TimStuff <- scale_TimStuff(
-        predict(FF_model, as.matrix(cbind(
-          rhp_ff$start_speed, rhp_ff$IVB, rhp_ff$HB, 
-          rhp_ff$z0, rhp_ff$x0, rhp_ff$extension))), 
-        0.07666808, 0.02507584)
-    }
-    
-    if(nrow(rhp_si) > 0) {
-      rhp_si$TimStuff <- scale_TimStuff(
-        predict(SI_model, as.matrix(cbind(
-          rhp_si$start_speed, rhp_si$IVB, rhp_si$HB, 
-          rhp_si$z0, rhp_si$x0, rhp_si$extension))), 
-        0.09427338, 0.01524531)
-    }
-    
-    if(nrow(rhp_ct) > 0) {
-      rhp_ct$TimStuff <- scale_TimStuff(
-        predict(CT_model, as.matrix(cbind(
-          rhp_ct$start_speed, rhp_ct$IVB, rhp_ct$HB, 
-          rhp_ct$z0, rhp_ct$x0, rhp_ct$extension))), 
-        0.01484248, 0.0001241943)
-    }
-    
-    if(nrow(rhp_sl) > 0) {
-      rhp_sl$TimStuff <- scale_TimStuff(
-        predict(SL_model, as.matrix(cbind(
-          rhp_sl$start_speed, rhp_sl$IVB, rhp_sl$HB, 
-          rhp_sl$spin_rate, rhp_sl$z0, rhp_sl$x0, rhp_sl$extension))), 
-        0.06669426, 0.01006894)
-    }
-    
-    if(nrow(rhp_cb) > 0) {
-      rhp_cb$TimStuff <- scale_TimStuff(
-        predict(CB_model, as.matrix(cbind(
-          rhp_cb$start_speed, rhp_cb$IVB, rhp_cb$HB, 
-          rhp_cb$spin_rate, rhp_cb$z0, rhp_cb$x0, rhp_cb$extension))), 
-        0.07480712, 0.01038754)
-    }
-    
-    if(nrow(rhp_ch_spl) > 0) {
-      rhp_ch_spl$TimStuff <- scale_TimStuff(
-        predict(CH_SPL_model, as.matrix(cbind(
-          rhp_ch_spl$start_speed, rhp_ch_spl$IVB, rhp_ch_spl$HB, 
-          rhp_ch_spl$z0, rhp_ch_spl$x0, rhp_ch_spl$extension, 
-          rhp_ch_spl$velo_diff, rhp_ch_spl$IVB_diff, rhp_ch_spl$HB_diff))), 
-        0.09255855, 0.01739746)
-    }
-    
-    # Combine all RHP pitch types
-    rhp <- rbind(rhp_ff, rhp_si, rhp_ct, rhp_sl, rhp_cb, rhp_ch_spl)
-    print("Rhp")
-    print(rhp)
+  if(nrow(game_complete) > 0) {
+    pred_matrix <- as.matrix(game_complete[, feature_vars])
+    predictions <- predict(model, pred_matrix)
+    game_complete$TimStuff <- scale_TimStuff(predictions, -0.002620635, 0.006021368)
   }
   
-  # Process left-handed pitchers by pitch type
-  if(nrow(lhp) > 0) {
-    # Split by pitch type
-    lhp_ff <- lhp %>% filter(pitch_name %in% c("Fastball","FourSeamFastBall"))
-    lhp_si <- lhp %>% filter(pitch_name %in% c("OneSeamFastBall","Sinker","TwoSeamFastBall"))
-    lhp_ct <- lhp %>% filter(pitch_name == "Cutter")
-    lhp_sl <- lhp %>% filter(pitch_name %in% c("Slider","Sweeper"))
-    lhp_cb <- lhp %>% filter(pitch_name == "Curveball")
-    lhp_ch_spl <- lhp %>% filter(pitch_name %in% c("Changeup", "ChangeUp","Splitter","Knuckleball"))
-    
-    # Apply models for each pitch type with appropriate scaling
-    if(nrow(lhp_ff) > 0) {
-      lhp_ff$TimStuff <- scale_TimStuff(
-        predict(FF_model, as.matrix(cbind(
-          lhp_ff$start_speed, lhp_ff$IVB, lhp_ff$HB, 
-          lhp_ff$z0, lhp_ff$x0, lhp_ff$extension))), 
-        0.07666808, 0.02507584)
-    }
-    
-    if(nrow(lhp_si) > 0) {
-      lhp_si$TimStuff <- scale_TimStuff(
-        predict(SI_model, as.matrix(cbind(
-          lhp_si$start_speed, lhp_si$IVB, lhp_si$HB, 
-          lhp_si$z0, lhp_si$x0, lhp_si$extension))), 
-        0.09427338, 0.01524531)
-    }
-    
-    if(nrow(lhp_ct) > 0) {
-      lhp_ct$TimStuff <- scale_TimStuff(
-        predict(CT_model, as.matrix(cbind(
-          lhp_ct$start_speed, lhp_ct$IVB, lhp_ct$HB, 
-          lhp_ct$z0, lhp_ct$x0, lhp_ct$extension))), 
-        0.01484248, 0.0001241943)
-    }
-    
-    if(nrow(lhp_sl) > 0) {
-      lhp_sl$TimStuff <- scale_TimStuff(
-        predict(SL_model, as.matrix(cbind(
-          lhp_sl$start_speed, lhp_sl$IVB, lhp_sl$HB, 
-          lhp_sl$spin_rate, lhp_sl$z0, lhp_sl$x0, lhp_sl$extension))), 
-        0.06669426, 0.01006894)
-    }
-    
-    if(nrow(lhp_cb) > 0) {
-      lhp_cb$TimStuff <- scale_TimStuff(
-        predict(CB_model, as.matrix(cbind(
-          lhp_cb$start_speed, lhp_cb$IVB, lhp_cb$HB, 
-          lhp_cb$spin_rate, lhp_cb$z0, lhp_cb$x0, lhp_cb$extension))), 
-        0.07480712, 0.01038754)
-    }
-    
-    if(nrow(lhp_ch_spl) > 0) {
-      lhp_ch_spl$TimStuff <- scale_TimStuff(
-        predict(CH_SPL_model, as.matrix(cbind(
-          lhp_ch_spl$start_speed, lhp_ch_spl$IVB, lhp_ch_spl$HB, 
-          lhp_ch_spl$z0, lhp_ch_spl$x0, lhp_ch_spl$extension, 
-          lhp_ch_spl$velo_diff, lhp_ch_spl$IVB_diff, lhp_ch_spl$HB_diff))), 
-        0.09255855, 0.01739746)
-    }
-    
-    # Combine all LHP pitch types
-    lhp <- rbind(lhp_ff, lhp_si, lhp_ct, lhp_sl, lhp_cb, lhp_ch_spl)
+  if(nrow(game_complete) > 0 && nrow(game_na) > 0) {
+    game_result <- rbind(game_complete, game_na)
+  } else if(nrow(game_complete) > 0) {
+    game_result <- game_complete
+  } else if(nrow(game_na) > 0) {
+    game_result <- game_na
+  } else {
+    game_result <- game
   }
   
-  # Combine all processed data
-  game_complete <- rbind(rhp, lhp, game_na)
-  
-  return(game_complete)
+  return(game_result)
 }
 summary_table <- function(game) {
   game <- calculatetimstuffall(game)