Fix drive sequence number

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: nflfastR
 Title: Functions to Efficiently Access NFL Play by Play Data
-Version: 4.5.1.9005
+Version: 4.5.1.9006
 Authors@R: 
     c(person(given = "Sebastian",
              family = "Carl",

NEWS.md

@@ -6,6 +6,8 @@
 - The function `calculate_player_stats_def()` no longer errors when small subsets of pbp data are missing stats. (#415)
 - The function `calculate_series_conversion_rates()` no longer returns `NA` values if a small subset of pbp data is missing series on offense or defense. (#417)
 - `fixed_drive` now correctly increments on plays where posteam lost a fumble but remains posteam because defteam also lost a fumble during the same play. (#419)
+- nflfastR now fixes missing drive number counts in raw pbp data in order to provide accurate drive information. (#420)
+
 
 # nflfastR 4.5.1
 

R/helper_additional_functions.R

@@ -55,57 +55,6 @@ clean_pbp <- function(pbp, ...) {
   } else {
     user_message("Cleaning up play-by-play...", "todo")
 
-    if(any(pbp$season >= 2022)){
-
-      # user_message("Loading pbp player ID patch files", "info")
-
-      patch_seasons <- unique(pbp$season[pbp$season >= 2022])
-
-      patch_ids <- nflreadr::load_from_url(
-        glue::glue("https://github.com/nflverse/nflverse-data/releases/download/misc/pbp_patch_ids_{patch_seasons}.rds")
-      ) %>% suppressMessages()
-
-      patchable_ids <- pbp  %>%
-        dplyr::select(
-          dplyr::any_of(c(
-            "game_id", "play_id",
-            "passer_id", "passer_name" = "passer",
-            "receiver_id", "receiver_name" = "receiver",
-            "rusher_id", "rusher_name" = "rusher",
-            "fantasy_id", "fantasy_name" = "fantasy",
-            "fantasy_player_name"
-          )),
-          dplyr::matches("player_id|player_name")
-        )  %>%
-        tidyr::pivot_longer(
-          cols = -c("game_id","play_id"),
-          names_to = c("stat",".value"),
-          names_pattern = c("(.+)_(id|name)"),
-          values_drop_na = TRUE
-        )  %>%
-        dplyr::filter(is.na(.data$id)) %>%
-        dplyr::left_join(patch_ids, by = c("game_id","play_id","name"))  %>%
-        dplyr::mutate(
-          id = dplyr::coalesce(.data$id,.data$gsis_id),
-          gsis_id = NULL,
-          club_code = NULL,
-          name = NULL
-        )  %>%
-        tidyr::pivot_wider(
-          names_from = "stat",
-          values_from = "id",
-          names_glue = "{stat}_id"
-        )
-
-      if(nrow(patchable_ids) > 0){
-        pbp <- tibble::tibble(pbp)  %>%
-          dplyr::rows_patch(patchable_ids, by = c("game_id","play_id"))
-      }
-
-      # cli::cli_alert_success("{my_time()} | Patched {nrow(patchable_ids)} missing gsis_id field{?s}")
-
-    }
-
     # drop existing values of clean_pbp
     pbp <- pbp %>% dplyr::select(-tidyselect::any_of(drop.cols))
 

R/helper_scrape_nfl.R

@@ -108,6 +108,32 @@ get_pbp_nfl <- function(id, dir = NULL, qs = FALSE, ...) {
 
       plays <- raw_data$data$viewer$gameDetail$plays %>% dplyr::mutate(game_id = as.character(game_id))
 
+      # We have this issue https://github.com/nflverse/nflfastR/issues/309 with 2013 postseason games
+      # where the driveSequenceNumber in the plays df is NA for all plays. That prevents drive information
+      # from being joined.
+      # In this case, we compute our own driveSequenceNumber by incrementing a counter depending on the
+      # value of driveTimeOfPossession.
+      # driveTimeOfPossession will be a constant value during a drive so this should actually be accurate
+      if (all(is.na(plays$driveSequenceNumber))){
+        plays <- plays %>%
+          dplyr::mutate(
+            # First, create a trigger for cumsum
+            drive_trigger = dplyr::case_when(
+              # this is the first play of the first drive
+              is.na(dplyr::lag(driveTimeOfPossession)) & !is.na(driveTimeOfPossession) ~ 1,
+              # if driveTimeOfPossession changes, there is a new drive
+              dplyr::lag(driveTimeOfPossession) != driveTimeOfPossession ~ 1,
+              TRUE ~ 0
+            ),
+            # Now create the drive number by accumulationg triggers
+            driveSequenceNumber = cumsum(drive_trigger),
+            # driveSequenceNumber should be NA on plays where driveTimeOfPossession is NA
+            driveSequenceNumber = ifelse(is.na(driveTimeOfPossession), NA_real_, driveSequenceNumber),
+            # drop the helper
+            drive_trigger = NULL
+          )
+      }
+
       #fill missing posteam info for this
       if (
         ((home_team %in% c("JAC", "JAX") | away_team %in% c("JAC", "JAX")) & season <= 2015) |