Syracuse Can Throw It Until the Field Gets Short

Week 3 2026 — Syracuse 13 at Pittsburgh 27 — exhibits for publication

library(tidyverse)
library(arrow)
library(gt)
library(ggrepel)
library(scales)

options(cfb.lib_only = TRUE)
source("../R/02_team_game_metrics.R")   # classify_plays(), NEEDED
source("../R/04_storyline_engine.R")
source("../R/05_viz_style.R")

TEAM <- "Syracuse"; OPP <- "Pittsburgh"
SEASON <- params$season; WEEK <- params$week
FIG_DIR <- sprintf("figures/%d_wk%02d_story", SEASON, WEEK)
SCOPE   <- sprintf("%d Week %d | %s at %s", SEASON, WEEK, TEAM, OPP)

hist <- read_parquet("~/cfb-data/processed/team_game_metrics.parquet")
cur  <- read_parquet("~/cfb-data/processed/team_game_metrics_2026.parquet")
fbs  <- readRDS("~/cfb-data/fbs_teams_2026.rds")

load_week <- function(w) {
  p <- read_parquet(sprintf("~/cfb-data/raw/pbp_inseason/pbp_%d_wk%02d.parquet", SEASON, w))
  # Raw week files are cached BEFORE harmonisation, so divisions are rebuilt
  # here from the cached FBS team list.
  p$home_team_division <- ifelse(p$home %in% fbs, "fbs", "fcs")
  p$away_team_division <- ifelse(p$away %in% fbs, "fbs", "fcs")
  p$wk <- w
  p
}
raw <- map(1:WEEK, load_week)

su_plays <- map(raw, function(p) {
  p |> select(all_of(intersect(NEEDED, names(p))), wk) |>
    filter(pos_team == TEAM) |> classify_plays() |> filter(is_scrimmage)
}) |> list_rbind() |>
  mutate(type = ifelse(pass == 1, "Pass", "Rush"),
         zone = ifelse(yards_to_goal <= 40, "Inside opponent 40", "Outside the 40"))

peers <- cur |> filter(team %in% fbs, !low_volume, is_fbs_matchup, week == WEEK)
su3   <- cur |> filter(team == TEAM, week == WEEK)

Exhibit 1 — The passing game collapses in scoring range

# Weeks 2-3 only: both FBS opponents, so the comparison is clean. Week 1's FCS
# game is excluded here rather than averaged in.
zone_split <- su_plays |>
  filter(wk >= 2) |>
  group_by(zone, type) |>
  summarise(plays = n(), sr = mean(success, na.rm = TRUE),
            epa = mean(EPA, na.rm = TRUE), .groups = "drop") |>
  mutate(zone = factor(zone, levels = c("Outside the 40", "Inside opponent 40")))

p_zone <- ggplot(zone_split, aes(x = zone, y = epa, fill = type)) +
  geom_hline(yintercept = 0, color = SU$muted, linewidth = 0.4) +
  geom_col(position = position_dodge(width = 0.72), width = 0.62) +
  geom_text(aes(label = sprintf("%+.3f\n%.0f%% success\n%d plays", epa, 100*sr, plays),
                vjust = ifelse(epa >= 0, -0.25, 1.15)),
            position = position_dodge(width = 0.72),
            size = 3.1, fontface = "bold", lineheight = 0.95, color = SU$ink) +
  scale_fill_manual(values = c(Pass = SU$orange, Rush = SU$blue)) +
  scale_y_continuous(expand = expansion(mult = c(0.28, 0.28))) +
  labs(
    title = sprintf("%s throws well until the field gets short, then it falls apart", TEAM),
    subtitle = "EPA per play by field zone | Weeks 2-3 only, both FBS opponents | Higher is better",
    x = NULL, y = "EPA per play", fill = NULL,
    caption = su_caption("2026 Weeks 2-3 | vs California and Pittsburgh")
  ) +
  theme_su(legend_pos = "top")

save_fig(p_zone, "01_pass_collapse_in_scoring_range", height = 5)

Outside the 40, Syracuse’s passing game is above average. Inside it, the success rate is cut roughly in half and EPA per play falls by more than eight tenths of a point. The run game declines in scoring range too — every offense does, as the defense gains a shorter field to defend — but nothing like this.

Exhibit 2 — Three weeks, three steps backward

by_week <- su_plays |>
  group_by(wk, type) |>
  summarise(epa = mean(EPA, na.rm = TRUE), sr = mean(success, na.rm = TRUE),
            .groups = "drop")

ends <- by_week |> filter(wk == WEEK)

p_trend <- ggplot(by_week, aes(wk, epa, color = type)) +
  annotate("rect", xmin = 0.6, xmax = 1.4, ymin = -Inf, ymax = Inf,
           fill = SU$muted, alpha = 0.10) +
  annotate("text", x = 1, y = Inf, vjust = 1.6, label = "FCS opponent",
           size = 3, fontface = "italic", color = SU$ink2) +
  geom_hline(yintercept = 0, color = SU$muted, linewidth = 0.4) +
  geom_line(linewidth = 1) +
  geom_point(size = 3) +
  geom_text_repel(data = ends, aes(label = sprintf("%s %+.3f", type, epa)),
                  nudge_x = 0.18, size = 3.4, fontface = "bold",
                  segment.color = NA, force = 0) +
  scale_color_manual(values = c(Pass = SU$orange, Rush = SU$blue), guide = "none") +
  scale_x_continuous(breaks = 1:WEEK, labels = c("Wk 1\nNew Hampshire",
                                                 "Wk 2\nCalifornia",
                                                 "Wk 3\nPittsburgh")[1:WEEK],
                     expand = expansion(mult = c(0.05, 0.22))) +
  labs(
    title = "The run game recovered against Pitt. The passing game did not.",
    subtitle = sprintf(
      "EPA per play by game | Pass: %+.3f to %+.3f to %+.3f | Week 1 came against an FCS defense and is shaded accordingly",
      by_week$epa[by_week$type=="Pass" & by_week$wk==1],
      by_week$epa[by_week$type=="Pass" & by_week$wk==2],
      by_week$epa[by_week$type=="Pass" & by_week$wk==3]),
    x = NULL, y = "EPA per play",
    caption = su_caption("2026 Weeks 1-3")
  ) +
  theme_su(legend_pos = "none")

save_fig(p_trend, "02_pass_vs_rush_by_week", height = 5)

by_week |>
  pivot_wider(names_from = type, values_from = c(epa, sr)) |>
  left_join(su_plays |> filter(pass == 1) |> group_by(wk) |>
              summarise(sack_rate = mean(sack == 1, na.rm = TRUE), .groups = "drop"),
            by = "wk") |>
  transmute(Game = c("Wk 1 — New Hampshire (FCS)", "Wk 2 — California",
                     "Wk 3 — Pittsburgh")[wk],
            `Pass EPA` = epa_Pass, `Pass success` = sr_Pass,
            `Rush EPA` = epa_Rush, `Rush success` = sr_Rush,
            `Sack rate allowed` = sack_rate) |>
  gt() |>
  tab_header(
    title = md("**The decline is not a protection problem**"),
    subtitle = md("Sack rate allowed held steady while passing efficiency fell — the line is not the story")
  ) |>
  fmt_number(columns = c(`Pass EPA`, `Rush EPA`), decimals = 3) |>
  fmt_percent(columns = c(`Pass success`, `Rush success`, `Sack rate allowed`), decimals = 1) |>
  cols_align("center", columns = -Game) |>
  tab_style(style = cell_text(weight = "bold"), locations = cells_body(columns = Game)) |>
  tab_source_note(md(paste0("**", su_caption("2026 Weeks 1-3"), "**"))) |>
  gt_theme_su()

The decline is not a protection problem
Sack rate allowed held steady while passing efficiency fell — the line is not the story
Game	Pass EPA	Pass success	Rush EPA	Rush success	Sack rate allowed
Wk 1 — New Hampshire (FCS)	−0.112	51.4%	0.035	55.0%	8.1%
Wk 2 — California	−0.159	33.3%	−0.093	37.5%	5.6%
Wk 3 — Pittsburgh	−0.204	31.4%	0.130	41.7%	5.7%
Data: CollegeFootballData via cfbfastR | 2026 Weeks 1-3

Exhibit 3 — Where that ranks nationally

lg <- raw[[WEEK]] |>
  select(all_of(intersect(NEEDED, names(raw[[WEEK]]))),
         home_team_division, away_team_division) |>
  filter(home_team_division == "fbs", away_team_division == "fbs") |>
  classify_plays() |> filter(is_scrimmage, pass == 1) |>
  group_by(team = pos_team) |>
  summarise(n = n(), pass_epa = mean(EPA, na.rm = TRUE), .groups = "drop") |>
  filter(n >= 15) |>
  mutate(is_su = team == TEAM)

su_pass_rank <- as.integer(rank(-lg$pass_epa)[lg$team == TEAM])

lg <- lg |> arrange(desc(pass_epa)) |> mutate(team = factor(team, levels = team))

p_rank <- ggplot(lg, aes(x = pass_epa, y = team, color = is_su)) +
  geom_vline(xintercept = median(lg$pass_epa), linetype = "dashed",
             color = SU$muted, linewidth = 0.4) +
  geom_point(aes(size = is_su)) +
  geom_text_repel(data = filter(lg, is_su),
    aes(label = sprintf("%s — %+.3f EPA per dropback\n%s of %d",
                        team, pass_epa, scales::ordinal(su_pass_rank), nrow(lg))),
    nudge_x = -0.33, size = 3.4, fontface = "bold", segment.color = NA,
    lineheight = 0.95, force = 0) +
  scale_color_manual(values = c(`FALSE` = SU$muted, `TRUE` = SU$orange), guide = "none") +
  scale_size_manual(values = c(`FALSE` = 1.6, `TRUE` = 3.6), guide = "none") +
  labs(
    title = sprintf("%s ranked %s of %d in passing efficiency in Week 3",
                    TEAM, scales::ordinal(su_pass_rank), nrow(lg)),
    subtitle = sprintf("EPA per dropback | FBS teams with 15+ dropbacks against FBS opponents | League median %+.3f | Dashed line = median",
                       median(lg$pass_epa)),
    x = "EPA per dropback", y = NULL,
    caption = su_caption(SCOPE)
  ) +
  theme_su(legend_pos = "none") +
  theme(axis.text.y = element_text(size = 5.5))

save_fig(p_rank, "03_pass_epa_vs_peers", height = 8)

Exhibit 4 — What it cost

drives <- raw[[WEEK]] |>
  filter(pos_team == TEAM) |>
  group_by(drive_id) |>
  summarise(start = first(drive_start_yards_to_goal),
            closest = min(yards_to_goal, na.rm = TRUE),
            pts = first(drive_pts), result = first(drive_result),
            plays = n(), .groups = "drop") |>
  filter(!is.na(start), is.finite(closest))

opps <- drives |> filter(closest <= 40) |> arrange(closest)

pr_pts <- {
  v <- peers$off_pts_per_opp
  as.integer(round(rank(-v, na.last = "keep")[peers$team == TEAM]))
}

opps |>
  transmute(`Drive started` = sprintf("own %.0f", 100 - start),
            `Got as close as` = ifelse(closest == 0, "the goal line",
                                       sprintf("the %s %.0f", OPP, closest)),
            Plays = plays, Result = result, Points = pts) |>
  gt() |>
  tab_header(
    title = md(sprintf("**Seven trips inside the %s 40 produced 13 points**", OPP)),
    subtitle = md(sprintf(
      "%.2f points per scoring opportunity — %s of %d nationally, and the third straight weekly decline",
      su3$off_pts_per_opp, scales::ordinal(pr_pts), nrow(peers)))
  ) |>
  cols_align("center", columns = c(Plays, Points)) |>
  cols_align("left", columns = c(`Drive started`, `Got as close as`, Result)) |>
  data_color(columns = Points, method = "numeric", palette = SU_SEQ) |>
  tab_style(style = list(cell_fill(color = "#FFF3E6"), cell_text(weight = "bold")),
            locations = cells_body(rows = Points == 0)) |>
  tab_source_note(md(paste0("**", su_caption(SCOPE),
    "** — a scoring opportunity is any drive reaching the opponent's 40."))) |>
  gt_theme_su()

Seven trips inside the Pittsburgh 40 produced 13 points
1.86 points per scoring opportunity — 100th of 114 nationally, and the third straight weekly decline
Drive started	Got as close as	Plays	Result	Points
own 17	the goal line	14	FG	3
own 35	the Pittsburgh 1	13	TD	7
own 25	the Pittsburgh 10	7	MISSED FG	0
own 41	the Pittsburgh 16	11	FG	3
own 16	the Pittsburgh 28	15	PUNT	0
own 20	the Pittsburgh 37	12	DOWNS	0
own 25	the Pittsburgh 40	6	INT	0
Data: CollegeFootballData via cfbfastR | 2026 Week 3 | Syracuse at Pittsburgh — a scoring opportunity is any drive reaching the opponent’s 40.

Figures written to disk

figs <- list.files(FIG_DIR, pattern = "\\.png$", full.names = TRUE)
if (length(figs) == 0) cat("No figures written.\n") else {
  cat("300-dpi PNGs ready for the piece:\n\n")
  for (f in figs) cat("  ", f, "  (", round(file.info(f)$size/1024), " KB)\n", sep = "")
  cat("\nCopy the gt tables from the browser into Google Docs — they paste as\n")
  cat("native editable tables.\n")
}

## 300-dpi PNGs ready for the piece:
## 
##   figures/2026_wk03_story/01_pass_collapse_in_scoring_range.png  (127 KB)
##   figures/2026_wk03_story/02_pass_vs_rush_by_week.png  (151 KB)
##   figures/2026_wk03_story/03_pass_epa_vs_peers.png  (387 KB)
## 
## Copy the gt tables from the browser into Google Docs — they paste as
## native editable tables.