Reading the Run Game

The question: Who carried the Lions’ run game in 2025, and who was actually efficient doing it?

This is the first rep every analyst runs: turn a pile of play-by-play rows into a clean leaderboard, then read it honestly. Volume and efficiency are not the same stat — a back with 1,000 yards on 280 carries is a different story than one with 600 yards on 90.

Install (the concept)

Raw football data comes one row per player per week. To get a season leaderboard you have to collapse those rows into one row per player. That’s three moves:

GROUP BY player — one bucket per player
SUM(...) — add up the weekly numbers inside each bucket
ORDER BY ... DESC — put the leaders on top

Then one analyst’s instinct: a rate stat. Total yards rewards whoever touched the ball most. Yards per carry (ypc) tells you who did the most with each touch. We compute both.

Note: column names in nflverse extracts occasionally differ (recent_team vs team, etc.). If a column errors, run DESCRIBE SELECT * FROM player_stats_2025; in The Film Room to see the real names, then adjust.

Run the drill

SELECT
  player_display_name        AS player,
  position,
  SUM(carries)               AS carries,
  SUM(rushing_yards)         AS rush_yds,
  ROUND(SUM(rushing_yards) * 1.0 / NULLIF(SUM(carries), 0), 2) AS ypc
FROM player_stats_2025
WHERE recent_team = 'DET'
GROUP BY player_display_name, position
HAVING SUM(carries) >= 20      -- drop noise: tiny sample sizes
ORDER BY rush_yds DESC;

Why the pieces matter

NULLIF(SUM(carries), 0) guards against divide-by-zero (a receiver with a single end-around).
HAVING filters after aggregation — it trims players with too few carries to read into. WHERE can’t do that; it runs before the SUM exists.

Read the result

Look for the split between volume and rate. The top of the rush_yds column is your workhorse. Now re-sort the same result by ypc in your head — does the order change? When the volume leader and the efficiency leader are different players, that gap is the actual analytical finding. That’s the sentence you’d put in front of a coordinator.

Visualize it

A leaderboard is a ranking, and rankings read best as a horizontal bar chart: long bars = more yards, sorted top-to-bottom. Here’s the Vega-Lite spec the Film Room renders from your result set:

{
  "$schema": "https://vega.github.io/schema/vega-lite/v5.json",
  "data": { "name": "rows" },
  "mark": { "type": "bar", "cornerRadiusEnd": 3 },
  "encoding": {
    "y": {
      "field": "player",
      "type": "nominal",
      "sort": "-x",
      "title": null
    },
    "x": {
      "field": "rush_yds",
      "type": "quantitative",
      "title": "Rushing Yards — 2025"
    },
    "color": { "value": "#0076B6" }
  },
  "config": {
    "background": null,
    "view": { "stroke": null },
    "axis": { "labelFont": "JetBrains Mono", "titleFont": "JetBrains Mono", "grid": false },
    "font": "Saira"
  }
}

The lesson inside the spec — encoding is everything:

y is nominal (categories: players). Categories go on the axis you can stack vertically and label.
x is quantitative (a magnitude: yards). Magnitude maps to bar length — the thing your eye compares fastest.
sort: "-x" sorts the categories by the quantitative value, descending. This one line is what turns a bar chart into a leaderboard.

If you swap which field goes on x vs y, or change a type, the chart changes meaning. That’s the whole grammar: you’re not “making a bar chart,” you’re mapping data fields to visual channels.

Run the next rep (challenge)

Color by position. Replace the color block with:
```
"color": { "field": "position", "type": "nominal", "scale": { "range": ["#0076B6", "#B0B7BC", "#34E27A"] } }
```
Now you can see at a glance whether the yards came from RBs or got spread to other positions.
Switch the story to efficiency. Change the x field from "rush_yds" to "ypc" and the title to "Yards per Carry". Re-run. Does the leaderboard reorder? Write one sentence explaining the gap between the volume chart and the efficiency chart — that sentence is your first piece of analysis.

Unlock: finish both challenges to bank this rep toward Scout.