Shot Chart Analysis

Use this guide when you want to turn fact_shot_chart into a reliable film-room view of where shots came from and how efficient they were.

Pick the first question

If you need…	Start here	Finish with
A fast zone read before plotting	Zone efficiency query	A table you can sort by attempts or FG%
A first visual sanity check	First-pass shot chart	A half-court map with makes and misses
Hot and cold area reads	Efficiency heatmap	A location-based FG% view
Range profile by distance	Distance read	A table you can bin or chart later

Filter to one player and one season for your first pass. Get the court rendering right before you widen the sample to teams, lineups, or league-wide heatmaps.

Scope the possession first

Before plotting anything, lock in three choices:

Who are you charting? One player, one team, or the whole league
When are you charting them? A specific season_year and, usually, a season_type
What question are you answering? Volume, efficiency, zone splits, or location clustering

What `fact_shot_chart` gives you

Each shot record includes four main buckets of context:

Bucket	Useful columns
Location	`loc_x`, `loc_y`
Zones	`shot_zone_basic`, `shot_zone_area`, `shot_zone_range`
Result	`shot_made_flag`, `event_type`, `shot_type`
Game context	`period`, `minutes_remaining`, `seconds_remaining`, `shot_distance`, `game_id`, `player_id`

Court coordinates refresher

The NBA shot-coordinate frame used here is the standard half-court orientation:

Origin (0, 0) is at the basket
loc_x runs left to right
loc_y runs baseline toward midcourt
A first-pass plot usually focuses on the offensive half court rather than the full floor

Build the analysis in four passes

Pass 1 — Zone efficiency query

Start here when you want the cleanest table before you touch plotting code.

import duckdb

conn = duckdb.connect("nbadb/nba.duckdb")

zones = conn.sql("""
    SELECT
        p.first_name || ' ' || p.last_name AS player,
        sc.shot_zone_basic,
        COUNT(*) AS attempts,
        SUM(sc.shot_made_flag) AS makes,
        ROUND(100.0 * SUM(sc.shot_made_flag) / COUNT(*), 1) AS fg_pct,
        ROUND(
            CASE
                WHEN sc.shot_type = '3PT Field Goal'
                THEN 3.0 * SUM(sc.shot_made_flag) / COUNT(*)
                ELSE 2.0 * SUM(sc.shot_made_flag) / COUNT(*)
            END,
            3
        ) AS pts_per_attempt
    FROM fact_shot_chart sc
    JOIN dim_player p ON sc.player_id = p.player_id
    JOIN dim_game g ON sc.game_id = g.game_id
    WHERE p.last_name = 'Curry'
      AND p.first_name = 'Stephen'
      AND g.season_year = 2024
    GROUP BY 1, 2, sc.shot_type
    ORDER BY attempts DESC
""").pl()

print(zones)

Pass 2 — Build a first-pass shot chart

Start here when you want to confirm the coordinate frame and basic court rendering.

import matplotlib.pyplot as plt
import matplotlib.patches as patches
import polars as pl

shots = conn.sql("""
    SELECT loc_x, loc_y, shot_made_flag
    FROM fact_shot_chart sc
    JOIN dim_player p ON sc.player_id = p.player_id
    JOIN dim_game g ON sc.game_id = g.game_id
    WHERE p.last_name = 'Curry'
      AND p.first_name = 'Stephen'
      AND g.season_year = 2024
""").pl()

fig, ax = plt.subplots(figsize=(12, 11))

made = shots.filter(pl.col("shot_made_flag") == 1)
missed = shots.filter(pl.col("shot_made_flag") == 0)

ax.scatter(missed["loc_x"], missed["loc_y"], c="red", alpha=0.3, s=10, label="Missed")
ax.scatter(made["loc_x"], made["loc_y"], c="green", alpha=0.3, s=10, label="Made")

# Minimal half-court frame
court = patches.Rectangle((-250, -50), 500, 520, fill=False)
hoop = patches.Circle((0, 0), 7.5, fill=False)
ax.add_patch(court)
ax.add_patch(hoop)

ax.set_xlim(-260, 260)
ax.set_ylim(-60, 480)
ax.set_aspect("equal")
ax.legend()
ax.set_title("Stephen Curry Shot Chart 2024-25")
plt.show()

Pass 3 — Convert it into an efficiency heatmap

Start here once the court frame looks right and you want a denser efficiency read.

fig, ax = plt.subplots(figsize=(12, 11))

hb = ax.hexbin(
    shots["loc_x"],
    shots["loc_y"],
    C=shots["shot_made_flag"],
    reduce_C_function=lambda x: sum(x) / len(x),
    gridsize=25,
    cmap="RdYlGn",
    mincnt=3,
)

plt.colorbar(hb, label="FG%")
ax.set_xlim(-260, 260)
ax.set_ylim(-60, 480)
ax.set_aspect("equal")
ax.set_title("Shot Efficiency Heatmap")
plt.show()

Pass 4 — Add a distance read

Start here when the real question is range profile rather than just court position.

distance = conn.sql("""
    SELECT
        shot_distance,
        COUNT(*) AS attempts,
        SUM(shot_made_flag) AS makes,
        ROUND(100.0 * SUM(shot_made_flag) / COUNT(*), 1) AS fg_pct
    FROM fact_shot_chart sc
    JOIN dim_game g ON sc.game_id = g.game_id
    WHERE g.season_year = 2024
      AND g.season_type = 'Regular Season'
    GROUP BY 1
    ORDER BY 1
""").pl()

Sanity checks before you trust the chart

Check	Why
Confirm the player, season, and season type filter	Most misleading charts are really scope mistakes
Join `dim_game` when you need season-level filtering	Shot tables alone do not carry every analysis filter you need
Validate the coordinate frame before tuning colors	A beautiful chart on the wrong floor is still wrong
Say when you mix `Regular Season` and `Playoffs`	Otherwise the read is not comparable possession-to-possession

DuckDB Query Examples for more SQL patterns
Parquet Usage if you want to run the same drill over Parquet files
Schema Reference / Facts if you need to compare fact_shot_chart against nearby fact tables

Shot Chart Analysis

Shot Chart Analysis

Pick the first question

Scope the possession first

What `fact_shot_chart` gives you

Court coordinates refresher

Build the analysis in four passes

Pass 1 — Zone efficiency query

Pass 2 — Build a first-pass shot chart

Pass 3 — Convert it into an efficiency heatmap

Pass 4 — Add a distance read

Sanity checks before you trust the chart

Stay in the same possession

Analytics Quickstart

Visual Asset Prompt Pack

On this page