A Dungeons & Dragons data story · Jake Biondolillo

The Spellcaster’s
Compendium

A character sheet is a small record of an imagined life. What patterns appear when we read thousands of them together?

In Dungeons & Dragons, a class gives a character a set of abilities. This story focuses on six classes closely associated with magic—from a Wizard’s study to a Warlock’s pact.

BardClericDruidSorcererWarlockWizard
  1. 01 / Who’s in the archive?
  2. 02 / What’s on the spell sheet?
  3. 03 / What changes when zeros count?
01 /

The spellcasting classes

Wizards and Clerics are the largest groups in this six-class archive. Each accounts for less than one in five character IDs. The rest of the picture is spread across four other ways of making magic.

Character IDs by Starting Class

Count and share of 223,284 character IDs across six starting classes

  • Wizard41,906 18.77%
  • Cleric41,219 18.46%
  • Warlock37,051 16.59%
  • Bard36,829 16.49%
  • Sorcerer34,281 15.35%
  • Druid31,998 14.33%

One latest record per character ID, with five unresolved IDs excluded before the class filter. Multiclass records stay under their starting class.

02 /

Spells on the character sheet

A cantrip is a spell of level 0. These are the most commonly listed cantrips for each of the six classes in the smaller sheet collection. Vicious Mockery and Eldritch Blast stand out within the lists we can read.

Here, the unit changes to sheet records. A spell counts once per sheet, even if it appears more than once in its list.

Leading Listed Cantrip in Each Class

Cantrips (Spell Level 0)

Each denominator is that class’s sheets with at least one parseable spell at this level.

  • BardVicious Mockery170 / 211 sheets · 80.57%
  • ClericSacred Flame202 / 362 sheets · 55.80%
  • DruidShillelagh76 / 229 sheets · 33.19%
  • SorcererFire Bolt121 / 212 sheets · 57.08%
  • WarlockEldritch Blast180 / 250 sheets · 72.00%
  • WizardFire Bolt140 / 245 sheets · 57.14%

The denominator is sheets in that class with at least one parseable level-0 spell—not every class record. All-class rates and levels 1–2 are in the appendix.

03 /

A small decision
reverses the story.

The larger archive includes a note-length field. If we look only at records with positive lengths, Warlocks have the highest mean. Include valid recorded zeros, and Warlocks have the lowest.

That reversal says more about the measurement and its denominator than it can tell us about the people behind the sheets.

716.36WARLOCK MEAN · POSITIVE ONLY

Highest of the six classes.

226.88WARLOCK MEAN · INCLUDING ZEROS

Lowest of the six classes.

Mean Recorded Note Length, under Two Policies
Including recorded zerosPositive lengths only
  • BardIncluding zeros 267.06 (N=36,826)Positive only 687.66 (N=14,302)
  • ClericIncluding zeros 249.41 (N=41,216)Positive only 599.11 (N=17,158)
  • DruidIncluding zeros 256.93 (N=31,995)Positive only 635.28 (N=12,940)
  • SorcererIncluding zeros 233.25 (N=34,277)Positive only 623.44 (N=12,824)
  • WarlockIncluding zeros 226.88 (N=37,049)Positive only 716.36 (N=11,734)
  • WizardIncluding zeros 241.01 (N=41,903)Positive only 589.27 (N=17,138)

Each line connects two summaries of the same class, not a change over time. Negative values are excluded. The exact construction and unit of notes_len are undocumented.

Interpreting zero values

Zero may mean no notes were recorded—or that the sheet was incomplete. The source does not resolve that ambiguity, so both policies stay visible.

All six medians are zero in the inclusive view. The precise source unit is unknown, and note length cannot measure storytelling effort or personality.

Distributions, exclusions & full table →

Reflection

This project began as an exploration of D&D character creation. Working with the data revealed important limits: repeated rows are not new characters, levels are not a timeline, and a missing field is not a motive.

The data reveals several patterns. Some labels recur, some spells are common in recorded lists, and a simple filtering choice can reverse a ranking. Making those choices visible gives the reader a way to judge the evidence.

These are selected historical online records, not a representative survey of D&D players. Campaign context, content access and incomplete sheets remain outside the data.

Explore All Eight Visuals ↗

Class · level · race · background · subclass · spell · gold · notes

Read the methodology and sources