Origins of save scumming
Read the illustrated story: Just one more reload.
Players were undoing unwanted outcomes long before they called it save scumming. A February 1981 review of Zork already describes saving before dangerous moves and disparages the players who do it. The exact expression save-scumming is documented in Angband’s Usenet community by July 1996, well before the 2003 example sometimes presented as its origin. Neither date establishes an invention or a first-ever use. The tension was already familiar. Saving can preserve a session, support experimentation, or let a player reject consequences that a game expects them to accept. (BYTE’s contemporary Zork review, July 1996 Angband discussion)
| Date | Documented evidence | What it establishes |
|---|---|---|
| February 1981 | BYTE describes saving before danger in Zork. | The behavior and criticism of it already existed. |
| June 1981 revision header | Preserved Rogue 3.6 source resists repeated restoration. | Early software deliberately distinguished resuming from retrying. |
| December 1995 | Angband’s FAQ discusses backup resurrection and stair-scumming. | The practice sits within a wider vocabulary of seeking favorable results. |
| July 1996 | An Angband Usenet exchange uses save-scumming. | The expression was in use by this date. |
| April 2002 | Civilization III patch notes add a selectable random-seed rule. | Rerolling becomes an explicit player option. |
| October 2013 | An XCOM: Enemy Within designer describes its “Save Scum” option. | Community slang becomes an official feature name. |
The distinction between these dates matters. Merely providing a save command does not prove that people repeatedly restored to obtain better results. Saving immediately before danger and returning after failure is another step. Repeatedly rerolling a battle or copying a supposedly permanent character’s save file introduces different expectations again. A history that treats all four as interchangeable will assign the practice an unjustifiably precise birthday.
Saving before a dangerous move
The technical background reaches into the 1970s. Mike Markowski’s documentation of the preserved 1977 Adventure describes a SUSPEND facility based on saving a process memory image. That is evidence for the machinery of interrupted play, not contemporary testimony that players used it to reverse bad outcomes. It therefore belongs to the prehistory of save scumming, without establishing Adventure as its inventor. (Adventure source and port documentation)
By 1981, players were already arguing over the legitimate purpose of a save. In BYTE’s February issue, Bob Liddil describes Zork players preserving their position before dangerous moves and calls them “cowards.” He presents continuing a long game and protecting against power failures as the proper uses. Liddil says he played an advance copy without instructions. His verdict tells us how one reviewer understood the rules. (BYTE, February 1981, p. 262)
An early Apple II Zork manual gives the opposite advice. It explicitly encourages snapshots around difficult or dangerous situations and restoration after getting lost or killed, describing the user as a “cautious or prudent player.” The museum identifies the artifact as the first Infocom-branded release, although its precise printing year is uncertain. The advice in the manual shows that recovering from experiments could be an intended way to play. Modern definitions broad enough to include this behavior reach back into a tradition that officially welcomed it. (Zork manual, pp. 8–9, museum artifact record)
One dungeon, one attempt
Rogue created a different relationship with saved progress. Preserved Rogue 3.6 source bearing a June 16, 1981 revision header checks whether a save has been manipulated and removes it when ordinary play resumes. A comment states the purpose: “defeat multiple restarting from the same place.” Saving therefore supports interruption, while the program actively resists using that saved position repeatedly. A player seeking do-overs must work around the intended lifecycle. The June 1981 revision dates these safeguards. Which ones the original 1980 release already included remains uncertain. (Rogue archive and provenance, archival 3.6 source, save.c)
Co-creator Glenn Wichman’s retrospective account helps explain the attraction of this approach. He and Michael Toy wanted a dungeon generated by the computer so that even its creators could encounter surprises, unlike fixed adventure puzzles whose solutions their programmers already knew. That structure offers a reason to accept a failed run. Another attempt can produce a different adventure. That helps explain the design tradition, but Wichman’s account does not establish the motivation for each individual save-file restriction. (Wichman, A Brief History of Rogue)
When Angband players called it save scumming
Jason Holtzapple’s Angband FAQ, posted December 15, 1995, describes restoring external copies after death as backup savefile abuse. It also explains stair-scumming: repeatedly generating dungeon levels until a desirable one appears. A November 1998 discussion explicitly groups stair-, town-, and save-file-scumming together. These sources support reading save scumming as one member of a family of techniques for exploiting or repeatedly seeking favorable conditions. They do not establish which compound came first, why the word scum was originally chosen, or a connection with LucasArts’ SCUMM engine. (1995 Angband FAQ, 1998 explanation of scumming)
The earliest exact expression located appears in the July 1996 Angband thread “YAPD.” A July 19 reply quotes a message attributed to John May on July 18 containing “without save-scumming.” A separate September 19 post uses the spaced form while discussing wins achieved with and without it. The dated extracts preserved in the archive index place the term in use by July 1996. Who first coined it remains unknown. (July thread, September post)
This materially predates the December 2003 Battle for Wesnoth example cited by Know Your Meme as its earliest discovered occurrence. That account leaves room for earlier usage. Nor does the earlier Angband evidence prove exclusive invention in that game: an archive’s oldest located occurrence is a surviving witness, not a birth certificate. (Know Your Meme’s account)
Roguelike attitudes were not unanimous. The 1995 Angband FAQ recommends eventually playing without backup resurrection, but also says, “As long as you enjoy the game, who cares?” It asks players to disclose backups or cheating modes when posting victories. That distinction makes the historical dispute more intelligible: private enjoyment and the meaning of a publicly claimed achievement are separate questions. The same community could defend personal choice while preserving standards for comparing successful runs. (Angband FAQ, sections 6a and 7c)
By March 1999, an Angband exchange describes not just reversing death but selling an item to discover what it is, then reloading with that knowledge. In July 2000, a NetHack player requests save-scumming utilities and recalls writing a DOS batch file for the purpose. These examples show the breadth of the practice: resurrection, undoing mistakes, learning from discarded outcomes, and automating backups. The modern debate inherits all of them. (1999 Angband exchange, 2000 NetHack utilities discussion)
When game designers gave players the choice
Other designers debated the underlying behavior without using this vocabulary. In his 1989 essay Why Adventure Games Suck, Ron Gilbert argues against arbitrary deaths and dead ends that force players to save defensively. His preferred solution is to remove the need for constant recovery. Daggerfall’s manual approaches the issue from another direction: repeatedly perfecting every decision can erase interesting failures from a character’s story. It encourages continuing after failed quests or getting caught stealing, while allowing restoration after death or catastrophe. These are different objections—one to frustrating design, the other to lost narrative possibilities. Neither amounts to a blanket rejection of restoring saves. (Gilbert’s essay, Daggerfall manual, p. 2)
Strategy games made another distinction explicit: changing a decision and rerolling the same decision are different operations. Civilization III co-designer Soren Johnson recalls preserving the game’s random-number state so repeated combat would produce the same result. Some players appreciated the restraint. Others objected to losing their ability to reroll. The developers made the rule optional, a change documented in April 19, 2002 patch notes. Preserving the random sequence discourages repeated identical attempts while leaving players free to return to an earlier position and choose a different strategy. (Johnson’s firsthand account, contemporary patch notes)
XCOM illustrates how these preferences continued to coexist. In 2012, lead designer Jake Solomon explained how Ironman’s restriction on reloading made losses more intense and consequential. In October 2013, Enemy Within lead designer Anand Gupta described an option literally called “Save Scum,” which resets the random seed on loading and permits repeated attempts at a shot. The slang had become recognizable enough to label an official feature, even within a series that championed accepting losses. (Solomon interview, Gupta interview)
The strongest historical interpretation is that save scumming acquired a name within communities where a save was supposed to preserve one continuing attempt, while the underlying desire to revisit consequences was already familiar elsewhere. Its broader use brings several different bargains under one provocative label: protecting invested time, experimenting with a world, selecting favorable luck, and defeating intended permanence. Those bargains explain why the same action can feel like prudence in one game and a broken rule in another. The surviving accounts leave open who first tried it and who gave it a name.