Adventure is the fons et origo of all later dungeon-crawling computer games, the grandaddy of interactive fiction, and one of the hallowed artifacts of hacker folklore.

The very first version was released by Crowther in 1976, in FORTRAN on the PDP-10 at Bolt, Beranek, and Newman. (Crowther was at the time writing what we could now call firmware for the earliest ARPANET routers.) It was a maze game based on the Colossal Cave complex in Kentucky, including fewer of the D&D-like elements now associated with the game.

Adventure as we now know it, the ancestor of all later versions, was released on a PDP-10 at the Stanford AI Lab by Don Woods in 1977 (some sources, apparently erroneously, say 1976). That version is sometimes known as 350-point Adventure.

Between 1977 and 1995 Crowther and Woods themselves continued to work intermittently on the game. This main line of development culminated in the 1995 release of Adventure 2.5, also known as 430-point Adventure

The earliest port to C was by Jim Gillogly under an early Unix running at the Rand Corporation in 1977; this version was later, and still is, included in the BSD Games collection. I have it from Don Woods directly that "[Jim Gillogly] was one of the first to request and receive a copy of the source" but that Woods did not actually know of the BSD port until I briefed him on it in 2017. (This contradicts some implications in third-party histories.)

Many other people ported and extended the game in various directions. A notable version was the first game shipped for the IBM Personal Computer in 1981; neither Crowther nor Woods nor Gillogly were paid royalties.

The history of these non-mainline versions is complex and murky. Functional differences were generally marked by changes in the maximum score as people added puzzles and rooms; however, multiple ports of some versions existed - some in FORTRAN, some in C, some in other languages - so the maximum point score is not completely disambiguating.

Same articles at [DA] are a narrative of the history of the game. There is an in-depth study of its origins at [SN]. Many versions are collected at The Interactive Fiction Archive [IFA]; note however that IFA’s historical claims are thinly sourced and its dates for the earliest releases don’t match either comments in the code or the careful reconstruction in [SN].

An attempt to untangle and document a lot of the non-mainline history has been made by Arthur O’Dwyer at [QUUX]. For our purposes, it will suffice to explain the chain of provenance that led from the original Adventure to the Open Adventure distributed with this document.

The original 350-point ADVENT on the PDP-10 had been one of my formative experiences as a fledgling hacker in 1976-77. Forty years later, in February 2017, while doing some casual research into the history of text adventure games, I looked through some source code at [IFA] and was delighted to learn of Adventure 2.5, a version of the Crowther-Woods mainline later than I had ever played.

Adventure 2.5 had been shipped long enough ago that today’s conventions of open-source licensing were not yet fully established. The Makefile contained a rights reservation by Don Woods and that was it.

I wrote to Don asking permission to release 2.5 under 2-clause BSD; he replied on 15 May 2017 giving both permission and encouragement.

Here is what Don said about differences between the original Adventure and 2.5:

> The bulk of the points come from five new 16-point treasures.  (I say "bulk"
> because I think at least one of the scores included some padding and I may
> have tweaked those.)  Each of the new treasures requires solving a puzzle
> that's definitely at the tricky end of the scale for Adventure.  Much of the
> new stuff involves trying new directions and/or finding new uses for stuff
> that already existed; e.g. the forest outside is no longer a small number of
> locations with partially random movement, but is a full-fledged maze, one
> that I hope has a character different from either of the previous two.
>
> As the text itself says, V2.5 is essentially the same as V2, with a few more
> hints.  (I think I came up with a better one for the endgame, too.)  I don't
> seem to have a copy of the similar text from V2, so I don't know whether/how
> it described itself to new and seasoned players.
>
> The other big change, as I mentioned above, was I added a way of docking
> points at a certain number of turns.  This was my second attempt to do what
> the batteries had been for: require being efficient to achieve top score.
> Alas, the batteries led to players deliberately turning the lamp off/on
> whenever they weren't moving or were in a lit area, making the game take
> even longer!  I set the requirement at what felt like a hard but fair
> number of turns, then applied several sneaky tricks to shave off another
> twenty.
>
> I hacked up a wrapper around the game (still in Fortran, most likely, but
> I forget) that would try each initializing the RNG using each second of a
> given day, while feeding in a script that either worked or aborted early
> if anything went wrong (such as a dwarf blocking my path).  As I recall,
> it took less than a day's worth of RNG seeds to find one that worked.
>
> I verified my script could work given a favorable RNG, and stuck
> that number in the message.
>
> I like how that final puzzle, unlike the game itself, does not readily
> succumb even given access to the game source.  You really need to fit
> together not only the goals and the map and use of inventory space, but
> also details like just what _can_ you do in the dark...?

Earlier non-influences

There is record of one earlier dungeon-crawling game called "dnd", written in 1974-75 on the PLATO system at University of Illinois [DND]. This was in some ways similar to later roguelike games but not to Adventure. The designers of later roguelikes frequently site Adventure as an influence, but not dnd; like PLATO itself, dnd seems not to have become known outside of its own user community until rediscovered by computer historians many years after Adventure shipped.

There was also Hunt The Wumpus [WUMPUS], written by Gregory Yob in 1972. There is no evidence that Yob’s original (circulated in BASIC among microcomputer enthusiasts) was known to the ARPANET- and minicomputer-centered culture Crowther and Woods were part of until well after Adventure was written.

(I was a developer of the Nethack roguelike early in that game’s history, in the late 1980s; we knew nothing of PLATO dnd. We did know of Hunt The Wumpus then from its early Unix port, but it didn’t influence us either, nor in any apparent way the designers of other early roguelikes. After my time the wumpus was included as a monster in Nethack, but this was done in a spirit of conscious museumization after historians rediscovered Yob’s game.)

Neither of these games used an attempt at a natural-language parser even as primitive as Adventure’s.

Sources