1
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
4
<title>DM4 §49: Structure</title>
5
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
6
<link rel="stylesheet" type="text/css" href="dm4.css">
10
<a href="index.html">home</a> /
11
<a href="contents.html">contents</a> /
12
<a href="ch8.html" title="Chapter VIII: The Craft of Adventure">chapter VIII</a> /
13
<a href="s48.html" title="§48: A triangle of identities">prev</a> /
14
<a href="s50.html" title="§50: The design of puzzles">next</a> /
15
<a href="dm4index.html">index</a>
18
<a id="p385" name="p385"></a>
19
<h2>§49 Structure</h2>
21
<p class="normal"><span class="atleft"><img src="dm4-378_1.jpg" alt=""></span>
22
Games from all of the major design houses of the 1980s share a common
23
structure, partly because they were planned as if they were Hollywood
24
films, which even today retain the shape of a three-act play from
25
nineteenth-century theatre. Designers would begin by making a formal
26
pitch, writing two or three-page synopses of the action, and serious
27
coding did not begin until such a synopsis had been talked through.
28
Structural breakdowns of what they produced are often revealing.</p>
30
<p class="dotbreak">� � � � �</p>
32
<p class="normal"><i>Size and density.</i>
33
There was a time when the sole measure of quality in advertisements
34
for cheaper adventure games was the number of rooms. Level 9's most
35
original work, ‘Snowball’, claimed to have over 7,000
36
locations, of which 6,800 made up an unusually wearisome maze. Even
37
a 200-room game meant only minimal room descriptions and simple puzzles,
38
scattered thinly over the map. Whereas ten of the Infocom games have
39
fewer than sixty rooms, with ‘Seastalker’ (30 rooms),
40
‘The Witness’ (30) and ‘The Hitchhiker's Guide
41
To The Galaxy’ (31) the most geographically concise. Today's
42
custom is that, barring a few junctions and corridors, there should
43
be something interactive in every room.</p>
45
<p class="indent">Today's design systems impose few restrictions on
46
game size or construction, but designers still have an unconscious idea
47
of a “budget” for a game design, if only to keep its
48
proportions in balance. A typical medium-to-large game contains
49
250 objects, counting items, rooms and other sundries (the player,
50
the compass directions) as objects. Many items are not portable but
51
are instead walls, tapestries, thrones, control panels, coal-grinding
52
machines and, as a rule of thumb, three items to one room is about right.
53
We might therefore expect 60 rooms, and the next step in budgeting would
54
be to share these out among game regions. The 180 or so items might divide
55
as 50 portable items and 130 furnishings. (‘Wishbringer’:
56
250 objects, 52 rooms, 34 takeable items.) As for the text, ‘Enchanter’
57
contains 20,100 words, ‘Trinity’ 32,200 and ‘Curses’
58
44,000, but the latter are large games and 25,000 words is more typical:
59
around a quarter of the length of a typical novel, for an average of
60
only 400 words per location. The size of the source code varies dramatically
61
with the design system, but for instance: ‘Spellbreaker’
62
17,800 lines, ‘Christminster’ 13,000, ‘Once and
63
Future’ (Gerry Kevin Wilson, 1998) 35,000.</p>
65
<a id="p386" name="p386"></a>
66
<p class="indent">Limitations can be a blessing in disguise, because
67
they force a designer to keep asking if this part or that part is
68
altogether necessary. Here is Brian Moriarty, whose research went
69
as far as looking up geological surveys:</p>
71
<blockquote>The first thing I did was sit down and make a map of the
72
Trinity site. It was changed about 50 times trying to simplify it
73
and get it down from over 100 rooms to the 40 or so rooms that now
74
comprise it. It was a lot more accurate and very detailed, but a
75
lot of that detail was totally useless.</blockquote>
77
<p class="normal">This reduction to 40 rooms would have been worthwhile
78
even if memory and disc space had been unlimited. Redundant locations
79
can be an indication of too much prose and too little interaction.
80
‘The Light: Shelby's Addendum’ (C. A. McCarthy, 1995)
81
contains much that is praiseworthy but its reviewers took a dim view
82
of its having over twenty locations in which nothing happens.</p>
84
<p class="aside"><span class="warning">▲</span>
85
It is not always realised how technical constraints have influenced the
86
literary style of the classic games, and so of what we consider to
87
be good today. Infocom's writers were mostly working under an absolute
88
ceiling of 255 objects (including rooms) and memory was more pressing
89
still. Lebling summarised the implementor's frustration <i>en passant</i>
90
in a 1987 memo: “I made a pitch for a more-unlimited game system…
91
Brian [Moriarty] is already running out of table space in his game, Amy
92
[Briggs]'s game is too big and not even in Beta yet, all our EZIP games
93
have had to be cut, and ‘Bureaucracy’ [Douglas Adams, 1987]
94
had to become EZIP instead of LZIP.” (We would now call both
95
EZIP and LZIP version 4 of the Z-machine, but LZIP games were small
96
enough to run on a Commodore 64 – the biggest source of sales
97
– and EZIPs weren't.) Lebling lamented in 1998 that “a lot
98
of lovely shivers” had to be cut out of his almost-finished
99
‘The Lurking Horror’ as a result: whereas in spite of
100
the amount of her game (‘Plundered Hearts’) that ended
101
on the cutting-room floor, Amy Briggs felt that “The constraints
102
of running on the Commodore 64 helped the games be richer, I believe,
103
than if we had been writing then for the Pentium Pro”
104
(<i>XYZZYnews</i> 12). Similarly, Scott Adams's games, running on
105
smaller computers with tape decks instead of disc drives, were obliged
106
to show extreme economy with objects and textual messages, but
107
coded rules and what we would now call daemons so efficiently that these
108
few objects ended up tightly interlinked, with side effects and multiple
111
<p class="dotbreak">� � � � �</p>
113
<p class="normal"><i>The prologue.</i> Most games divide into
114
three phases: prologue, middle game and end game. These phases are
115
often closed off from each other, so that once a phase has ended it
116
cannot be returned to, though the prologue sometimes offers premonitions
117
of the end, or conversely the end game echoes back to the prologue.
118
Like the children in C. S. Lewis's tales of Narnia, the player is
119
always going “further up, and further in”. Stu Galley, in
120
debugging the mystery ‘The Witness’, found himself obliged
121
to enforce the plotting of the passage between prologue and middle
124
<a id="p387" name="p387"></a>
125
<blockquote>[The play-testers] discovered significant “branches”
126
in the story that I had overlooked. For example, what if the player sneaks
127
into the house or doesn't go in at all until too late? The first possibility
128
raised too many complications, so we decided to lock all the outside doors.</blockquote>
130
<p class="normal">The duties of the prologue are to establish an atmosphere,
131
to foreshadow what is to come and give out a little background information,
132
while giving the player enough entertainment to want to continue. The
133
interactive aspect of this is that the player has to pick up the game's
134
special skills, using commands, tools or actions special to the setting.
135
(The prologue to ‘The Meteor, The Stone and a Long Glass of Sherbet’
136
(Angela M. Horns, 1997) uses an easy puzzle involving a telescope and
137
a guide book to provide practice with them.)</p>
139
<p class="indent">The task of passing into the middle game should be
140
reasonably straightforward, but at the same time involved enough that
141
the player has a feeling that time spent on the game is time rewarded.
142
The designer would be wise to imagine that the player of the prologue
143
is really only toying with the game at this stage, and isn't drawing
144
a map or being at all careful. If the prologue is too big, the player
145
will quickly get lost and give up. If it is too hard, then many players
146
simply won't reach the middle game.</p>
148
<p class="aside"><span class="warning">▲</span>
149
The passage from the prologue to the middle game is often also the passage
150
from the mundane to the fantastical, so that the prologue answers the
151
question “How did I get into all this?” The prologue of
152
‘Advent’ is an above-ground landscape, whose presence
153
lends a much greater sense of claustrophobia and depth to the underground
154
bulk of the game. On the other hand, a few games drop the player right
155
in at the deep end, as in ‘Plundered Hearts’, which opens
156
to a sea battle in full swing.</p>
158
<p class="aside"><span class="warning">▲</span>
159
Notable prologues include the streets and meadows outside the apparently
160
impenetrable ‘Christminster’ college (4 locations), the
161
undemolished planet Earth of ‘The Hitchhiker's Guide To The
162
Galaxy’ (6) or 221B Baker Street in ‘Sherlock’ (again,
163
6), but some have been as large as the guild house of ‘Sorcerer’
164
(Steve Meretzky, 1984), at 13 locations, or the seaport of ‘Crobe’
165
(Jonathan Partington, 1986), at 18. ‘Advent’ in its classical
166
form had an 8-location prologue, but some extensions (such as Level
167
9's) fleshed out the above-ground substantially, making the volcano
168
visible as a precursor to its underground discovery late in the
171
<p class="dotbreak">� � � � �</p>
173
<p class="normal"><i>The middle game.</i> The middle game is
174
the one which least needs detailed planning in advance, because it
175
is the one which comes nearest to being a miscellaneous collection
176
of puzzles. On the other hand, since it is also the largest part
177
it is the most in need of some rough subdivision into segments.
178
Working through these segments, one by one, provides a sequence of
179
problems and rewards for the player. A first-draft design of the middle game may just
180
<a id="p388" name="p388"></a>
181
consist of a rough sketch of these segments,
182
with some general ideas for objects, places and characters. Slotting
183
actual puzzles in can come later in a more improvisatory, freewheeling
186
<p class="indent">The obvious way to subdivide is to carve up the
187
map, perhaps with a pattern to the regions, perhaps even sharing regions
188
out to different authors (as in the AGT game, ‘Shades of Gray’,
189
1993). Regions correspond perhaps to time zones, to the four winds
190
or the twelve signs of the Zodiac, or else are delineated from each
191
other through simple geography: cave games are especially prone to
192
this, often having a node-like room with exits in all eight cardinal
193
directions. Thus ‘Zork II’ (Marc Blank and Dave Lebling,
194
1981) has seven areas arranged at compass points around the Carousel
195
Room, with the area northeast serving as prologue. Sometimes the
196
same locale occurs more than once, revisited with a different perspective.
197
The innovative ‘Spider and Web’ (Andrew Plotkin, 1998),
198
in which a player is being interrogated about what really happened
199
when a secret installation was broken into, features repeated but
200
varying hypothetical versions set in the same locations.</p>
202
<p class="indent">Other designers structure the game around performances:
203
in ‘Ballyhoo’ the player enacts the full repertoire of
204
circus skills, and in a region of ‘The Quest for the Sangraal’
205
all seven deadly sins must be committed. More often, though, dramatic
206
actions are intended to become the turning points in a story. Gareth
207
Rees (in <i>XYZZYnews</i> 6):</p>
209
<blockquote>In ‘Christminster’, I identified a set of key
210
scenes each of which was an event or experience that affected the
211
player character, and moved the story forwards towards the conclusion,
212
and yet could plausibly be implemented as a section of an adventure
215
<p class="normal">In chronological terms, the plot literally moved
216
forwards: the clock, in ‘Christminster’, chimes the half-hour
217
when a key event takes place, advancing towards dinner in hall, but
218
time hangs heavy in an endless Cambridge afternoon while the player
219
is stuck. Puzzles were slotted in later, often around the needs of
220
the plot. One of the game's most enjoyable sequences, with the player
221
exploring a pitch-dark secret tunnel in the company of Professor Wilderspin,
222
in fact exists only to oblige the player to spend time in the Professor's
223
company so that he can do a good deal of talking. Like most traditional
224
interactive fiction, ‘Christminster’ has a plot with little
225
overall variation except for the order in which the player does things.
226
But some radical designers see events not as milestones but forks in
227
the plot. Thomas M. Disch:</p>
229
<blockquote>… any computer-interactive text deconstructs itself
230
as you write because it's always stopping and starting and branching off
231
this way and that … With ‘Amnesia’, I found myself
232
working with a form that allowed me to display these erasures,
233
these unfollowed paths.</blockquote>
235
<a id="p389" name="p389"></a>
236
<p class="normal">There are games, though, in which an entirely improvised
237
middle game is compensated for by a tautly controlled prologue and
238
end game. Andrew Plotkin (in <i>XYZZYnews</i> 14):</p>
240
<blockquote>Since ‘So Far’ [1996] is pretty much pure
241
surrealism, I didn't have a plot in mind originally. I had a theme,
242
and was co-inventing puzzles and scenes and events all at the same
245
<p class="normal">This sounds potentially shambolic, but in an interview
246
(<i>XYZZYnews</i> 13) in the immediate aftermath of testing ‘So Far’,
247
Michael Kinyon found it enormously affecting. Steve Meretzky:</p>
249
<blockquote>Sometimes you have only a sketchy outline and are just beginning
250
to coalesce the geography. Sometimes the geography coalesces around the
251
puzzles. Sometimes it's both together.</blockquote>
253
<p class="normal">Geography coalescing around puzzles is evident in
254
Meretzky's work, in which events often spread across multiple locations,
255
as in the case of the ‘Sorcerer’ flume ride.</p>
257
<p class="aside"><span class="warning">▲</span>
258
The middle game is likely to have the largest area of playable map that
259
the player will face. In laying this out, it adds to the interest to
260
make connections in the half-cardinal compass directions – northeast,
261
northwest, southeast, southwest – and to steer away from a feeling
262
that the game has a square grid. (One of the few defects of the
263
‘Trinity’ middle game, though possibly that was the price
264
to be paid for one of its better puzzles.) Equally, a few, possibly long,
265
loops which can be walked around reduce retracing of steps and avoid
266
the appearance of a bus service map in which half a dozen lines have
267
only one exchange.</p>
269
<p class="aside"><span class="warning">▲</span>
270
The passage from middle game to end game often takes the form of a scavenger
271
hunt: throughout the middle game a number of well-hidden objects are
272
collected and only when they are combined can the end game be entered. See
273
‘Lords of Time’ (Sue Gazzard, 1983) or indeed almost any game
274
produced in the wake of ‘Advent’, as it was almost taken
275
for granted that any game must have “treasures”. Soon
276
enough it became a cliché, and one which games like ‘Leather
277
Goddesses of Phobos’ or ‘Hollywood Hijinx’
278
(“Hollywood” Dave Anderson and Liz Cyr-Jones, 1986) send
279
up, but it's still not a bad idea, because it enables many different
280
problems to be open at once. You can be stuck finding sprocket 2 and
281
go and work on finding sprocket 5 for a while instead.</p>
283
<p class="dotbreak">� � � � �</p>
285
<p class="normal"><i>End game.</i> End games serve two purposes.
286
First, they give the player a sense of being near to success (they used
287
to be called the “Master Game”), and can be used to culminate
288
the plot, to reveal the game's secrets. They also serve to stop the
289
final stage of the game from being too hard to play, narrowing it to
290
only a few accessible rooms or objects. In cave games like ‘Zork’,
291
the final puzzle would be made exceptionally difficult but today's designers
292
<a id="p390" name="p390"></a>
293
usually prefer to give the player the satisfaction of finishing,
294
and themselves the satisfaction of knowing that their story has been
297
<p class="indent">A mark of the last pieces of the puzzle falling into
298
place is that loose ends are tied neatly up and the characters sent
299
away with their fates worked out and futures settled. Looking back,
300
from the point of view of a winning player, can you understand what
301
has happened and why? Can you also see what is to happen to the protagonist
304
<p class="indent">The final message is another important one and, as
305
with the overture, the coda is all the better for being brief. To quote
306
examples would only spoil their games. A popular device is to make
307
the two scenes which open and close the story “book-ends”
308
for each other, symmetrical and matching.</p>
310
<p class="aside"><span class="warning">▲</span>
311
To speak of “the” final message or “the” last
312
step is a little presumptive. Multiple outcomes are not to every designer's
313
taste, but Daniel Ravipinto's ‘Tapestry’ (1996), with
314
its sense of tragedy, and its misleading portents and advice, drew
315
much of its strength from an open end. (Its plot owes much to the 1993
316
<i>Star Trek: The Next Generation</i> episode of the same name.)</p>
318
<p class="aside"><span class="warning">▲</span>
319
Gerry Kevin Wilson suggests that the end game should feature “your
320
Big Nasty™. The Big Nasty™ is the final challenge, be it monster,
321
man, maze, or whatever. This is where you want to ham up your writing
322
and get a sense of urgency going. There needs to be a time limit…”
323
In this view, which is not universally shared, the end game is like
324
a video game's “Boss Level”.</p>
326
<p class="aside"><span class="warning">▲</span>
327
Like prologues, end games vary in size: from a one or two location single
328
closing scene (‘The Lurking Horror’, ‘Sorcerer’)
329
to a new game region (the Dungeon Master's lair in ‘Zork III’
330
(Marc Blank and Dave Lebling, 1982), 7 locations, or Roman Britain in
331
‘Curses’, 16).</p>
333
<p class="dotbreak">� � � � �</p>
335
<p class="normal">Games in the style of ‘Advent’ are very
336
wide, with around thirty or so puzzles, all easily available and soluble
337
in any order. Others, such as the Melbourne House adaptation of Tolkien
338
(‘Lord of the Rings Game One’ (later a.k.a. ‘The
339
Fellowship of the Ring’), ‘Shadows of Mordor’, ‘The
340
Crack of Doom’, Philip Mitchell, 1985, 1987, 1989), are very
341
narrow: a long sequence of puzzles, each of which leads only to a
342
chance to solve the next. Wide games are dull, since no problem solved
343
will lead to any radical change. Narrow games are difficult to pitch:
344
if the one puzzle open at a time is easy then play is too rapid, but
345
if it is hard then the player will be abruptly slammed into a wall
348
<p class="indent">Towards the end of design it can be helpful to draw
349
out a lattice diagram of the puzzles. At the top is a node representing
350
the start of the game, and then lower nodes represent solved puzzles.
351
An arrow is drawn between two puzzles if one has to be solved before
352
the other can be. Here is the lattice diagram for
353
<a id="p391" name="p391"></a>
354
‘Ruins’, with subscripted numbers showing the points scored on reaching each
355
given position: each of the artifacts is worth 5 except the jade mask,
356
worth 10. [<a href="colophon.html#ld_lattice">description</a>]</p>
358
<p><img src="dm4-391_1.jpg" longdesc="colophon.html#ld_lattice"
359
title="Lattice diagram for ‘Ruins’"
360
alt="The diagram is simple, but balanced, showing which puzzles are
361
prerequisites for solving other puzzles, from the start to the end of
362
the game. It also shows that for most of the game, the player has at most
363
two puzzles to work on, and at three points in the game, only one."></p>
365
<p class="normal">This diagram is useful for three reasons. Firstly,
366
it checks that the game is soluble at all: for example, if the jade
367
mask had been kept in Xibalbá, there would be no solution.
368
Secondly, it shows how much of the game happens in different areas.
369
Most usefully of all, the diagram shows whether the game is wide or
370
narrow and which puzzles are likely to be bottlenecks, with large
371
parts of the game depending on their solution. This tall, spindly
372
diagram is indicative of a fairly linear plot, not surprisingly as
373
the game is so small. The problem of entering the Shrine is evidently
376
<p class="aside"><span class="warning">▲</span>
377
A long arrow on a lattice diagram is a caution that some action very
378
early in play is essential even though it has no effect until some
379
other action much later on. If the early action becomes impossible
380
later, for instance because it is in a prologue which cannot be returned
381
to, the player will legitimately feel aggrieved. In ‘Christminster’,
382
“getting invited to dinner” is theoretically an early puzzle
383
because access to the Master's lodgings occurs at the outset of the
384
middle game, but the puzzle never goes away and remains accessible
385
right up to dinner time.</p>
387
<p class="aside"><span class="warning"><b>•</b>
388
<b>REFERENCES</b></span><br>
389
C. E. Forman exhibits the lattice diagram for ‘Enchanter’,
390
which clearly shows its prologue, middle game, end game structure, in
391
<i>XYZZYnews</i> 4. Replying in issue 6, Gareth Rees argues that game
392
analysis is an aid to, rather than an integral part of, game design.</p>
396
<a href="index.html">home</a> /
397
<a href="contents.html">contents</a> /
398
<a href="ch8.html" title="Chapter VIII: The Craft of Adventure">chapter VIII</a> /
399
<a href="s48.html" title="§48: A triangle of identities">prev</a> /
400
<a href="s50.html" title="§50: The design of puzzles">next</a> /
401
<a href="dm4index.html">index</a>