Puzzle Design

Hey everyone!

I’m looking for some help coding/designing the final puzzle for my game. This is more a plea for brainstorming than a hard coding question.

In my hybrid IF/graphical game, you take the role of a reporter working your way through a murder. Basically, by the end of the game, the player will have accumulated heaps of information, some of it relevant, some of it not. If the player (not the PC) knows the significance of the info, they will be able to piece it together into a correct account of events, pinning down who did what, why, and how. I’m envisioning a puzzle in the form of the reporter PC’s notes. I can do this visually, so hopefully that allows for some options that might not be possible in straight IF.

Basically, there are a couple of potential suspects and a couple of things the victim has been involved in that could motivate a murder. It’s gonna be the peak of the game, so several elements and layers of complexity would be cool.

Some things that I was thinking were:

[spoiler]A simple series of more or less direct questions: Who was the murderer? Where did it happen? What was the weapon? This will accomplish what I want (testing the player’s, not the PC’s knowledge), but it feels really heavy-handed.

A wheel showing all the NPCs with the victim in the middle. The player would draw lines between the NPCs (both the suspects and the bystanders) to indicate their relationship: Alice hates Bob, Carol betrayed Alice. The problem is, I don’t know how to guide the player unless I give a list of relationships (hate, betrayal, jealousy) that would feel ridiculously artificial. I could set it up so the player actually types in the relationship, but this also feels a little odd.

Otherwise, I could do a timeline where the player puts events in the victim’s life in order. When I think about the motive and what led up to the murder, it makes most sense to think of it sequentially. But when things actually happened isn’t a puzzle, so the puzzle would be knowing what was significant. The problem here is how do I ask the player that? If I just put out a list, then the significance becomes obvious. If I ask the player to type it, the events are too complicated to spell out.

There’s already a notes system in place that I quite like. The PC jots down relatively innocuous statements automatically as he/she discovers them, but it is up to the player to connect two separate pieces.

For example, say there’s a bunch of notes, in two different sections (one about the body and one about Adam), it’d look something like this:


  1. Fact: Zachary was murdered on Tuesday between 0200 and 0600
  2. Fact: Body had rope marks on wrists
  3. Fact: Zachary was shot with a .44


  1. Adam: “I haven’t seen Zack since raquetball on Sunday”
  2. Fact: Adam was seen at the Slee-Z motel Tuesday @ 2300, out at 0800 the next day
  3. Fact: Adam worked at a bank
  4. Adam: “I was visiting my mother Tuesday night”

So if you select 1A and 2B a conclusion pops up saying “Adam couldn’t have been the murderer”. If you select 2B and 4B you’d get something like “Adam was lying about where he was the night of the murder”. If you select two unrelated ones (say 2B and 3B), no conclusion would appear.

I think maybe if I use this notes system it’d be fine to just put the important conclusions in a timeline: the player will have already made the inference.[/spoiler]

Anyway, what do you guys thing? Have you seen similar puzzles in other games? What works? What doesn’t?

All the best,


(Maybe this should be in General Game Design?)

I think the nature of the final puzzle should be informed by the manner and format in which the data has been accumulated over the course of the game.

To speak to your suggestion, since you’ve kept a virtual reporter’s notebook for the player over the course of the game (27 June, 2011. Alice barely polite, implicated Bob.) it would be perfectly natural to lay out those observations either as if they’d been cut-and-pasted into logical categories, or even (since you’re using a Txt/Gfx hybrid) laying it out on a murder board with annotated links between places/events/people/times.

As for the endgame, some of this depends on what you intend to be the fun part for your players. There are three steps involved in detective work: collecting, analyzing, and reaching a conclusion. Analysis involves arranging the data into a usable format, and will probably be most enjoyed by casual gamers (if GUI’d well) and graphic adventure gamers if you try to implement it in the game itself. While I could be wrong, I think most people derive the greatest satisfaction from the third step (not that the rest isn’t enjoyable as well).

Emphasizing Step 1: Collecting the Data
While I might get myself in trouble, I’d say there are very few interesting [to me!!] examples of games which focus on collecting, although hidden object games like Big Finish Games’ 3 Cards to Midnight/Dead Time epitomize this and sneak in a great deal of story.

Emphasizing Step 2: Analyzing the Data
A common puzzle in graphic adventure games requires the player to reconstitute a message written on shredded or otherwise disassembled pieces, and is a spatial relations puzzle emphasizing the second step. In the classic logic puzzle, the fundamental collection has been done (artificially haphazardly, of course) and the conclusion naturally presents itself once the data has been properly described. In IF, Jon Ingold’s Make it Good is a fairly recent example of a game which stresses placing disparate information into a rational pattern.

Emphasizing Step 3: Reaching a Conclusion
An Encyclopedia Brown or Five-Minute Mystery book does the first two steps for you, leaving the reader to determine the logical conclusion.

Emphasizing Step 1 with a Smattering of 3, While Doing 2 for You Completely
In Knights of the Old Republic, several puzzles revolve around studying a crime scene and interviewing multiple suspects. The questions you can ask and the statements you can make in deciding the case depend on the player’s thoroughness in investigation. At the end of the scene, you can choose from several different statements which may be correct according to the information you’ve acquired and a minor application of reasoning. This emphasizes the first and third steps and does the fundamental analysis behind the scenes.

All Three Steps
Emily Short’s Bronze, on the other hand, requires you to perform all three steps in a casual, friendly way. You need to visit numerous rooms in the castle, make particular observations, correlate Story This with Painting That and determine Infernal Something Else. Make it Good requires all three explicitly.

Depending on how IF-centric this work is, I’d try to minimize the amount of suggesting the game does at the end, leaving more of the solution to the player’s own reasoning rather than it feel like clicking 1A and 2B give a positive result, 1A and 3B give a meaningless result, etc. KOTOR’s investigator-lite approached worked well in a quest-based action RPG, but you might be able to request more of your players in this regard. It’s hard to say without knowing more about what type of hybrid you’re working on: Meanstreets? Gateway? Starship Titanic?

I think you should figure out how you want to tell your story, and what parts of the investigative process make sense to be implemented given the medium and your goals for the player’s involvement.

Moderators, I wonder if this post would get more attention if it were moved out of the Inform coding forum and into the design section? The issues it raises aren’t really platform-dependent.

Jason, maybe you could provide some more info on the nature of the knowledge you’d like the player to demonstrate her mastery of? Your first example implies that the questions are the classic, object-focused Clue questions–who, where, with what weapon–but the later examples seem much more focused on abstractions like motive and relationships. So I’m not really sure what the parameters are. But I’ll still try to comment on a few of your suggestions.

You mentioned bringing in the graphical notes–>inferences interface and using it in the final puzzle. I think that this might be awkward in a couple of different ways. First, the notes system strikes me as an optional aid to the player, one she can choose to use as she goes along or not. Unless it’s renamed and brought more toward center stage throughout the game, I think that players who haven’t made consistent use of the notes would find it annoying to suddenly have to do so to complete the final puzzle. Also, it seems as though quite a few of these could pile up, making it a lot to sort through if a player hasn’t been keeping up with it all along. I’m also not sure that the timeline model is direct enough–it seems pretty abstracted from the questions of whodunit/how/why.

This also seems too abstract to me, at least if it is meant to be the final puzzle. Most of the connections drawn would be incidental to the killing, which would seem to make for a less than climactic moment. The drawing of a single line that reads “killed” between the murderer and victim might be fun, but how would the player know which relationships beyond that matter enough to be added? Also, again, these relationships seem very abstract compared to the questions of who/what/where/why/how…

Would these questions be more interesting if elicited during a dramatically structured scene? For example: The PC is a reporter and is probably doing some things that the cops wouldn’t like as part of the investigation. Could some dramatic tension be got out of an interview with the detective in charge of the case that would allow the player to answer these questions, but also feel like she’s in some kind of danger herself?

Another idea might be to require the player to do things, rather than tell them, to demonstrate that they understand what happened. For example, the PC gets an anonymous tip that the murderer has another victim in his house right now, so we must race to the murderer’s house to prevent a second killing! If we get to the mayor’s house and there’s no murder in progress, well, that means we dead wrong about the mayor. This method might be harder to structure plausibly, as the player has to be made well aware when her actions are meant to be assertions of fact, and when they are just actions. But handled with finesse, it might make for a reveal more like those in Dashiell Hammett or Raymond Chandler than those in Agatha Christie.


A side note: I ran across your post on TIGsource the other day, and noticed that the Mac package you’re offering for testing includes Spatterlight. That’s probably not a good idea–your testers may very well think the game is buggy when they try to play it: Spatterlight is very slow, especially with graphics, and it requires double mouse clicks instead of single ones to register mouse input. Gargoyle would be a better choice for Mac users, as it also supports sound and is much faster and up to date (though there is a killer bug in the current version that will screw up hyperlinked text.)

Nitpick: You mean Big Finish Games, not Big Finish Productions. (My initial reaction: “Wait, Big Finish makes games as well as audio drama? I have to check this out!” Slightly disappointed now…)


As someone who’s still carrying a torch for Tex Murphy, I assure you your disappointment is no greater than my own.

Conversation with a character can be agood way to establish that the player understood all the details, because the NPC can ask questions that the player then has to answer with evidence. I recommend Chris Huang’s Act of Murder, which uses that method (combined with a notebook of observed facts) to pretty good effect.