Most of you reading this have probably heard about the well-known essay from The Alexandrian called the “Three Clue Rule” that deals with the following problem in mystery and clue gathering scenarios:

The PCs will end up veering wildly off-course or failing to find a particular clue and the entire scenario will grind to a screeching halt or go careening off the nearest cliff. The players will become unsure of what they should be doing. The GM will feel as if they’ve done something wrong. And the whole evening will probably end in either boredom or frustration or both.

The Alexandrian’s and other classical approaches say to either key1 locations better, or to key them more:

For any conclusion you want the PCs to make, include at least three clues.

I felt this approach just masks the underlying problem and puts a ton of pressure on my prep work. Not to mention that I would have to re-key most of the published materials I want to run, where this problem is rather endemic.

The difficulties I had with my own and most of the published materials seemed to run deeper, and I found myself pondering about how we play games, handle information, and construct our imaginary worlds.

The problem, as I defined it eventually, arises from the fact that we are hard-coding (hard-keying) the information to our mental models of the shared fictional narrative. The method I adapted to tackle this I named soft-keying.

This statement needs a lot of unpacking, so please join me on this journey as I try to build my case, one clue at a time.

The Hard-keyed Mansion

Let’s create a simple hard-keyed location.

The crew of misfits is tasked with breaking into an old mansion and stealing the MacGuffin.

The Krisdithas Estate & Ballroom
The Krisdithas Estate & Ballroom by Dyson Logos

I take time to hard-key my information to each room, I place a dozen clues throughout, and hide the MacGuffin in a secret room, behind a bookshelf in room 21. To open the secret passage players need to figure out a sequence of the mansion owner’s favorite books arranged as notes in his late wife’s favorite song. I plant the book and music information throughout the mansion as clues.

Here is the common outcome of that kind of prep work:

Session night comes and long story short, instead of a quick and fun heist we all spend the night in frustration as the players are bumbling around not able to either find the secret passage that leads to the MacGuffin or figure out how to open it.

What happened there I ask myself? I meticulously placed all the clues, deliberately crafted all the descriptions, did everything short of saying the solution to my players. Of course, I can’t tell them the solution or present the facts as that would make the whole affair rather meaningless. The same goes for asking for a check and on a success revealing the secret combination - it is tempting, but not a rewarding way to play. One additional temptation is to railroad the players towards the secret, denying their agency to fail, or any meaningful choices they’ve made so far.

On the player side, they took in the information, they pieced together most of it, it’s just that no one connected the books with music all the way through to solve the bookshelf in room 21 (which they might not even know is there). They are now frustrated, the energy low after 4 hours of pixel hunting and trying to touch and turn and pull every suspicious object in the mansion except the correct one.

You’ve had it happen to you, haven’t you? It happened to me and it will happen again. Not always, but often enough that the group might start to dislike these kinds of scenarios, or the way I run games. The “Three Clue Rule” can help, but is not a silver bullet, and can make the problem even worse if there is too much information clouding players’ minds. You could make several secret passages to the MacGuffin, but this mansion already took so many hours to prep - imagine how much information you would need to add to allow for several contingencies and come up with even more puzzles for players to solve.

Let’s discuss a few things first before we move on to the soft-keying approach.

Parallel Worlds of the Theater of Mind

From an adventure writer to the GM over to each player’s mind, every instance of the shared narrative, whether a scene or a location, is a parallel world of its own. Every instance will be slightly different than the other. Some will differ in color, some in spatial representation, some will be vibrant, and some sparse with details.

Traditionally we want to minimize these discrepancies by drawing maps, using handouts, and crafting elaborate descriptions. But, I think these differing subtleties are a feature, not a bug. In the same way a novel can have wildly different visualizations and interpretations depending on the reader, it’s OK for our shared narrative to remain different for each person at the table. We can play with that, and I would argue it improves the experience rather than hinders it.

Brain Cloud by David Revoy CC BY 3.0
Brain Cloud by David Revoy CC BY 3.0

The Mind’s Eye

The imaginary theater is built by our information model in each of the player’s minds. They interact with it by visualizing information, creating a mental map or node map. Let me illustrate.

Imagine a room with a door, a desk, a painting, and a window.

Here is what you might have so far in your mind’s eye, displayed here as a mental map with nodes.

The Room

  • A door
    • wooden
    • brown
    • brass handle
  • A desk
    • massive
    • writing
    • drawers
    • chair
  • A painting
    • a portrait
    • hanging on a wall
  • A window
    • closed
    • there is a street outside
    • trees

Now let’s add the following information to your mind:

You are a thief.

Your new mental map might add various new nodes similar to these:

The Room

  • A door
    • locked/unlocked?
    • someone might come through
    • do I hear something?
    • wooden
    • brown
    • brass handle
  • A window
    • open
    • came through it to the room
    • escape
    • night outside
    • closed
    • there is a street outside
    • trees
  • A Painting
    • is there a safe behind it?
    • on which wall is it exactly?
    • is it tripped with an alarm?
    • a portrait
    • hanging on a wall
  • A desk
    • what is in the drawers?
    • are the drawers locked?
    • is there a secret compartment?
    • are there any letters on the desk?
    • massive
    • writing
    • drawers
    • chair

What happened here is I activated a trope in your head, and you started thinking like a thief. More importantly, you added new nodes to your mental map, and they probably got shuffled up and down in relevance. By playing a thief you will interact with those mental nodes and add information to the map like so:

Thief in a room > Painting > Portrait > Alarm trip? > No > Move the painting > Is there a safe> Yes! > Crack the safe > Grab the loot > Escape through the window.

Where Am I Going With This?

So we already established that with five people at the table we have five parallel worlds, one in each mind. This means we have five different mental maps with different nodes. They might be similar but are not the same.

Sintel character by David Revoy CC BY 3.0
Sintel character by David Revoy CC BY 3.0

So if I hard-code the safe in my own mental map’s Painting node, there is no guarantee it will appear in any of the players’ Painting nodes! I can add as many clues and connections as I like but there is no guarantee that players will connect these to the correct nodes because their mental maps are all different. The more complex a scene is the less likely the nodes will match.

And this, my friends, is the heart of my problem with hard-keying information. I have a hard-coded mental map and to succeed, the players need to recreate my mental model to exactly match their own - yet we already know that our mental maps cannot possibly be the same!

Even worse, when the frustration strikes, players will go on a form of pixel hunting trying to connect/disconnect random nodes to each other, attempting to guess which ones will produce results. And as you might imagine, the more nodes and information they have, the more difficult that prospect will be. The more pixels there are on the screen, the fewer chances there are to nab the right pixel.

Soft-keying Approach

To fix that, I must avoid hard-coding information in my mental model.

Imagine a room with a door, a desk, a painting, and a window. You are a thief.

In my simple room example, I might leave the Safe node hanging out, not attached to the Painting node (I might attach it to the overall Mansion node). By doing it this way I allow the Safe node to appear in any node in the player’s mind. If my thief decides to investigate the Desk node, I can attach the Safe to it and say: “Yes, there is a safe behind the desk”. The safe is now in the same node in all of the player’s minds and is hard-coded as soon as it enters our shared imaginary world. Success!

Here it looks like I’m just putting things where the players expect to find them, which is true for this example, but it’s just because I left the gaming part out for the clarity of the argument.

The Game is Afoot!

So what is exactly gameable here? First, let’s set some design goals!

  • I want players to create their mental map out of the information available so that once connected in a logically consistent way it leads to a successful state.

  • I want them to achieve that by making meaningful choices and maintaining the player agency so that a fail state is always a possible outcome.

  • I want them to achieve that by both using player skills and activating their character skills.

  • I don’t want players trying to recreate my mental map to achieve a successful state.

  • I don’t want to prep contingencies. The scenario should work with a minimum amount of information.

  • I don’t want “quantum ogre”2 nodes.

Information Hierarchy

Before we make this gameable, I need to talk about information hierarchy, as it lies at the heart of our game of questions, answers, and meaningful choices. The shared fictional narrative we play through consists entirely of information, so we need a good model that provides a framework for our players to play through. I spent quite some time researching this until I came across the DIY & Dragons blog and appropriated the model presented there whole cloth. What I love the most about this simple model is that it’s inherently logical and immediately gameable.

Here it is in brief, copied from the DIY & Dragons blog:

Information can be landmark or hidden, or it can be secret.

This division can apply to locations on overland maps, objects within dungeon rooms, and even details about locations and objects that the players encounter.

Landmark information is automatic and free. Players hear landmark information the first time without asking, and if they ask, they can be reminded of it as freely as they heard it at first. Learning landmark information doesn’t take up any fictional time and doesn’t pose any risks.

Hidden information isn’t automatic - players have to ask to learn it. And it often isn’t free - there is often some fictional cost that must be paid to learn hidden information. There are two costs to learning hidden information. The first cost, which is possible but not mandatory, is time. If your game keeps track of time, then it’s possible that learning hidden information will require allowing some to pass. The second cost is risk. What’s hidden might not be beneficial, or might include both benefits and harms. To learn it, a character must be close enough to touch the thing, must interact with it directly.

Secret information has no guarantees at all. It is the opposite of automatic, and it’s always expensive. It’s not just that players have to ask for secret information, as they do with hidden; there is also a chance the judge will continue to withhold the information, unlike any previous type. To learn secret information, players must roll the dice and win. That extra risk, not just of injury but of failure, is what makes secret information more costly than hidden.

A device that causes a bookshelf to rotate out of the way, revealing a doorway when a particular combination of books are tilted at specific angles, is an example of a secret. The bookshelf is a landmark. The existence of the device is hidden, but any character who inspects it closely will notice that the bookshelf is perfectly flush with the wall and that the floor is scratched and scuffed in a half-circle in front of it. The operation of the device, however, is a secret. It’s not enough to spend time trying to activate the device. There is a chance the characters will try but still fail.

Seriously you should read the whole piece, it’s great!

The Wind Rises by Studio Ghibli
The Wind Rises by Studio Ghibli

The Procedure

Armed with all that theory we can finally create our gameable procedure.

By gathering landmark and hidden information, players need to piece them together to form valid clues, and by combining at least two valid clues they will unlock a secret.

To successfully combine two clues both need to be logical on their own, and their combination needs to be logical as well. Player skill is involved, character skill can be invoked and there is a chance of failure, so we tick our design boxes.

Here is the breakdown:

  • Gather all the information you can. Some are landmarks, others are hidden.
  • Piece together information to form a clue. A clue must be an inherently logical conclusion to be valid. Players don’t know if a clue is valid or not.
  • Piece together at least one more clue.
  • Combine at least two valid clues logically by interacting with an environment to either unlock a chance to roll or automatically discover a secret.

The Soft-keyed Mansion

Back to our mansion example, first, we will unkey some of our secret and hidden information from room nodes and attach them to the mansion node. This means that secrets can appear anywhere within the mansion and take any form. In a larger sense, they don’t have to even be in the mansion, but anywhere really, we don’t care.

Secret Information

  • hidden space that contains the MacGuffin
  • hidden entry to MacGuffin space is opened by combining Lady’s favorite music with Lord’s favorite books.

I know that the MacGuffin is in some kind of space, but I don’t want to know precisely what kind as the nature of it will depend on where the entry is. I could imagine a cellar space, a vault, a hidden room, a safe, etc.

I abstracted the procedure for opening the hidden entry as there can be many ways to combine the two clues about music and books, and although I will think of at least one way it might work, I will leave it to the players’ creativity to combine them. Additionally, I might add info for more combinations like names and birthdays of their kids, their allegiance to organizations, favorite hobbies, and such. I will wait for players to come up with a creative solution for success.


I tilt the books in a sequence of notes from the Lady’s favorite song, starting from Lord’s favorite book as the first note.

This kind of super-specific idea would not require a roll. Similar but a tad vague idea would call for a dice roll:

I check the bookshelf to see if there are any patterns with Lord’s favorite books and Lady’s music

That could be an INT check with a bonus if the character has skills in music or literature.

Arrietty by Studio Ghibli
Arrietty by Studio Ghibli

Players might surprise me with a different tactic. Let’s say they have gathered various info and formed the clue that the Lord is a wine expert. They go down to the wine cellar and inspect the wine shelf. They get nothing. Now one player says: “Wait a minute, is there any specific order to these bottles, like in names or something?” At this point, if they have a clue of “Salinger”, I might say: “Give me an INT check, Sommelier and Literature skills apply” He passes the check, and I give him this: “The row with the finest wines is sorted like this: R, E, G, N, I, L, A, S” If they figure it out and sort the bottles to spell SALINGER, the passage will appear before their eyes.

Hidden Information

  • entry to MacGuffin space

From my information model, I know that finding the location of the entry requires at least some interaction from the players as it is hidden. Here I’m waiting on any sensible exploration actions from the players on potential likely places. They might find the bookshelf in a study, or a wine shelf in a wine cellar. Sensible action has to include player skill and/or character skill. Example: “I check the floor in front of the bookshelf for any circular marks” That’s a sensible action right there if you ask me, and I might rule that indeed there are some marks on the floor. Hooray!

Hidden Information (continued)

  • late Lady Enora’s favorite music piece is Beethoven’s Piano Sonata no. 29
  • Lord’s Highaven’s favorite book is The Catcher in the Rye by J.D. Salinger

Here I included specific bits of hidden information as I have a clear idea of how to combine the two. Both are fully formed clues that can be combined logically within the mansion. As I mentioned, I might add even more specific info that I don’t know how to combine and leave it to the players to solve. As we saw in the wine cellar example piecing at least two clues in any logical form that applies in a given scene or a situation will yield at least a roll if not an automatic success.

After the specific hidden clues, I add as much hidden info I can think of like:

Hidden Information (continued)

  • (locked) Salinger books (drawers, nightstand,…)
  • (hacking) computer history full of wine review sites
  • (locked) wine review magazines

Landmark Information

  • A piano with a notebook on it. There is a real size statue of a woman sitting at the piano
  • Desk (Lord’s Highaven room) with some books on it
  • Master bedroom nightstand with a book on it
  • Gramophone with a record in it
  • Three paintings side by side with children on it.

I put as much landmark info as I can think of, some specific and some abstract, some can be hard-coded to specific rooms, and others can be free-floating, waiting to be found in a convenient place.

Player perspective

This procedure works great from the players’ perspective, even if they know how you do it. They know that there is work in the form of risking their characters’ lives in gathering all the info they need, and they know they will need player skills to piece together valid clues. There will be a chance to activate character expertise, and above all, they know they can fail at each step along this path and accomplish nothing.

Characters might trip alarms, activate traps, get into fights, and die. Players might fail rolls, create dumb, non-valid clues, and generally suck badly as players and accomplish nothing.

In the end, their choices and reasoning will matter. And with that, I tick all my design boxes and rest my case.


Having an information hierarchy in place means that hacking is very easy to implement whether it’s done with technology or via magical means.

Hacking Landmark information might allow the character to get the information without the need to observe it directly.

Hacking Hidden information might allow character access to hidden information without the need to interact with the environment.

Hacking might allow characters to piece together valid clues or even access secret information.

Information rules would still apply, so the closer a hacker is to the source and the more they interact with the world, the better results might be.

Genghis Jones by David Revoy CC BY 3.0
Genghis Jones by David Revoy CC BY 3.0

Monster & Bounty Hunting

Having a good information structure lends itself to other, similar gameplay modes as well. If we imagine players taking the role of Witchers and going on a monster hunt, I can easily take monster information, its modus operandi, whereabouts, stats, strengths and weaknesses, and structure it as landmark, hidden and secret information. With that, most of my prep is done. As players uncover bits and pieces by finding tracks, talking to witnesses, and examining attack scenes, they slowly form clues and inch towards the monster.

The same procedure can be used for bounty hunting missions, or even missing person missions.

  1. The Process of marking a location with numbers or letters and providing additional information via corresponding text blocks. 

  2. Palette shifting, meaning the quasi-railroading practice of substituting one encounter or location for another, to make sure the PCs experience it. Denies players’ agency by making their choices meaningless