Make the player show the work

Detective fiction sells the labour. A film like Mystic River, 2003, comes apart through interviews, a scene, and attention. The pleasure is watching someone earn a theory. A game that hands you a list of suspects and waits for the correct portrait has skipped the labour and kept the costume.

The playable version makes the theory expensive to state and possible to reach. You assemble it from facts that were in the world, and you are wrong in a way that teaches you which fact you misread. A multiple-choice accusation is not automatically lazy. It is lazy when one option is glowing. Typing a name, or building the case out of pins, forces you to have the theory before the interface will accept it. That is the whole genre. Everything else is a coat.

Detective fiction sells the work itself. Watch Mystic River and the killing of Katie Markum comes apart through interviews, leads, physical evidence, and patient attention to detail. The detectives read the crime scene, press the people who knew her, register the things that refuse to line up, and slowly the case takes a shape they can see.

Games in the genre promise the player that same experience: hold the clues and the witness statements in your head, rotate them, and wait for the moment everything clicks. What is odd is how rarely the promise lands, despite decades of attempts. Point-and-click adventures, investigation side quests in open worlds, whole games built around a single case — all of them reach for it, and in most of them the player never quite feels like the person who cracked it.

So the design question is narrow and practical: what does a game have to do to let a player be Sherlock? The most useful move is to break detective work into its parts and look at each one separately. Information has to be gathered. Lies have to be caught. Leads have to be chased. Evidence has to be linked, deductions have to be made, and in the end an accusation has to be proved.

Gathering information is where every case starts, and it is the part games already handle well. The detectives in Mystic River pull a bullet from the scene, find phone numbers in a backpack, take a call from the police, and talk to the victim's parents, her friends, and a witness who saw it happen. All of that maps onto game verbs without much strain.

LA Noire turns evidence into 3D objects you can lift, turn over, and study from every angle. Condemned gives you forensic gadgets that raise fingerprints and bloodstains out of a room. Dialogue trees show up almost everywhere in the genre, and they are a reasonable fit for working through witnesses and suspects. None of this is the hard problem.

The trouble starts once the player is holding the information. Collecting clues is not detective work by itself — a detective uses clues to catch a lie, chase a lead, spot a link, and reach a conclusion. Games stumble at that point for a specific reason: the moment you check whether the player has worked something out, you risk telling them what there was to work out.

The usual answer is a multiple-choice question. The Wolf Among Us lets Bigby examine a clue and then asks him what it means, with a short set of readings to pick from. Sherlock Holmes: Crimes and Punishments runs the same idea through its interviews — at chosen moments a button appears, you challenge the witness, and you select the line that catches them out.

What undermines it is the prompting. The available answers do most of the thinking; the player scans them for the one that sounds right rather than forming a thought and then finding a way to state it. Guessing is always on the table. If a wrong pick only produces a shrug and another attempt, there is barely a test left at all.

Underneath this sits the central difficulty of the whole genre. Asking the player whether they have solved something leaks the fact that there was something to solve, and often its shape as well. The instant a game says "combine these clues now," the player has stopped discovering anything. They are answering a comprehension question about a case someone else has already closed.

One way to strengthen the model is to widen it. Detective Grimoire has you assemble a thought out of separate parts — pictures, statements, slots that combine in a great many ways. The deduction is still prompted, but with that many permutations, guessing at random stops paying.

The Trace works along similar lines: you answer by dragging in clues and observations you have already found, frequently needing more than one to complete a single response. Ace Attorney has the sharpest version of catching a lie. A witness testifies, and you present the piece of evidence that contradicts one particular statement. With several statements to choose between and a fat inventory of evidence and profiles, the choice takes real thought, and a life bar that drains on a bad objection makes brute force costly.

Done well, this produces genuine detective moments. Life is Strange puts an evidence board in front of you and asks which clues establish where Nathan Prescott took someone after a party. The pool of possible clues is big enough that guessing is unattractive, so you go looking for actual links — a date, a time, a broken taillight that pins down a car.

The prompting has not gone away, though. The question itself announces that a puzzle exists here, and usually hints at the sort of answer that will satisfy it. The player ends up cast less as the detective than as the assistant being quizzed by somebody who already knows how it ends.

The better move is to hand the player a tool for calling out contradictions and connections whenever they see one, with no narrow question opening the door first. Contradiction is built on exactly that. You interview suspects, accumulate a long list of statements, and can knock any two statements together at any time to catch somebody out.

Compare that with Ace Attorney, where you always know the current testimony contains a flaw, because the scene will not advance until you find it. Contradiction lets you walk away from a witness mid-thought, question somebody else, come back later, and carry several conversations at once. Nobody has told you who is lying or which pair of lines hides the problem. You have to work the case.

Papers, Please is not a detective game, but its discrepancy system explains why this shape is so effective. You cannot turn someone away at the border without stating why. You raise the inspection tool and highlight two things that cannot both be true — an expired passport against today's date, say, or a detail on a document against the rule book.

In both games the player has stopped reacting and started asserting. They are informing the game that they have noticed something. Papers, Please is the more elegant of the two, because any two pieces of text on screen can be paired. The combination count is far past brute force, and anything visible can turn out to matter, not only the objects a designer decided to file into an inventory.

Links and lies are well served by systems like those, but deductions are a harder problem. Discworld Noir shows the promise alongside the catch. Its hero writes anything useful into a notebook automatically, and you form a deduction by combining two notes. One puzzle asks you to put a strange phrase daubed on a wall together with the fact that the victim was hung upside down.

As a puzzle, it holds up. The cost is that the game must first mint notebook entries for both the phrase and the inverted body — and that second note says too much. Once thoughts arrive pre-cut as puzzle pieces, the player is not always reasoning. Often they are just clicking together reasoning the game prepared in advance.

The Shivah finds a cleaner route. Reading a dead man's email, you keep seeing messages from an Ethan G. The open question is his surname. He could be the Ethan Goldwater named in another message, or the Ethan Goldberg tied to an accounting firm in a ledger. Neither possibility is handed over as an inventory item. You open a fictional search engine and type a name. Goldwater returns nothing. Goldberg does.

The strength of that moment is the round trip. The information leaves the fiction, sits in the player's head or on their notepad, and comes back as an action they chose to take. There is no menu of candidate deductions. There is a thought the player built, and a way to find out whether it was right.

Few games commit to that idea as hard as Her Story. It hands you a database of interview clips concerning the death of Simon Smith, and the only way to reach any of them is to type search terms into a computer. Each search returns the first five results in chronological order, and no more.

Tangible rewards are almost absent: no clue inventory, no formal list of observations, no notebook stocked with finished deductions. Yet deduction never stops. A name, a phrase, an object, an inconsistency in one clip becomes a term worth trying. If the term surfaces new footage, the case advances. When it lands, the payoff is among the strongest eureka moments the genre has produced.

Search also fixes the part of detective work games most often automate: following a lead. In Blackwell, a place name mentioned in conversation or glimpsed in text can be typed into an in-game internet search, which returns an address, which becomes a map location. The name is only worth anything once the player has recognized that it is worth something.

Sherlock Holmes: Consulting Detective applies the same principle to a newspaper and a city directory. You comb articles for names, businesses, and details that might bear on the case, then look those leads up among a very large set of entries. The names that matter sit inside a mass of red herrings, so reading closely and choosing what deserves your attention is the actual play.

With the facts assembled, the case has to end in an accusation, and the weakest version is walking up to somebody and naming them. Certain Assassin's Creed investigations run this way: name the wrong person, lose a few points, try again; name the right one and the case closes. Nothing there asks the player to demonstrate that they understand what took place.

A stronger system makes you show your work. Eagle Eye Mysteries, a startlingly good DOS mystery for children, has you collect a large pile of statements and then, when you are ready, submit the key ones alongside the suspect's profile. Because the pile is deep and several of the right statements are required, a lucky stab is not enough — you need to know how the crime fits together.

Return of the Obra Dinn gets there differently. A pocket watch drops you into the frozen instant of each death, and your job is to establish who every person was and what killed them. The Crew Muster Roll is where those deductions get recorded, name by name and fate by fate. The combination space is enormous, so guessing is hopeless. You have to already know before the book will accept it.

The Vanishing of Ethan Carter closes its scenes with a timeline, asking you to number ghostly tableaux in the order the murder actually unfolded. Even a multiple-choice questionnaire can carry real weight at the end of a case, as Consulting Detective shows — by then you are walking into court, and you need a settled theory of who did it and why.

Detective games are an unusual design problem because the fantasy is a mental one. We admire detectives for knowing which question to ask, which lead to run down, how two facts connect, and how to look at the same evidence from an angle nobody else tried. A game can put the player inside that process in a way no film or novel can.

Plenty of detective games then give the feeling away by leading the player by the hand. If the interface poses the decisive question, displays the plausible answers, and converts every observation into an obvious puzzle piece, what the player solves is the interface, not the mystery.

The systems that hold up test the player's intelligence without tipping their hand. They let you gather information and notice a contradiction on your own schedule. Chase a lead. Form a deduction somewhere the interface cannot see, search for a name you personally thought was interesting, and finally prove the accusation with evidence. Whether yours does that is not a thing you can settle on paper — you build it, play it, and watch where your own attention goes. A good detective game asks for more than the right answer. It leaves the player enough room to have the thought.

In the Flockbay app, place three facts in the world and require the player to state the conclusion without a highlighted option. Play it cold, as if you do not know the answer. If you cannot get there, the facts are not in the world. Move them. Do not add a glow.

A detective mystery template and the AI point-and-click maker are the coat. The facts are the job.