Detective games


Here is an interesting video about detective games I came across last week. It is mostly about input methods and what they do or don’t give away about the correct answer.


Mark Brown’s vids are great.


Coincidentally to this getting a bump, I noticed PC Gamer did an article on detective games.


Cool list, though “detective game” is a pretty liberal definition there. I’m surprised that Thimbleweed Park didn’t make the list. For that matter, The Witness (given their loose definition).


Thanks. That video gets one key point absolutely right: the player needs the ability to show they’re understanding the case unprompted. This was easily my biggest problem with some of the games discussed earlier.

But search boxes alone aren’t sufficient; the part where Her Story fell down for me was the extreme ease of accidentally finding important clips. Full text search simply isn’t expressive enough to allow distinguishing between the player displaying their mastery of the case vs. just stumbling on a key piece of evidence. It works better when the pieces of information you’re searching for something much more specific than that.

Also, for showing that you know exactly what happened, the kind of timeline structure shown in the video (“put these 12 events in the right order”) is just rubbish. It suffers from exactly the same problem that the video author is complaining about with regards to deduction. The player is spoon-fed the solution, with no sense of accomplishment.

What I’ve been imagining as the perfect solution for video games is having the player reconstruct the murder using the normal modes of interaction (e.g. if the investigation is a point and click adventure, so is the reconstruction, using the same set of verbs and locations). At any point in time you can reset the state to before the murder, take over any character, and play the pre-murder part of the game from that suspects perspective exactly the same way you played the investigative part. If you reconstruct the murder well enough, you’ve clearly solved the case.


I keep thinking of ancient games for this thread. Two that came to mind to me this morning: the old Telarium illustrated text adventure Perry Mason: The Case of the Mandarin Murder. It’s primarily a court game of course but you’re trying to figure out what happened in the crime. I never did, though.

The other is Activision’s Murder on the Mississippi, set on a river boat. That one had a remarkable approach to conversations where your character would click on words in a conversation to record them in a journal for future use/evidence/to bring back up later. It was the sort of game that I think was a bit ahead of its time, and that I wish had inspired more follow-ups.


Well, if we’re going back to the stone age…


There’s a great Digital Antiquarian post on Zinderneuf:

I find it hard to believe that you could do even mediocre procedural mystery generation with today’s technology. Let alone doing it on 8-bit micros. So the graphical game of that era that I find most promising is Killed Until Dead.


What about Covert Action? It did a reasonably good job generating mysteries while still fitting on one floppy disk. I don’t see any reason why you couldn’t improve greatly on it now.


I LOVED Murder on the Zinderneuf.


The Trace is a mobile detective game that looked kind of interesting in some brief clips in the Mark Brown video from last month.

Turns out that brief video clips can be misleading… It’s actually really miserable, one of the least engaging video game experiences I can think of. Most of the gameplay is investigating small (3-4 rooms) environments. In practice this means one of two things:

  • Randomly tapping on anything that looks like it might be an object of interest that you zoom into, nothing, or a piece of scenery that gives a bit of flavor text. There’s a ~1s delay between tapping and the effect, so this is excruciating.
  • Trying different kinds of touch gestures to try to open up doors, drawers, cupboards, etc. The system is very finicky either about the exact gestures or the locations, so it can often take a few tries to actually get it to react even if you’re using the right gesture. So there’s some cupboard doors that probably took me ten attempts to open (without even knowing for sure it could actually be opened).

So what do you find when investigating:

  • Items that go into your inventory, used for very light and totally obvious adventure game puzzles. (There’s a lever that’s stuck and gives a “you need some lubricant” message; there’s a can of lubricant; I wonder what two items I could combine?).
  • Questions (look at an open window, you’ll get the question “how was the window opened?”)
  • Facts/observations (zooming in on the window frame gives you the fact “the window shows signs of forced entry”)

You’ll then use the observations to deduce answer the questions. In the above example you’d deduce that the answer to the question “how was the window opened” is “the window was forced open”. I’m not making this shit up; that’s a genuine and representative “puzzle” from this game. (Sorry for the spoilers…)

The strange part is that the plot would have some elements well suited for non-obvious deductions. All the evidence is there. But the master detective you’re playing just totally ignores those. “Huh, there’s some zip-ties in the bedroom with the dead body, probably not important. But let me figure out if this other guy might have a drinking problem.”

So the investigation happens, you’re given no chance to show that you’re understanding what’s going on, and then the game does its big reveal. This system is just no good. Either get rid of the hand-holding of answering canned questions, or actually use those canned questions for the important plot points rather than just to drive a procedural investigation.


The Lost Files of Sherlock Holmes is a 1992 point and click adventure game from EA.

The story works surprisingly well for an adventure game from that era. The mystery starts out small, but expands and changes in nature through most of the game. There’s a large cast of characters and locations, which open up very naturally through the game. And for at least some of the locations there’s reason to visit them several time through the game. So it feels surprisingly non-linear.

The actual detective-work is a very mixed bag. There’s a ton of pixel-hunting. Most of it’s not too bad, but there’s some that were as unfair as anything I’ve ever seen. Most of the conversations with people are purely about exhausting the full conversation tree, and maybe coming back to ask more question if you’ve found out something new that they might react to. There is no need for you as the player to have any understanding of the case, Holmes does it all for you.

Except… There’s a few places where it looks like they initially had a lot more ambition than that. For example in the initial scene you have a chat with Lestrade, who thinks that the murder you’ve been called to is the work of Jack the Ripper. You can try to convince him otherwise with a menu-driven conversation. And it’s a conversation tree with a bunch of total nonsense, non-sequiturs, untrue claims of facts, and then the actual things that you’ve found out from the crime scene that point out to the killer being someone else.

Or in a couple of other places you’re tracking down specific people based on vague clues you’ve found out about them. You describe a person to a shopkeeper, who then wants to know how tall the guy was, their build, the color of their hair, was there anything special about them, etc. And none of this is spelled out, you did actually need to look at the clues and think about the answer. I think you can’t even answer all of them with the information you have: and if you try to guess about the things you can’t possibly know, the conversation will be shunted to a dead-end branch. You need to actually admit when you don’t know something.

It’s this glimpse into what they were trying to do that kept me playing. Unfortunately being wrong in these discussions has no consequences, they’re pretty rare, and they’re always strictly about facts. They aren’t really about things you’ve cleverly deduced.

The puzzle design is odd. There’s way too many hidden compartments etc. around. Or places where you need to repeat the same action multiple times with no sign that you’re going to eventually get a different result. But then there’s a few places that are just hilarious misdirections.

Like someone tells you “I won’t believe that X is dead unless I see something official, like a death certificate”; you go to morgue, and indeed the medical examiner has just produced one! Too bad they absolutely refuse to give that to you; it’s absolutely against policy. It’s going to be yet another one of those puzzles where you need to get them out of the room so that you can just brazenly steal the stuff, right? Or one where you need to chase people around for a permission slip. But really… The original guy you needed to talk to wasn’t looking specifically for a death certificate, just something more than merely your word. How about you just buy a newspaper? :-D

Anyway, TLFoSH isn’t really worth playing. It had some interesting ideas, and a larger scope than any more modern detective game I’ve played. But there’s just too much '90s adventure game bullshit to deal with, too much watching Holmes be a genius, and not enough of actually having to think about the case yourself.


I’ve been playing Sherlock Holmes: The Devil’s Daughter and it’s been an ok experience. I’m a big fan of the stuff Frogware has done with these games, and though they’ve added some weird stuff like more quick-time events and annoying mini games, the experience of playing as Sherlock with your faithful companion Dr Watson is still pretty solid.

But last night I hit a bug that annoyed me. I was in the fourth chapter, so I imagine in the home stretch, and I encountered a big I couldn’t move past, I was supposed to go into the sewers to continue my investigation but the game essentially locked up if I did so. I googled for info and found a post to the Steam forum by the developer that said yeah, you’ve pretty much got to restart the case, sorry about that. So I lost about an hour which, yeah, could be worse but still pretty annoying. Not sure I can give a thumbs up for this game. But I’m hopeful I’ll complete it.


Detective / investigation boardgames are really flooding on the market these days. Chronicles of Crime is another one. The gimmick is that much of the interfacing with the game happens through a smartphone app. Want to interrogate a character? Scan their QR code, and you’ll get some text to read out loud. Want to ask them about a specific item or person you’ve found during the investigation? Scan the QR code of that item while still in the discussion UI for that character. Etc.

There’s also a secondary bit of functionality in the app where, when visiting a new location, one player looks at the scene on the phone app (potentially in VR) and describes items of interest, other people try to find matching items from a deck of cards.

The primary upside of this app should be that it’ll let the scenario designer provide different responses depending on the information you have. And a smaller benefit is that there will be intermediate points in the story where the players prove that they’ve figured something out (by asking an astute question), not just the traditional “answer these five questions at the end of the case to prove you solved it” model.

Also, new cases can be sold as DLC to the phone app, no need for full board game expansions. All the physical components are generic (e.g. having a deck of 50 items and 50 persons), the story is completely in the app. But I don’t know that I really want to call it a “benefit”.

Unfortunately the execution just isn’t there. The amount of text you get for a single action is a couple of sentences. This makes the experience of playing the game just kind of slow and fragmented: there’s no chance for whoever is reading to even get into the narration before it’s cut short and you have to scan another QR code. And since the snippets are so short, there’s very little room for subtle hints. It’s just the most heavy-handed tripe imaginable. This is probably because the authors had to provide tons of filler text as answers to non-essential conversation branches, and couldn’t polish the parts that actually matter for the story.

The one scenario we played was just boring. There was none of that “we have no idea of what’s going on here or where to go next, so let’s toss around some ideas” discussions that’ll usually happen during Consulting Detective. It was just following up on the obvious clues one by one.

I doubt I’ll ever play a second scenario of this. But at least it solved the problem of the game giving static responses regardless of the players’ theoretical / actual knowledge of the case. My group has thought it was a real issue with the whole Consulting Detective-like genre. Now we know that the cure is probably worse than the disease.


Mythos Tales used Requirement Cards to at least somewhat let them track receiving clues and such (as well as having encounters based on time of day). It’s still a little clunky and the writing wasn’t as good as Sherlock Holmes, Consulting Detective, but it shows promise.


Yeah, the problem I have with that approach is that it’s obvious to the players that they didn’t get the entire scene. (And it also becomes obvious exactly when they gain whatever it takes to unlock the parts of the scene they missed out on. “Visiting this location gave us card 3, wasn’t there some place where we needed card 3?”). Or at least that’s been our experience with Time Stories, which uses a similar system.

So rather than being a reward for the players figuring out some part of the mystery (which is the experience I want), it just adds some hidden dependencies to which order you need to do things in.


I guess Return of the Obra Dinn is a detective game? It’s certainly got investigation and deduction. It’s also doing both of those parts in ways that are really interesting, though maybe not applicable anywhere outside of this particular game.

First the investigation part. In Obra Dinn you’re navigating through a sequence of 3d scenes, each one frozen in place at the time of someone’s death. Your goal is to figure out who died and how. There is very rarely (never?) enough information in just that one scene though, so you’ll need to scour a bunch of related (or perhaps unrelated!) scenes for further information.

The important part is this: there are no game mechanics involved with the investigation. You are doing nothing except looking at the scenes and trying to read it. There are no pre-defined points of interest to act on. There is no indicator that you’ve found something, let along found everything. It really is just you as a player looking at stuff and deciding what’s important. Or once you’re late into the game, looking for some very particular bit of evidence on a specific character that you’ve decided could work as the cornerstone of your next deduction step.

It’s just ludicrous how much more satisfying this open ended investigation is than going through the motions of gathering predefined facts in the more typical game.

Now, there is a downside to having the investigation part be so free-form. How do you prove to the game that you’ve understood the case, when the game doesn’t know what facts you have. The multiple choice systems like in Frogwares games would immediately leak information. The Consulting Detective style “write your answer down, then read the real answer, then decide whether you got it right” system only provides feedback at the end of the game and doesn’t give the player the option to refine the answer. And free text parsers are too hard.

The solution in Obra Dinn is kind of brilliant, kind of broken. What they do is only allow entering one form of deduction: what’s this character’s name and who died. There are 60 characters whose names you know from the crew manifest. You need to map their faces to their names. And there are probably 30 possible causes of death, with some additionally requiring you to name the killer.

Given this structured form of data entry, Obra Dinn requires you to correctly answer the same question 60 time for different characters. This almost entirely removes the issue with the questions themselves leaking information. It’s a standardized question with a ton of fixed options.

Since getting 60 answers right would be unreasonable, any time you get three answers right they’ll lock in. This is the kind of broken part: It’s a bit too forgiving, since unfortunately getting feedback for the correct answers implicitly gives feedback on the incorrect ones too. Even if you’re not explicitly brute-forcing the game, any time you get a set of three locked in it’s safe to go and change the answers you have for all the other characters a bit on whatever dimension you’re uncertain about.

I’ve complained in a bunch of posts in this thread about games only letting me deduce stuff at certain points in the story. It’s really frustrating to figure out something ahead of schedule, but getting no credit from the game since the question won’t be asked until the answer is totally obvious. Obra Dinn kind of does this: the game will inform you about whether you have enough information to determine their identity and the current difficulty of making that determination. But that’s just a guideline, and you can just ignore the warnings and enter your guesses (sorry; your iron-clad deductions).

Anyway, this system probably can’t be adapted to any other game. There’s not a lot of scenarios where triaging through 60 corpses makes sense.

As for the non-detective parts of Obra Dinn? The visuals are really striking once you move around in a scene, in a way that screenshots didn’t prepare me for. There’s a couple of really annoying UI issues (e.g. the way your first viewing of a scene is timed). And while piecing together the story is great, the final bit was a bit of a letdown.

Overall definitely worth playing.

Return of the Obra Dinn deserves a thread, matey!

Fun game, pretty enjoyable stories.


LK I am buying that NOW. I bet that’s fun. Spooky sounds edition!


Sherlock Holmes Consulting Detective: Jack the Ripper & West End Adventures is a standalone SHCD expansion. It contains six cases based on the classic West End Adventures expansion, and a four case Jack the Ripper campaign.

I’ve only played the Jack campaign, and it’s not good.

The writing is just terrible. The authors weren’t out to write a good yarn, but to make the point that Victorian London was actually a pretty shitty place to live and not worth romanticizing. Visit a place that really should be very relevant to the investigation, and you’ll probably get half a page of turgid exposition on just how a Victorian laundry functioned and how horribly exploited the workers were, followed by a paragraph on the actual character you came to see. Toward the end of the campaign we started just summarizing the bullshit flavor text rather than reading it out loud.

A second problem with the writing is that the descriptions of the crime scenes are gruesome as hell. Lovingly written descriptions on exactly how the victim was cut up, which body parts were cut off, the exact state of various organs, etc. And unlike the blah-blah-blah on Victorian social issues, you actually do need to read through this to solve the cases. It’s of course realistic, but it’s also actively unpleasant.

So, very little of the charm of Consulting Detective is there. What about the actual gameplay?

On that side Jack has two large structural issues. One is that in the three first cases you don’t really have a mystery to solve. Like, any attempt to actually investigate the whole crime will fail and the game pretty much even says that up front. Instead you’re supposed to be investigating some fairly arbitrary aspect of the crime for each case. Since the real world is messy and ambiguous and your goals are so vague, it’s hard to do well on these. It feels like half the job is deciding arbitrarily which bits of evidence to ignore totally.

And then in the fourth case you get another arbitrary task. After coming back with the answer, you’re told that it was critical for figuring out the case, and are thrown back into the game to consider all five murders (one case has two murders) as a single unit. The intended solution depends on leaps of intuition that are totally absurd even by the standards of Holmes, basically pattern matching from a single example. I think it’s genuinely unsolvable the way they intended, though brute forcing would have been an option.

But the most inexcusable thing is that we actually would have solved the crime in an earlier case using another approach, and tried to do it, but the game just outright shut the door on following that thread.

I don’t think the designers were at fault here though, other than the choice of subject matter. This was supposed to reflect a complex and ultimately unsolved real-world crime. That’s not at all compatible with making a whodunit that’s solvable by normal people.

I honestly don’t think the Jack campaign is worth playing, no matter how much you like the genre. The only value is in seeing just how badly things can go wrong when following the normal Consulting Detective template. The other cases should be known quantities, except that apparently the rewrites broke one of the cases again.