I haven't quite finished watching (around ~12 minutes in), but I couldn't help but feel the Ace Attorney games definitely fit this 'detective game' mold. In a talk by the series' founder, Shu Takumi, he describes the overarching structure for a case as follows: 1. Overarching Mystery -> 2. Smaller Mystery -> 3. Smaller Mystery Solved -> 4. A New Development -> 5. A Big Turnabout -> 6. The Case is Closed It's worth noting that he explicitly states that steps 2,3,4 are meant to loop into each other, and that when looked at in closer detail, those steps can be described as: 2. Smaller Mystery: The speaker claims to be innocent 3. Smaller Mystery Solved: Presenting evidence and solving contradictions, i.e. the main gameplay 4. A New Development: New information such as a new witness testimonial When Jon presented his flowchart of Discovery > Deduction > Action > (& loop), I couldn't help but think the similarities here between these separate concepts of the 'detective' loop were very interesting. The order's a bit different, but Step 4, "A New Development" maps to Jon's "Discovery" step. And then Step 3, "Smaller Mystery Solved" maps to Jon's "Action" step, because it requires you to select a specific segment of the witness' testimonial, and then select a piece of evidence in your inventory that proves that they are lying (Takumi's model doesn't have "Deduction", but it would slot between these two steps). And what's even cooler imo is that the games often do get more specific about demanding you understand what's going on, just like Jon describes - early in a case, usually all you have to do is present evidence (ex: a security camera photo) & the game will accept that, but later on, you do have to get more specific (ex: select the shadow in the corner of said photo that _proves_ that the victim was not alone when they died). Furthermore, with a total of 11 games in the franchise since the original title's release on the GBA in 2001 (counting spinoffs & Japan-only games), I feel like the series actually serves as a great existing example of how mystery games are 'repeatable', like Jon says. I think it's really interesting to look at how later games in the series advance from 2D pieces of evidence like photos (that work well on the GBA hardware) to more complex things that the player has to interact with (ex: 3D models of evidence, video clips, the sensations of the deceased in their final moments -- yes, AA gets a little weird). Here's a link to Chin Xiang Chong's translation of said talk, I think it's a fascinating read: medium.com/@waxinlyrical/ace-attorney-writing-and-narrative-design-talk-from-gcc-18-8c8951757c10 (okay I know this comment became an essay but I think this stuff is all super interesting. Thanks for the talk, Jon!)
With the exception of the fact that it requires the solution to be spoiled, and therefore makes entering a guess a high stakes final action, this is essentially the solution that the Sherlock Homes consulting detective board games use. Fill out a questionnaire, Sherlock Holmes goes over the solution and how he figured it out, and then you get points for correct answers and lose points if it took you more moves than Sherlock. Finally, rate yourself against a par system to see how well you did.
13:44 outer wilds is more of a "a few big mysteries with epiphanies" game (it's got 2-4) than a "many small puzzles game", even if it does both very well. great talk, gonna check out overboard.
I have been following your work for long time now , I have learned many things from your wit and genius. You are appreciated. Cheers and Take Care Jon.
I am more of a Return Of The Obra Dinn kinda guy than those textually and narratively defined mysteries (I prefer books for those) but this was very interesting anyways! 10:41 Like here where he talked about the runner-up and later RoTOD itself. 18:40 I wish more games had a "toolbox" of basic concepts of how the world works (things like "animals are startled by noise" or "the Berlin wall fell in 1989" or even "English Mansions usually have X staff") that you can put into verbs as ingredients, and which "boxes" you can use vary depending on the case or facts at hand. Only a few games did it, and even less were detective games. I want that feeling that you get in Cultist Simulator when you answer "What May Be Lost?" to progress in the Mansus, or when you recall a passing quote of Christopher Illopolly three hours later. Only thing that seemed close to me was Obra Dinn's causes of death, and it's not the same. 20:56 Took me 7 seconds. It's the green and turquoise ones in the almost bottom right and middle bottom left. My eyes just so happened to be there. My unreliable mind has advantages XD 21:48 I actually didn't notice they were pointing the other way! I was too focused on colors. That's why I have bad scores in half my logic tests. I miss the obvious and arrive at the answer backwards 25:42 Such is the goal. That seems incredibly hard
I agree and I think the best example of someone already doing something similar is the French comic book artist Ced. His Sherlock Holmes game book comics remind me a lot of Ace Attorney but also relies on deductions where you have to use both your knowledge of the real world and observations of the game world. It managed to make me feel much more like a detective than any video game has managed (Infocom's Deadline excepted, perhaps...). Some of Ced's books has been translated to English by Van Ryder Games.
"You've never finished a sudoku puzzle and finally put in that last number and thought 'oh it was a 5 all along! I knew it! That changes everything!" You should check out Cracking The Cryptic. Some of their puzzles feel EXACTLY like this. You half-solve the entire board leaving 3 possibilities in every square then BAM you get one digit and suddenly the whole puzzle falls into place.
As another fan of CTC, I think you're making the same point Jon is. The 81st cell is never going to be interesting to fill in, but the puzzle doesn't feel complete until it is. The last "aha" came when the puzzle was half-filled, and filling in the rest was trivial and tedious. In a game, you want the sense of completeness and the big reveal to line up. "We're not supposed to end narrative experiences with a process of tidying up." I think the overall message of the analogy is that sudoku puzzles aren't a good foundation for narrative detective games.
I especially liked the point about having a broad path to win. A narrow and obscure path made titles like The Last Express extremely frustrating to play (even back when it was released). I feel though that Overboard’s approach would not scale that well outside of replay-to-win because a lot of the ways the player demonstrates understanding is by doing something at the right time. So there are more things that would get lost in that transition.
Tangled tower is a really nice example of "the solve". Problem is that's it's the game providing the solve at the end of the story (like in a book), and not the player (or the reader) coming up with it.
What is being described here is actually a design concept used in a wide variety of titles known as "Show, Teach, Test" and it's typically used for teaching a player about mechanics as opposed to using tutorials or help text. Investigation games shuffle this slightly as items and info may Teach you something new, Show you exactly what to do next or both. It then Tests that you understood by following those instructions or making use of some new feature it introduced before giving you yet more info or items. This is why it's unreasonable to say this is limited to Detective games as this gameplay loop can apply to even Metroidvanias whereby you're unable to progress without having the burden of proof that you understood where to go or how to use some new item or ability.
This talk is AMAZING, Jon legit just went over my entire thought process of the past two years and made it better lol. My main concern with any kind of AI interpreter or grading system is that it could very easily come across as unfair. I imagine it is very hard to avoid the trap of making the player feel like they are just trying to satisfy the demands of some arbitrary, mindless system (which, granted, happens in real life). I’ve considered any number of systems to avoid this, but even a LLM/chatGPT-based system seems like it would still have this issue. I guess Jon did kind of give an answer to this when he mentioned rigorous testing and tuning, combined with authored possibilities. The only thing I can think of is to at least make sure the AI explains its conclusions, so even if you don’t like its argument you can theoretically at least understand it and not be as frustrated. That still doesn’t fix the underlying problem though which is that AI simply isn’t very good at deductive reasoning yet. Interested to hear other people’s thoughts.
I think you're imagining a far more complicated AI than he's suggesting. He describes how Overboard's AI works in the lecture, it doesn't involve any machine learning at all. Its AI in that you don't have to hand craft every possible outcome it gives, but it still only gives a finite number of outcomes which are all reproducible. I don't really see how machine learning would help here.
Deadline ruled my childhood. Great game! Yet now, every time i play a recent Sherlock Holmes game i just go ‘ughh’ .. due to that fact 1 + fact 2 = fact 3 thing.
Sounds a bit like trying to make single player among us A good way to frame this would be knowing the position of the pieces after a checkmate and the position of the pieces 10 moves earlier, then finding a way to reach the checkmate position, but not necessarily with optimal moves, maybe someone made a mistake, and the game transposed into that final position.
I haven't quite finished watching (around ~12 minutes in), but I couldn't help but feel the Ace Attorney games definitely fit this 'detective game' mold. In a talk by the series' founder, Shu Takumi, he describes the overarching structure for a case as follows:
1. Overarching Mystery -> 2. Smaller Mystery -> 3. Smaller Mystery Solved -> 4. A New Development -> 5. A Big Turnabout -> 6. The Case is Closed
It's worth noting that he explicitly states that steps 2,3,4 are meant to loop into each other, and that when looked at in closer detail, those steps can be described as:
2. Smaller Mystery: The speaker claims to be innocent
3. Smaller Mystery Solved: Presenting evidence and solving contradictions, i.e. the main gameplay
4. A New Development: New information such as a new witness testimonial
When Jon presented his flowchart of Discovery > Deduction > Action > (& loop), I couldn't help but think the similarities here between these separate concepts of the 'detective' loop were very interesting. The order's a bit different, but Step 4, "A New Development" maps to Jon's "Discovery" step. And then Step 3, "Smaller Mystery Solved" maps to Jon's "Action" step, because it requires you to select a specific segment of the witness' testimonial, and then select a piece of evidence in your inventory that proves that they are lying (Takumi's model doesn't have "Deduction", but it would slot between these two steps). And what's even cooler imo is that the games often do get more specific about demanding you understand what's going on, just like Jon describes - early in a case, usually all you have to do is present evidence (ex: a security camera photo) & the game will accept that, but later on, you do have to get more specific (ex: select the shadow in the corner of said photo that _proves_ that the victim was not alone when they died).
Furthermore, with a total of 11 games in the franchise since the original title's release on the GBA in 2001 (counting spinoffs & Japan-only games), I feel like the series actually serves as a great existing example of how mystery games are 'repeatable', like Jon says. I think it's really interesting to look at how later games in the series advance from 2D pieces of evidence like photos (that work well on the GBA hardware) to more complex things that the player has to interact with (ex: 3D models of evidence, video clips, the sensations of the deceased in their final moments -- yes, AA gets a little weird).
Here's a link to Chin Xiang Chong's translation of said talk, I think it's a fascinating read: medium.com/@waxinlyrical/ace-attorney-writing-and-narrative-design-talk-from-gcc-18-8c8951757c10
(okay I know this comment became an essay but I think this stuff is all super interesting. Thanks for the talk, Jon!)
I really do love the Inkle presentations 💚
Seriously. This and Meg's talks are burned into my brain. They're sooo good!
With the exception of the fact that it requires the solution to be spoiled, and therefore makes entering a guess a high stakes final action, this is essentially the solution that the Sherlock Homes consulting detective board games use. Fill out a questionnaire, Sherlock Holmes goes over the solution and how he figured it out, and then you get points for correct answers and lose points if it took you more moves than Sherlock. Finally, rate yourself against a par system to see how well you did.
This is a talk made specifically for me! Thank you!
13:44 outer wilds is more of a "a few big mysteries with epiphanies" game (it's got 2-4) than a "many small puzzles game", even if it does both very well. great talk, gonna check out overboard.
I have been following your work for long time now , I have learned many things from your wit and genius. You are appreciated. Cheers and Take Care Jon.
Love this talk and honestly find it hilarious how much shade you throw at Arthur Conan Doyle
Stopped halfway through the video, bought the game, played it, and I like it!
YEEEEEEES
Honestly, i've been waiting for a inkle gdc talk since overboard was announced
Inkle is great and thanks for the video!
Great talk, keep it coming GDC!
I am more of a Return Of The Obra Dinn kinda guy than those textually and narratively defined mysteries (I prefer books for those) but this was very interesting anyways!
10:41 Like here where he talked about the runner-up and later RoTOD itself.
18:40 I wish more games had a "toolbox" of basic concepts of how the world works (things like "animals are startled by noise" or "the Berlin wall fell in 1989" or even "English Mansions usually have X staff") that you can put into verbs as ingredients, and which "boxes" you can use vary depending on the case or facts at hand. Only a few games did it, and even less were detective games. I want that feeling that you get in Cultist Simulator when you answer "What May Be Lost?" to progress in the Mansus, or when you recall a passing quote of Christopher Illopolly three hours later. Only thing that seemed close to me was Obra Dinn's causes of death, and it's not the same.
20:56 Took me 7 seconds. It's the green and turquoise ones in the almost bottom right and middle bottom left. My eyes just so happened to be there. My unreliable mind has advantages XD
21:48 I actually didn't notice they were pointing the other way! I was too focused on colors. That's why I have bad scores in half my logic tests. I miss the obvious and arrive at the answer backwards
25:42 Such is the goal. That seems incredibly hard
I feel like the Ace Attorney games could really be improved with this approach.
I agree and I think the best example of someone already doing something similar is the French comic book artist Ced. His Sherlock Holmes game book comics remind me a lot of Ace Attorney but also relies on deductions where you have to use both your knowledge of the real world and observations of the game world. It managed to make me feel much more like a detective than any video game has managed (Infocom's Deadline excepted, perhaps...). Some of Ced's books has been translated to English by Van Ryder Games.
This was a hard topic to tackle. Very happy with the content and the in house approach these guys had to the detective mechanics.
Fascinating talk! Loved it
OMG, great topic!) And I love his Poirot references, he is soooo underappreciated
I like this topic. It reminds me game 'the deed'.
"You've never finished a sudoku puzzle and finally put in that last number and thought 'oh it was a 5 all along! I knew it! That changes everything!"
You should check out Cracking The Cryptic. Some of their puzzles feel EXACTLY like this. You half-solve the entire board leaving 3 possibilities in every square then BAM you get one digit and suddenly the whole puzzle falls into place.
As another fan of CTC, I think you're making the same point Jon is. The 81st cell is never going to be interesting to fill in, but the puzzle doesn't feel complete until it is. The last "aha" came when the puzzle was half-filled, and filling in the rest was trivial and tedious. In a game, you want the sense of completeness and the big reveal to line up. "We're not supposed to end narrative experiences with a process of tidying up."
I think the overall message of the analogy is that sudoku puzzles aren't a good foundation for narrative detective games.
I especially liked the point about having a broad path to win. A narrow and obscure path made titles like The Last Express extremely frustrating to play (even back when it was released). I feel though that Overboard’s approach would not scale that well outside of replay-to-win because a lot of the ways the player demonstrates understanding is by doing something at the right time. So there are more things that would get lost in that transition.
Tangled tower is a really nice example of "the solve".
Problem is that's it's the game providing the solve at the end of the story (like in a book), and not the player (or the reader) coming up with it.
Awesome talk. I'll gladly play his games :-)
What is being described here is actually a design concept used in a wide variety of titles known as "Show, Teach, Test" and it's typically used for teaching a player about mechanics as opposed to using tutorials or help text.
Investigation games shuffle this slightly as items and info may Teach you something new, Show you exactly what to do next or both. It then Tests that you understood by following those instructions or making use of some new feature it introduced before giving you yet more info or items.
This is why it's unreasonable to say this is limited to Detective games as this gameplay loop can apply to even Metroidvanias whereby you're unable to progress without having the burden of proof that you understood where to go or how to use some new item or ability.
52:14 "they look mo-ore ro-obo-oti-icccc" 🤣
Very insightful presentation, just a shame about the two painful distractions.
Great talk, but I wish he talked a least a little bit about his 2009 game "Make it Good"
This talk is AMAZING, Jon legit just went over my entire thought process of the past two years and made it better lol. My main concern with any kind of AI interpreter or grading system is that it could very easily come across as unfair. I imagine it is very hard to avoid the trap of making the player feel like they are just trying to satisfy the demands of some arbitrary, mindless system (which, granted, happens in real life). I’ve considered any number of systems to avoid this, but even a LLM/chatGPT-based system seems like it would still have this issue. I guess Jon did kind of give an answer to this when he mentioned rigorous testing and tuning, combined with authored possibilities. The only thing I can think of is to at least make sure the AI explains its conclusions, so even if you don’t like its argument you can theoretically at least understand it and not be as frustrated. That still doesn’t fix the underlying problem though which is that AI simply isn’t very good at deductive reasoning yet. Interested to hear other people’s thoughts.
I think you're imagining a far more complicated AI than he's suggesting. He describes how Overboard's AI works in the lecture, it doesn't involve any machine learning at all. Its AI in that you don't have to hand craft every possible outcome it gives, but it still only gives a finite number of outcomes which are all reproducible.
I don't really see how machine learning would help here.
Deadline ruled my childhood. Great game! Yet now, every time i play a recent Sherlock Holmes game i just go ‘ughh’ .. due to that fact 1 + fact 2 = fact 3 thing.
Reminds me of Ace Attorney.
There is some clear evidence here, that the bandwidth of his internet connection isn't sufficiant for giving an online presentation.
I would agree that I don't think Disco Elysium is a detective game. It's a game about detectives
Sounds a bit like trying to make single player among us
A good way to frame this would be knowing the position of the pieces after a checkmate and the position of the pieces 10 moves earlier, then finding a way to reach the checkmate position, but not necessarily with optimal moves, maybe someone made a mistake, and the game transposed into that final position.
@9:08 - "dective games live or die on the basis of grrgghsld".....WTF DID YOU SAY THERE?! PLEASE!!
I think it's "on the basis of verbs, right"