How come there are no good detective games?

[vovan]

Originally posted by Urban Ranger
As I understand it, you (as the author) define a set of criteria to judge the generation algorithm, and the various "crime cases" compete against each other for survival based on these rules. The closer they are to them, the "fitter" they are judged to be. Then the surviving cases get tossed back into the generator, which looks at them, mutate a variable here or there, and create a new batch.

Right, that's how a Genetic Algorithm works, but my point was: how do you actually define the rules? Having thought about for a while, I cannot really come up with a good set of criteria to judge the fitness of a particular crime chromosome by. I mean, you need to come up with some way to automatically analyze the coherency of a given case, which seems to me a rather daunting task. Instead, I would approach the generation of the crime in such a manner that simply cannot produce an incoherent crime, instead of generating random sets of variables and then weeding out the unfit ones.

[Urban Ranger]

Originally posted by vovan
First of all, I think that for this to be successful, one needs a strong information-sharing system. Basically, my idea was that every NPC would have a memoery of their own, represented by a semantic tree. At the root of it would be the most generic, almost useless information, and as you go deeper, more and more details are contained, so in the end the leaves have the most specific info - the one you are after.

That's a natural for OO.

Originally posted by vovan
Now, the deeper down the tree you go, the more details you discover right? So, my idea was that each NPC's memory would slowly decay.

I am not sure if you should do that, because there are so many factors affecting a person's memory. For example, a person can remember a very dramatic event vividly years after it occurred. OTOH, if it is a very traumatic event, your mind could try to bury it.

Another thing is most games are not time driven, but event driven, so the player has time to sit down and think about things. Without regular passage of time, there's no way you could implement memory decay.

[Urban Ranger]

Originally posted by vovan
Right, that's how a Genetic Algorithm works, but my point was: how do you actually define the rules? Having thought about for a while, I cannot really come up with a good set of criteria to judge the fitness of a particular crime chromosome by. I mean, you need to come up with some way to automatically analyze the coherency of a given case, which seems to me a rather daunting task.

You could always judge the scenes manually. This is going to be much slower than an automatic filter, but probably will still be faster than trying to write a concrete set of rules.

[vovan]

Originally posted by Urban Ranger
I am not sure if you should do that, because there are so many factors affecting a person's memory. For example, a person can remember a very dramatic event vividly years after it occurred. OTOH, if it is a very traumatic event, your mind could try to bury it.

Ah, but remember that we are considering a crime investigation game. I kind of assumed (incorrectly maybe) that the whole action would take within say 6 months of the crime, not much more. Hence, when I came up with the rough sketch of the memory system I outlined above, I rather thought about the minor details vs big things memory kind of thing. If a game were to span a few years, then of course your point may be valid, but it still seems almost an overkill. Besides, we are only concerned with memory relating to the crime in question. Hence, unless we are talking about the victim NPC, this may not even apply.

Originally posted by Urban Ranger
Another thing is most games are not time driven, but event driven, so the player has time to sit down and think about things. Without regular passage of time, there's no way you could implement memory decay.

Sure I could. Every turn the player takes takes a certain amount of time. Go from the police station to the victim's house - 30 min; question the victim - 90 min; search the house - 45 min; etc. As time wears on, people start forgetting stuff gradually. Of course, the time scale would need to be much exagerrated, like in the Sims, say, or the characters would seem much too forgetful, but...

Alternatively, in the simplistic turn-based system I proposed above somewhere, where the player gets ten turns to solve the crime: each two turns, say, each NPC forgets a bit of information. That doesn't seem too realistic, but would do perfectly well for a first prototype kind of thing.

[vovan]

Originally posted by Urban Ranger
You could always judge the scenes manually. This is going to be much slower than an automatic filter, but probably will still be faster than trying to write a concrete set of rules.

Ah, but that defeats the whole purpose of automatic crime generation, now doesn't it?

[Urban Ranger]

Originally posted by vovan
Ah, but that defeats the whole purpose of automatic crime generation, now doesn't it?

Ah, but the generation is automatic, just that the tuning of the generator is not.

[vovan]

Originally posted by Urban Ranger
Ah, but the generation is automatic, just that the tuning of the generator is not.

You could put it that way. But I still think it defeats the purpose of the generator. I mean, take Civ3. Suppose whenever you start a new game, 90% of the time, the map is unplayable. (Say, your settler ends up on a desrt tile surrounded by nothing but ocean.) Is that a good generator? Nope. Though it's plenty random.

Similarly with story generation. I can write a program that generates random numbers, uses them as an index into a dictionary and spits out the words. Sure I'm going to get a cool new novel written like this eventually, but what are the chances, and wouldn't I be better off using my own brain to come up with something creative?

Same here. Random assignment of values is just not good enough. There has to either be some way of assigning them so that they make sense together righht away, or to sort them out so that they make sense in the end.

[okblacke]

Hannibal Lecter: No, he covets. That's his nature. And how do we begin to covet, Clarice? Do we seek out things to covet? Make an effort to answer.

Clarice Starling: No. We just--

Hannibal Lecter: No. Precisely. We begin by coveting what we see every day. Don't you feel eyes moving over your body, Clarice? I hardly see how you couldn't. And don't your eyes move over the things you want?

I have a friend who's building a game (an MMORPG, but really nothing like Everquest and that ilk), to which I'm making some modest contributions.

Since the backdrop of the game is supposed to be very lively, part of what needs to happen is for the NPCs to interact with each other in many degrees, from casual to intimate, and with various degrees of ethical-ness.

The approach we worked out was to give the characters jobs and other relationships with other characters, and to move them around, and in that way crimes--when they occur--will arise "organically", if you will.

[child of Thor]

Although this is actualy opposite to what your talking about, it incorporates many of the things mentioned - just from the criminals perspective.
Its also a very enjoyable game to boot mobygames info on 'The Clue'
If you google for it you will find much better info on it.

[MattH]

Dammit, I didn't get any e-mail updates on this thread. Rrrrr...
Anyways, I kind of slacked off last week. Soon I'll post my non-finished code so you can see what I'm really thinking.

[vovan]

Originally posted by MattH
Soon I'll post my non-finished code so you can see what I'm really thinking.

Cool

[Urban Ranger]

Originally posted by vovan
You could put it that way. But I still think it defeats the purpose of the generator.

What I was saying was the generator "learns" from your feedback and adjust the parameters accordingly. It still is a genetic algorithm, sort of, with a manual filter instead of an automatic one.

[vovan]

Originally posted by MattH
Anyways, I kind of slacked off last week. Soon I'll post my non-finished code so you can see what I'm really thinking.

MattH? How goes it?

[vovan]

MattH! I know you are there.

[bipolarbear]

POST!

[MattH]

Ok, ok I got the e-mail.
Dev status:
Coding completed: 10-15%
Though process: 40-50%
Knowldge: 70-80%
We got 17+ new computers last wednesday and this afternoon was spent learning server 2003. But I SWEAR REALLY that I will give you something BY SATURDAY.

[vovan]

Originally posted by MattH
Dev status:
Coding completed: 10-15%
Though process: 40-50%
Knowldge: 70-80%

Nice

We got 17+ new computers last wednesday and this afternoon was spent learning server 2003.

Fun.

But I SWEAR REALLY that I will give you something BY SATURDAY.

Hehe, I was just wondering how it was coming along, really. I know I personally can't keep myself concentrated on a coding project for long, unless:

a) I see results right away. (For which to happen, the thing must be trivially simple, and those are no fun any way.)

OR

b) Someone is there to prod me along, literally or virtually. (That is, the project is for work/school, or someone keeps posting the "are we there yet" posts after I've announced my project.)

Hence, thinking that you could be someone like me, I figure an occasional "Are we there yet?" Post couldn't hurt any.

Good to see you are still going.

[J Bytheway]

It's probably too late to mention this, but if you're using a purely text-based interface it might be worth using a language tailored for it, like inform or TADS. Both are OO (in fact, I learnt OO programming using them, which does give me a rather twisted view of it, but I get by ).

The advantage of this is that you get a very good parser for the types of commands people use in text games, and also it handles objects in a fashion suitable for representing objects in the real world, rather than objects as they are usually used in mainstream programming languages.

[okblacke]

Originally posted by J Bytheway
The advantage of this is that you get a very good parser for the types of commands people use in text games, and also it handles objects in a fashion suitable for representing objects in the real world, rather than objects as they are usually used in mainstream programming languages.

But the engines that drive adventure games would be largely different from the engine that would drive a mystery game. (And yes, I know TADS and Inform, too, and am aware that they are capable of a great deal.)

Adventure engines are geared toward telling a single story, in as descriptive and unique a manner as possible.

The trick to a game like this would be to combine as many random elements as possible, yet make them seem as though they'd been individually crafted.

Some elements of these compilers-environments would help, in particular text-formatting, cross-platform capability, and if you were going to use a parser, these two are probably the best around. But I suspect other aspects might hinder.

In particular, if I were doing it (and as I mentioned, I am doing it, but in a different context), I would want as much as possible to come from a database of options. In other words, I'd be less inclined to use a language's object model and more inclined to design one that was constructed by reading data at runtime.

But again, that's just my US$.02.

[J Bytheway]

Yep, those are significant problems.

How about a C++ program to auto-generate source code for an inform game?

[loinburger]

Back in the day I enjoyed playing The Scoop, which changed depending on what you did in the game (f'ristance, if you're not at the docks when the killer comes back from Le Havre, then you can still pin the murder to him by blackmailing the chick who's running the heroin ring instead of turning her in to the police). It was ridiculously hard, so there was plenty of replay value despite the crime never changing.

[okblacke]

Originally posted by J Bytheway
Yep, those are significant problems.

How about a C++ program to auto-generate source code for an inform game?

I once designed a mystery game (in TADS) where what you did influenced what had happened (what the motive was, what had led to the crime, etc.). It was kind of cool.

When I realized how much work it would be to finish it, I nearly barfed.

[J Bytheway]

Originally posted by okblacke
I once designed a mystery game (in TADS) where what you did influenced what had happened (what the motive was, what had led to the crime, etc.). It was kind of cool.

It influenced what had happened, not what subsequently happened? That's slightly bizarre, but it would certainly increase replayability.

When I realized how much work it would be to finish it, I nearly barfed.

Alas, such is so often the case.

[okblacke]

Originally posted by J Bytheway
It influenced what had happened, not what subsequently happened? That's slightly bizarre, but it would certainly increase replayability.

And IF games are not usually very replayable.

The point more was that the user would end up unconsciously shaping the story. The concept was sort of like the old movie "Laura": Our hard-boiled detective walks into a dead girl's apartment and in the process of investigating the crime falls in love with the victim.

Only, in this case, whether he picks up a letter, looks out the window, etc., he gets a different clue, which creates a different story.

[vovan]

How goes it then, MattH?

[duke o' york]

Anyone heard anything about the CSI game? I saw a review of it in the newspaper today.