D&D General On simulating things: what, why, and how?

[Oofta]

As I have been saying, I'm not categorically deprecating ALL use of the term simulation. Yes, a D&D fiction can be thought of, in some particulars perhaps, as a 'proposed scenario', though I think we need to be careful to say that is only possible to the extent your setting corresponds with reality (which is very likely for basic stuff like gravity and food). If we get into more 'systemic' stuff like economies, demography, geography, biology, etc. I think it becomes a good bit less clear we have A) the information needed to simulate anything, and B) that the world is meaningfully similar to ours at that level.

I'm not saying everything in D&D is simulation and never have. On the other hand I probably believe that quite a bit more of it can fall under that label than you do. Obviously some things such as initiative are purely game rules because we can't handle the truth ... umm ... can't handle simultaneous actions easily.

I think its very possible for video games to be illustrations of modeling reality and playing them could be considered a simulation. I doubt that very many games are realistic enough to be meaningfully simulations, but my understanding is the military itself has developed such simulations, using largely 'game tech'.

You've never played a driving sim. Race car drivers use them to practice.

Oh, I don't agree with that! One of the primary reasons for simulation is in order to understand the system being simulated. You build a model, and to the degree that the simulation mirrors the system's behavior you can then hypothesize that the model is representative, internally, of processes present in the actual system. This is the ENTIRE THRUST practically of climate modeling for instance. Same with simulations of industrial chemical processes, for example. I mean, sure, sometimes you don't care, often you REALLY REALLY DO. If the object is playing an RPG, OK, you are probably not that concerned with improving your understanding of reality.

You don't believe simulations have "black boxes"? How do you think they work? Simulations involve large number of people all the time and include behavior randomized based on averages. It's pretty core to many simulations. If you're testing how your emergency services respond to a fire in a skyscraper, you don't care how the fire started. If it's possible for there to be structural collapse or explosions, you don't care what causes those events, you care about how well the emergency services handle those events.

The simulation is not concerned with what triggers an event, it's concerned with the flow of the system and how it responds to events. In D&D "the system" is the PCs, how they respond to events. Not sure how else to say it, it's pretty clear to me. The DM sets the stage, the world and inhabitants the PCs interact with, the simulation is the actions and results of the PCs interacting with that world not the world itself. The events only have to mimic what the world would look like.

But we certainly do have a wide variety of situations that characters in a typical D&D game find themselves in where the speed and power needed to stand up to a dragon in melee combat would be a HUGE benefit, yet the character is just depicted as a (albeit fairly exceptional) normal human. Being strong enough to push back on a dragon when it decides to just walk over top of you, at 10 tons, certainly would have serious implications!

But again, this is just an outlier effect of simplified rules that doesn't really affect 99% of combats. In addition, no fighter can shove a dragon without magic or some supernatural ability. You can't normally shove someone more than 1 size larger than your PC. On the other hand, a huge or larger dragon (being 2 sizes bigger than your PC) can walk right over the top of you.

I think in D&D terms this is not an unreasonable position when talking about something like modeling athletic ability or similar stuff. If you move to something like simulating the market dynamics of a fantasy city? I think it simply doesn't work treating it like a black box. Those are really complex non-linear systems that IME (having done some of these things) require deeper modeling and analysis where you are going to want to know quite a bit about processes within the model, and it will really need significant iteration. I mean, there are some basic economic models that are THEORETICALLY supposed to give you something like a price as a direct output with certain inputs of costs and demand and information, etc.

If you're simulating a market economy, you're on your own in house ruling territory so I'm not sure how that applies. The game is silent on the issue. Also, good luck, economists have been trying to model economies for a long, long time.

Right, so this is a perfectly good topic of conversation that I would think addresses both the sense of what the OP poses, and relates that to the question about what do we mean by simulation. Honestly, I'm not sure why people are so invested in that word! I want to depict certain things, and your argument, which I mostly agree with, is that we don't care a whole lot about the NATURE of what we are depicting, we just want some results that give us a feeling of verisimilitude, or produce a sense of suspension of disbelief. I don't think 'realistic' itself even really enters into it so much, but it is more "which things do you want to mimic in order to get that."

Yeah, I want my game to give results that feel like the PCs are in a high fantasy game. Reality sucks at times, D&D is an escape.

Like with combat, for me, I'd like it to match up with my sense of what the character is generally capable of. So a reasonably consistent view of their physical abilities across various things, whether fantastical or maybe the character is just pretty mundane. In the later case I'd expect that combating things like large monsters would NOT involve getting into melee with them, as multi-ton (or even half ton, check out tigers) magical monstrosities are going to just rip up normal humans! I mean, speed could replace strength there, but without some supernatural level of ability in one or the other even melee with a bugbear is going to be dicey, and forget something ogre sized!

People have been hunting tigers, bears, mammoths without the Renaissance level tech PCs have for a long, long time. Put on full plate and a bear will knock you down, but it will have a hard time chewing through high quality steel. Eventually it will, of course. I don't think animals are particularly well modeled, bears for example are immensely strong. But I also understand that there are only so many options given how the math is set up. If a bear had a 30 strength, no hunter would survive an encounter with one but obviously they do. But again, that's just an indication that it's not a particularly accurate simulation.

Don't get me wrong, D&D combat is a bit over the top/silly at times for the sake of grand adventure. Some of that's simplification - if you're hunting wild boar you want a specialty spear designed with a cross guard for example. But it's just not worth the effort to put in that kind of fidelity. Some of it's just fun, our PCs are super heroic and beyond the human normal.

 

[AbdulAlhazred]

I'm not saying everything in D&D is simulation and never have. On the other hand I probably believe that quite a bit more of it can fall under that label than you do. Obviously some things such as initiative are purely game rules because we can't handle the truth ... umm ... can't handle simultaneous actions easily.

You've never played a driving sim. Race car drivers use them to practice.


You don't believe simulations have "black boxes"? How do you think they work? Simulations involve large number of people all the time and include behavior randomized based on averages. It's pretty core to many simulations. If you're testing how your emergency services respond to a fire in a skyscraper, you don't care how the fire started. If it's possible for there to be structural collapse or explosions, you don't care what causes those events, you care about how well the emergency services handle those events.

The simulation is not concerned with what triggers an event, it's concerned with the flow of the system and how it responds to events. In D&D "the system" is the PCs, how they respond to events. Not sure how else to say it, it's pretty clear to me. The DM sets the stage, the world and inhabitants the PCs interact with, the simulation is the actions and results of the PCs interacting with that world not the world itself. The events only have to mimic what the world would look like.

but you ABSOLUTELY DO care about how fire propagates through a structure, and airflow and etc. etc. etc. and you MODEL these things, often using discrete models that incorporate simulation of physical process! These are CLASSIC realistic modeling methods, trust me. Yes, you may also use, often, statistical mechanical methods when you have large numbers of identical discrete elements. OTOH you may NOT. In galaxy formation simulations with 10's of billions of particles EACH ONE is individually modeled, every gravitational interaction is calculated out. There's no 'one way' that simulations work.

If you're simulating a market economy, you're on your own in house ruling territory so I'm not sure how that applies. The game is silent on the issue. Also, good luck, economists have been trying to model economies for a long, long time.

Heh, you need not tell me. I've worked with people trying to do it. I mean, you actually CAN do a lot of stuff, particularly with modern ML techniques. I mean, I could easily create a commodity model that will tell you the price of some commodity with 80%+ accuracy, but it could be wildly far off the other 20%, lol. Silly people thought they could get rich running that model. Tried to tell them better, but fools and money...

People have been hunting tigers, bears, mammoths without the Renaissance level tech PCs have for a long, long time. Put on full plate and a bear will knock you down, but it will have a hard time chewing through high quality steel. Eventually it will, of course. I don't think animals are particularly well modeled, bears for example are immensely strong. But I also understand that there are only so many options given how the math is set up. If a bear had a 30 strength, no hunter would survive an encounter with one but obviously they do. But again, that's just an indication that it's not a particularly accurate simulation.

I don't believe anyone ever hunted a bear this way. They built traps, deadfalls, pits, drove them with dogs, treed them, shot them with arrows, and maybe finally some crazy young stud closed in after the bear was on its last legs and plunged in the spear that finished it off. That is pretty clear from the terminology and stories of this kind of hunting.

Don't get me wrong, D&D combat is a bit over the top/silly at times for the sake of grand adventure. Some of that's simplification - if you're hunting wild boar you want a specialty spear designed with a cross guard for example. But it's just not worth the effort to put in that kind of fidelity. Some of it's just fun, our PCs are super heroic and beyond the human normal.

I entirely agree with you on the 'fun' part, and that's kind of my point. It doesn't happen so much anymore, but a lot of people burned a lot of brain cells imagining they were really simulating stuff in some profound way and believing that making it a 'better sim' was something worthwhile (maybe sometimes).

 

[Hussar]

/snip

But I've never heard anyone call CaGI it simulationist, so something must have gone wrong somewhere!

Just to clarify - I'm using the definition I posted in bold. If you're using the Forge term "simulationist" then, sure, totally not going to count Come and Get It.

But, for me, yeah, it qualifies fairly well. There's a pretty direct correlation between the stated actions of the player and the events in the game. Come and Get it is a bit wonky because it does rather skip over what the fighter actually does to attract the enemies. So, I'd say it's much more on the edge of a simulation, but, yeah, I'd say it's close enough for government work. I mean, heck, most of the powers in 4e actually do line up pretty well as simulations - Hammer and Anvil, for example, works pretty much like it says on the tin - I hit the target and an ally adjacent gets a free shot too. Seems pretty easy to narrate. A (I hit) ->B (ally gets a free shot due to the nature of my hit -> C enemy is attacked twice, if the first attack hits.

Because of the granularity of most of 4e's powers - each power is a distinct action with distinct effects - I'd argue that in combat at least, 4e leads to the closest example of D&D actually being a simulation of combat. (Note, simulation as defined as " a system which will, through the use of that system, directly inform the narration of that event."

 

[Hussar]

Yes, but I'd rather use coherent interpretation which doesn't change depending on how many people perceive the event.

Fair enough.

But, that's YOU. You've added that. There's nothing in the game that says we should even consider that. If I rolled a 6 or 16 or 106, the narration doesn't change. All that matters is that I succeeded. At no point do the mechanics actually influence the narrative. After all, you're third person could also roll a 5. Which means that it doesn't matter.

The whole point of a simulation in an RPG is that the mechanics MUST inform the narrative, which also means that they must invalidate some other narratives. You cannot have mutually exclusive narratives resulting from a simulation.

 

[Hussar]

Umm, I followed the link that @Oofta supplied above and this was the quote from the very first paragraph -

A simulation imitates the operation of real world processes or systems with the use of models. The model represents the key behaviours and characteristics of the selected process or system while the simulation represents how the model evolves under different conditions over time.

Additionally, there's this line, right below:

This visual simulation should include details of timings, rules, resources and constraints, to accurately reflect the real-world process.

Those bolded bits are pretty much the key points. If your simulation doesn't represent how the model evolves over time, then it's not a simulation. Which is why I push back so much on this. D&D mechanics don't tell us how very often. You start at the bottom of the hill, you finish at the top of the hill and the system does not tell you in any meaningful way how you got from bottom to top. All you know is you managed to beat an arbitrary number using bonuses to a d20 that are largely divorced from any sort of narrative of the character. Again, what does being proficient in Athletics actually mean? How do we narrate that?

 

[Hussar]

Y'know, funnily enough, I don't usually watch live plays, but, I've gotten into Viva La Dirt League's NPC D&D. Dunno why. it's just kinda amused me. But, in the latest episode, it highlights very nicely exactly what I'm talking about with skills in D&D. I made a 1 minute clip here, from episode 96, so, where the one player is trying to hide his true identity from a group of NPC's.

✂️ An example of the "Simulation" of Deception

59 seconds · Clipped by J. McDonald · Original video "Consequences of our actions - NPC D&D - Episode 96" by Viva La Dirt League D&D

youtube.com

Now, I know, I know, it's being played for laughs. Fair enough. Although, to be fair, I've certainly seen more than my share of this kind of moment at D&D tables over the years. But, the point right at the end there of, "How can he be good at deception" pretty much nails it on the head for me. Even though the DM has even taken into account the really, really bad attempt at deception, the character's bonuses are so high that he succeeds anyway and makes the whole scene a farce. Funny, at least to me. But, as far as a simulation of trying to deceive goes, well, it's pretty far removed.

I suppose you could argue that the DM shouldn't have asked for a roll, but, then again, that's pretty standard to ask for one at this point. Not unreasonable. But, really, there's a perfect example of how the simulation doesn't really work very well (like at all) because it doesn't actually match the narrative.

 

[Oofta]

but you ABSOLUTELY DO care about how fire propagates through a structure, and airflow and etc. etc. etc. and you MODEL these things, often using discrete models that incorporate simulation of physical process! These are CLASSIC realistic modeling methods, trust me. Yes, you may also use, often, statistical mechanical methods when you have large numbers of identical discrete elements. OTOH you may NOT. In galaxy formation simulations with 10's of billions of particles EACH ONE is individually modeled, every gravitational interaction is calculated out. There's no 'one way' that simulations work.

We're at a whole different scale. I'm thinking about dispatching firemen, police for traffic control, ambulances and how prepared hospitals are. You don't care about the fire itself other than to roughly model the consequences of the fire based on either random numbers or numbers the author puts in.

Heh, you need not tell me. I've worked with people trying to do it. I mean, you actually CAN do a lot of stuff, particularly with modern ML techniques. I mean, I could easily create a commodity model that will tell you the price of some commodity with 80%+ accuracy, but it could be wildly far off the other 20%, lol. Silly people thought they could get rich running that model. Tried to tell them better, but fools and money...


I don't believe anyone ever hunted a bear this way. They built traps, deadfalls, pits, drove them with dogs, treed them, shot them with arrows, and maybe finally some crazy young stud closed in after the bear was on its last legs and plunged in the spear that finished it off. That is pretty clear from the terminology and stories of this kind of hunting.

You might want to tell this guy that you can't hunt grizzlies with a spear.

I entirely agree with you on the 'fun' part, and that's kind of my point. It doesn't happen so much anymore, but a lot of people burned a lot of brain cells imagining they were really simulating stuff in some profound way and believing that making it a 'better sim' was something worthwhile (maybe sometimes).

Another example of a simulation would be Microsoft Flight Simulator (MFS). Whether you're flying a Cessna or a 747, the sim doesn't model the function of the engine be they gas engines driving a propeller or a jet engine, it just knows that given X throttle the engine provides Y thrust. If you fly into a thunderstorm, it just models typical storms with randomized wind gusts and conditions that fall in the range of the severity of the storm.

MFS also has things like a HALO drop ship, a shuttle craft that transports crew from a spaceship to earth. It's a totally fictional craft designed for a video game. The craft is simply not possible with any technology that we know of. But people still haven't dropped the "S" off of MFS.

Being a simulation doesn't stop being a simulation if it lacks granularity or details. I don't think it stops being a simulation if aspects are completely fictional.

 

[AbdulAlhazred]

We're at a whole different scale. I'm thinking about dispatching firemen, police for traffic control, ambulances and how prepared hospitals are. You don't care about the fire itself other than to roughly model the consequences of the fire based on either random numbers or numbers the author puts in.

Yes, you have to have SOME SORT of model of the fire. That is, suppose the firemen choose to do X, that will change how the fire spreads or doesn't spread (IE opening a door can RAPIDLY change what happens next! Trust me, or take a firefighting course). Now, obviously you may just assume the fire and the firemen do X, Y, Z and only build a simulation of ER operations and triage, OK. Nobody said that you have to simulate the whole world, but you will surely still have models of the things that are part of your simulated domain. If I give a guy blood, my blood inventory goes down, etc. There will be a stepwise through time progression where the state at each step will be output of the previous one, plus perhaps external inputs, modelling, and generation of the next state.

You might want to tell this guy that you can't hunt grizzlies with a spear.

But that is exactly what I mean. He doesn't, NOBODY DOES melee with a grizzly bear! It would be insane, you would die. No human can survive that (except by sheer luck) because a 1500+ lb grizzly is multiple times stronger and faster than us. Grizzly is about the size of an ogre in most systems. Think about it.

Another example of a simulation would be Microsoft Flight Simulator (MFS). Whether you're flying a Cessna or a 747, the sim doesn't model the function of the engine be they gas engines driving a propeller or a jet engine, it just knows that given X throttle the engine provides Y thrust. If you fly into a thunderstorm, it just models typical storms with randomized wind gusts and conditions that fall in the range of the severity of the storm.

Well, it does have a model of your 'engine', and it can even be told to simulate things like engine failure. Sure, those models are not literal physics engines that implement all the laws of physics to produce perfectly realistic engine behavior, nobody said THAT. However, if you input different commands to your 747's engines, they will react in a realistic fashion, which requires some fairly complex logic, it isn't just a single number or lookup table or something. Trust me on this, jet engines are complicated and the complicated aspects of them are highly relevant to flight. I worked on 747 avionics system design back in the day. I doubt MSFS handles ALL the variables, but it is known to be reasonably authentic and detailed. Obviously you can do something that 'feels like flying' which is MUCH simpler. I am not sure I would call it meaningfully a simulation though. Certainly there is some sort of dividing line.

MFS also has things like a HALO drop ship, a shuttle craft that transports crew from a spaceship to earth. It's a totally fictional craft designed for a video game. The craft is simply not possible with any technology that we know of. But people still haven't dropped the "S" off of MFS.

No, but I don't think that makes it a simulation! Again, there has to be a system you are building a model of, and without any actual system, no model is possible. I would agree that, if you could build a REALISTIC model, then you have a simulation. KSP is realistic enough that it can model some actual behaviors of real spacecraft in an emergent kind of way (IE you build something using modeled components and it exhibits behavior that is not coded into the thing from the start and which closely resembles things that real-world spacecraft do). THAT I would call a simulation, even if the specific rocket you build in KSP never existed. For example someone built a model of SpaceX Starship and illustrated how its aerodynamics implemented the 'belly flop' maneuver (although the KSP version looked a good bit different from the actual thing, so obviously the sim is not perfectly accurate).

Being a simulation doesn't stop being a simulation if it lacks granularity or details. I don't think it stops being a simulation if aspects are completely fictional.

I think it stops being a sim if it simulates something that is entirely fictional, but I don't think we really disagree there. I just think that it is not useful to use words like 'simulation' when talking about most RPG mechanics. They are vastly far from even the level of MSFS or KSP, which are themselves not good enough to be really useful beyond illustrating what something "might be like" to a degree.

 

[Crimson Longinus]

I think it stops being a sim if it simulates something that is entirely fictional, but I don't think we really disagree there. I just think that it is not useful to use words like 'simulation' when talking about most RPG mechanics.

And I don't think it is useful to have so strict definition of simulation that it disqualifies most RPG mechanics if the aim is to talk about RPG mechanics.

 

[Oofta]

Yes, you have to have SOME SORT of model of the fire. That is, suppose the firemen choose to do X, that will change how the fire spreads or doesn't spread (IE opening a door can RAPIDLY change what happens next! Trust me, or take a firefighting course). Now, obviously you may just assume the fire and the firemen do X, Y, Z and only build a simulation of ER operations and triage, OK. Nobody said that you have to simulate the whole world, but you will surely still have models of the things that are part of your simulated domain. If I give a guy blood, my blood inventory goes down, etc. There will be a stepwise through time progression where the state at each step will be output of the previous one, plus perhaps external inputs, modelling, and generation of the next state.

So we agree that you do not have to simulate everything, sometimes you can just have high level events that occur. Perhaps if you were doing a presentation you might give examples of how something happens for context and to get people to understand what could happen, but it's not relevant to the simulation.

But that is exactly what I mean. He doesn't, NOBODY DOES melee with a grizzly bear! It would be insane, you would die. No human can survive that (except by sheer luck) because a 1500+ lb grizzly is multiple times stronger and faster than us. Grizzly is about the size of an ogre in most systems. Think about it.

Okay. I thought about it. I seriously doubt "nobody" does this, on the other hand with plate armor people would be more likely to do so. It's also more about simulating things like the fight with the cave troll in LOTR.

Well, it does have a model of your 'engine', and it can even be told to simulate things like engine failure. Sure, those models are not literal physics engines that implement all the laws of physics to produce perfectly realistic engine behavior, nobody said THAT. However, if you input different commands to your 747's engines, they will react in a realistic fashion, which requires some fairly complex logic, it isn't just a single number or lookup table or something. Trust me on this, jet engines are complicated and the complicated aspects of them are highly relevant to flight. I worked on 747 avionics system design back in the day. I doubt MSFS handles ALL the variables, but it is known to be reasonably authentic and detailed. Obviously you can do something that 'feels like flying' which is MUCH simpler. I am not sure I would call it meaningfully a simulation though. Certainly there is some sort of dividing line.

No, but I don't think that makes it a simulation! Again, there has to be a system you are building a model of, and without any actual system, no model is possible. I would agree that, if you could build a REALISTIC model, then you have a simulation. KSP is realistic enough that it can model some actual behaviors of real spacecraft in an emergent kind of way (IE you build something using modeled components and it exhibits behavior that is not coded into the thing from the start and which closely resembles things that real-world spacecraft do). THAT I would call a simulation, even if the specific rocket you build in KSP never existed. For example someone built a model of SpaceX Starship and illustrated how its aerodynamics implemented the 'belly flop' maneuver (although the KSP version looked a good bit different from the actual thing, so obviously the sim is not perfectly accurate).

I think it stops being a sim if it simulates something that is entirely fictional, but I don't think we really disagree there.

We disagree. The HALO Pelican is totally fictional, anyone flying it is flying an imaginary vehicle that as far as we know is not possible. At least not until we invent magic.

I just think that it is not useful to use words like 'simulation' when talking about most RPG mechanics. They are vastly far from even the level of MSFS or KSP, which are themselves not good enough to be really useful beyond illustrating what something "might be like" to a degree.

I don't see the point of limiting ourselves to game theory definitions of simulation.