SimCity 5 Catch-All - Offline mode coming in next update

I know you guys might be getting tired of me harping on about this, but they have blatantly lied or embellished about pretty much everything about this game.

I'm not quite sure what is so awful about that video. Sims will go to the nearest place to find what they're looking for until that place is full. If you set up a pathological street system like that, you're going to force dumb behavior.

People need to stop bitching about the way the game works and start playing along with it. Sims are not little AI creatures. They are very simple automatons which follow very simple rules.

BadKen wrote:

I'm not quite sure what is so awful about that video. Sims will go to the nearest place to find what they're looking for until that place is full. If you set up a pathological street system like that, you're going to force dumb behavior.

The problem is that such a system creates really strange behavior (as shown above). That should actually be pretty easy to fix, as it seems like each 'sim' chooses it's destination when it spawns. That's why they all go to the same place despite the place only having room for the first car (or however many). If the first car 'claims' its spot upon spawning, you could then direct the rest of the traffic to more useful destinations.

People need to stop bitching about the way the game works and start playing along with it. Sims are not little AI creatures. They are very simple automatons which follow very simple rules.

EA dug their own grave here.

SimCity Site wrote:

Incredibly Deep Simulation — Every Sim, every car, every building has a purpose in SimCity. New data layer technology provides easy-to-read information, giving you the power to make choices that will shape your city and change the lives of the Sims within it.

That's a clever choice of words. There's certainly nothing incorrect about that statement. The sims in their cars do have purposes - they're just very stupid. The city-level stuff seems to be fairly deep (ignoring the various bugs related to specialization and resource sharing), but the implication there is that the sims themselves are part of the deep simulation, and that's clearly not the case.

Not that it matters in the end. The game will sell millions of units. Some people will play the game forever, others will get tired of it within a few weeks. Basically like any other game.

Anyway - that's it for me. I've said my piece and have no interest in continuing to harp on the downfalls of a game that other people are enjoying.

Expecting sophisticated AI in a game simulating thousands of independent characters at once may be asking a bit much for the current generation of hardware.

IUMogg wrote:

Videos like this make me feel stupid. 400K citizens, no public transportation, self-sufficient (no traffic in or out)

More of an explanation

After watching these videos, I was able to make my first city that reached over 40,000 population, and I got a mayor's mansion. The city is ready for further expansion as described in the video, by adding mass transit to multiply the availability of workers and shoppers. I'm interested to see what happens next!

This is also the first time that I've tried building a city with only llama speed. The city was making money early enough that I didn't have to rein in my spending too much. Much fun was had.

ruhk wrote:

Expecting sophisticated AI in a game simulating thousands of independent characters at once may be asking a bit much for the current generation of hardware. :?

Not really. By the sounds of it it wouldn't be asking too much to make the city simulate itself so it works correctly.

I know there are bugs and it's frustrating as can be when things don't work right. But overall I'm having a great time with the game! When I can get connected and play time passes a LOT quicker than I thought it would

To be fair, EA never claimed that their AI was not based on extensive study of gypsy nomads.

So what if the Sims are stupid? They eventually get to where they want to go, it looks like general traffic patterns when you don't look too closely, and it's fun. I have no doubt the other SimCity games never had sophisticated AI moderating the behavior of the citizens. In fact, I'm almost certain that the citizens were just numbers and nothing else. The fact that they're attempting a simulation at this level is impressive, so quit bitching about how it could have been better and just enjoy the game as it is.

And if anyone has a problem with the way the game "should have been" then they probably shouldn't have bought it. No one here is from EA and going to give two craps about the way things might have been.

namikaze wrote:

I have no doubt the other SimCity games never had sophisticated AI moderating the behavior of the citizens.

It's not SimCity, but 1994's SimTower (which was their localization of OpenBook's The Tower) did model the visitors and residents in your tower to the point where you could follow people around the tower as they went about their business. It was incredibly abstract (Sims were represented as silhouettes) but that's certainly a level of sophistication above the simple "shortest path for everything" simulation that's going on.

Also, as I mentioned up thread, there was an element of permanence in SimCity 4. Sims had set residences, set jobs, etc. You did not have a simulation where Sims switched jobs and homes every 24 hours. It might not have been agent based, but I'll gladly trade a more accurate model that's based on dice rolls than the model that's currently employed in the current iteration.

The Tropico series modeled every citizen, right down to eating, sleeping, going to church. But then it rarely got over 300 citizens at once.

The problem with SimCity 5 is that they programmed these sims to exist in an information vacuum.

shoptroll wrote:

Also, as I mentioned up thread, there was an element of permanence in SimCity 4. Sims had set residences, set jobs, etc. You did not have a simulation where Sims switched jobs and homes every 24 hours. It might not have been agent based, but I'll gladly trade a more accurate model that's based on dice rolls than the model that's currently employed in the current iteration.

If Maxis can add this functionality into the game, it'll make the game a million times more realistic. And from what I was reading, it sounds like Maxis is looking at adding more depth and functionality to the game already, starting with traffic management. The problem is, we can't know for sure what's on their development schedule. But predicting what might be in the game in the future is far more productive than complaining about what could have or should have been.

namikaze wrote:

If Maxis can add this functionality into the game, it'll make the game a million times more realistic. And from what I was reading, it sounds like Maxis is looking at adding more depth and functionality to the game already, starting with traffic management. The problem is, we can't know for sure what's on their development schedule. But predicting what might be in the game in the future is far more productive than complaining about what could have or should have been.

Right, but it's downright frustrating to watch them throw the baby out with the bathwater and claim it's an improvement.

namikaze wrote:

And if anyone has a problem with the way the game "should have been" then they probably shouldn't have bought it. No one here is from EA and going to give two craps about the way things might have been.

That seems like a case of "be careful what you wish for" if ever I've seen one.

Honestly, this simulation mechanism is pretty awesome, and the best I can tell is it needs a bit more randomness.

Explanation: Right now, what happens is that for each "kind" of driver destination, a distance is computed for each graph node. When a car reaches an intersection, it chooses which way to go based on which is closest. This is a pretty basic standard pathfinding algorithm. The reason it falls apart in the game is that many agents will frequently leave the same location at the same time, and they'll all book it for the same destination—especially if there's some sort of choke point (which you'll very commonly create when trying to make a layout that separates certain things like industry and residential). Further, when things get congested this further bunches up agents that may have started further away from each other, making it likely that there are far more agents together than room in the nearest regions of the map to satisfy them.

The way I would solve this is *not* to make some sort of "reservation" system. Why? Because that would involve much much more state, and likely be much much slower. Let me illustrate: In order for an agent to find their way to a given point on the map, they have to have a distance map showing how far away it is to that point from each intersection (as described above), and then use that to navigate. Right now, there's one of these maps per *kind* of destination—let's say this is on the order of tens of maps. In order to allow reservations to work, you'd need a map for each building that an agent is driving to or might want to drive to. This would be on the order of hundreds (if not more).

But, there are two things that could make it work just a little bit better, while still piling on players with a soul-crushing amount of traffic. The first and most important one involves randomizing things a bit. It's possible this is already there, but needs to be more random.

Instead of having driver agents navigate by choosing the optimal path each time, you can give them a chance of choosing the best path and a chance of choosing a less good path that still leads towards an objective eventually. And: this should [em]include[/em] the chance to drive into a given target location. (i.e. driving past the first home on the block and on towards the next one, and so on.) If this is already in the game (and it very well could be—they've described the streetcar system, and it includes some random chances), then they need to tune things up to make the suboptimal choices happen [em]more[/em] often. Particularly in cases of high congestion.

Anyway, this kind of thing is very interesting, because it's the random thoughtlessness of the system that actually makes it work. If you try to do this sort of simulation more "correctly", it becomes unreasonable to simulate (consider Dwarf Fortress, and how it slows down even on very beefy hardware when you get a lot of dwarves. This is precisely because of all of the path-finding that's happening, and DF *is* taking some shortcuts with things already. Now consider scaling that up from ~100 agents in DF to ~1000+ in SC.)

If you go with full randomness, you get behavior that's not very satisfying as a model of people. Consider the electrical and water grids in SC: those are purely random. Some people have complained about how if you're providing [em]just[/em] enough power, you can get brownouts at places distance from the power source as the random bouncing of electrical packets don't make it all the way out there. I think that's a perfectly fine outcome: it's predictable, you can work around it in a few different ways, and it provides an interesting problem to try to be efficient. However, it doesn't behave like people, at all.

Now consider the sewer system—that's a lot more like how people are modeled in SC: everything flows to one of a few destinations (outflow pipes). The distinction is that traffic has congestion, whereas the sewer network does not. So when every piece of sewage in the city piles in one direction, you're happy, but when all the traffic does? Not so much. (Note: In a real sewage system, you'd probably have to think about whether it's a good idea to stream everything in the same direction!)

I think the heart of the problem right now is that there's too much of that directed movement in the traffic, and it becomes obvious when you have a lot of it. Some of that results in a nice behavior of [em]you get gridlock[/em]. If you make the system too "good" that won't happen. But, there's also very nonsensical behavior that becomes obvious in some particularly egregious examples. If you add a bit more randomness, I think it will make that problem less obvious, and make the traffic appear more natural. It should still be weighted to cause gridlock, but not with the near certainty that it does right now.

(Oh, another thing that might work is to periodically recalculate the distance map, and take traffic congestion into account. This is hard to balance, though: if you do it too well, then the sims will behave with perfect global knowledge of traffic, and that will be pretty unrealistic. You want them to do it imperfectly, so that every sim on the NE side of the city doesn't suddenly switch directions because there's gridlock in the SW. Again: Add some randomness, weight things to only take congestion into account a *little* bit, and you're in much better shape. But you won't know if a given tuning is appropriate until you play it—and to be honest, I would not be at all surprised if you need to adjust the random factors based on the overall city population. (i.e. there's not just one "chance to take a poor turn", but rather that chance needs to change based on things outside the simulation. The parameters that make a low-population city behave with verisimilitude don't necessarily make the high-population city feel right. (And I suspect that is what we're seeing now: at some population densities, the way the simulation works looks great. At others, not so much.)

namikaze wrote:

The problem is, we can't know for sure what's on their development schedule. But predicting what might be in the game in the future is far more productive than complaining about what could have or should have been.

The thing that concerns me here is that they set up what is really a service, but seem to have planned to support it like an MMO or a single-player game with lots of DLC (i.e. The Sims). If they have a live team to support the game as if its service, like a typical Facebook game we could be seeing the beginning of a revolution in how AAA titles are developed. Wouldn't you like to get continually quality updates to the game that fix problems add more stuff for free (and lots more DLC)? But so far I have seen no evidence that they planned this properly. Paradox knows how to do this, Mojang knows how to do this, Zynga knows how to do this. But I'm not sure that EA does.

How the sims chose their destination and path seems fine to me. My only problem is when 2 cars merge into a lane and get stuck on each other permanently blocking a lane or two. Only fix I have found is to reload the city.

The only other problem I run into often is the number of workers a factory takes vs the number of workers that live in a similar sized residential building, depending on the tech level is something like 10-15 max density residential for each max density factory.

I saw the SimCity part of this on RPS, and my first thought was to add my own flavour to it: link

And that property (big factory employs way more people than big apartment complex houses) makes the traffic a nightmare, because of the optimal routing described above.

I haven't figured out the number-of-jobs-per-building type stuff, though. I should probably click on things and see if numbers of workers and such are listed.

Apparently, you do not even need commercial or industrial at all even in a 1-city region!

I watch that video, and my primary thought is still "Man, that looks gorgeous."

My short review after having played the came for about a dozen hours now:
1. It is fun and addictive.
2. It has flaws.
3. Since I generally stink at the game, those flaws go mostly unnoticed or are lower priority than my trying to learn how to play another SimCity game.
4. It is fully stable now for a while - not an issue for me at all.

Hypatian wrote:

I haven't figured out the number-of-jobs-per-building type stuff, though. I should probably click on things and see if numbers of workers and such are listed.

In the population screen it lists total number of jobs at each income level. In the comments of the YouTube video I posted previously, (or maybe in the actual video) the guy says he figured these numbers out. I think he just looked at available jobs, demolished a factory, and saw how that number changed.

The people to factory populations are fine, its the workers to factory that are not! Most of the cities are stay at home internet shoppers who never go out for any reason. (10-15% population seems to work)

Hypatian wrote:

Deep thoughts on pathing and destination reservation

How many buildings are in a high population city? 1000? You can very easily maintain counts of available/reserved/filled for a collection of that size. I haven't tackled any pathfinding stuff in over 10 years, but thinking about the problem for all of 30 seconds I'd say that I wouldn't have to do specific point-to-point paths at reservation time. I could break the city down to connected sub grids, and simply compute paths between my various grids and save them. When laying new roads, I have to recompute my grids. Then I can identify which grid my origin point is in, which grid my destination point is in, look up the appropriate saved path, and only have to do minimal work for the first and last few turns. Not to say this is trivial to do well (particularly figuring out which parts have to be re-calculated when new intersections are created), but it certainly seems achievable.

Pathing is such a major component of this game, it seems like a worthwhile investment of dev time.

Also, I like your idea of introducing some noise and not-quite-optimal paths. I'd like to see it add some weight to congested segments..."hmm, this segment is moving slowly, I should trigger a re-path for some agents in segments which flow into this one." Exactly what most commuters do. If you didn't check the traffic report before you left, you start looking for alternate routes as soon as your regular path starts going slowly, and not before.

I still have 9 or so plots available in West Goodjerville. Shoot me an Origin invite (Turrash is my name there) and I can get you. So far it's going well and nobody has gone up in flames as far as I know

I'm not going to get a chance to play until at least tomorrow night, but if you could save one of those plots for Ruhkfuture that'd be great.

I pinged Cobble via Origin too, but am posting to try to figure out what is the current primary active goodjer region (or regions, if we still have more than one that are considered the current ones) at this point? Between people abandoning early regions, and us all split over a bunch of servers, I have no idea who to ask for an invite to the "right" region anymore.

Mr Crinkle wrote:

Also, I like your idea of introducing some noise and not-quite-optimal paths. I'd like to see it add some weight to congested segments..."hmm, this segment is moving slowly, I should trigger a re-path for some agents in segments which flow into this one." Exactly what most commuters do. If you didn't check the traffic report before you left, you start looking for alternate routes as soon as your regular path starts going slowly, and not before.

How cool would it be if traffic jams started causing some individual cars to exit down side streets and make u-turns to find a way around, like you see in every traffic jam ever.

Farscry wrote:

I pinged Cobble via Origin too, but am posting to try to figure out what is the current primary active goodjer region (or regions, if we still have more than one that are considered the current ones) at this point? Between people abandoning early regions, and us all split over a bunch of servers, I have no idea who to ask for an invite to the "right" region anymore. :(

Man, I gots no idea at this point...lol