• Welcome to SC4 Devotion Forum Archives.

Gridlockcity

Started by ldog, November 14, 2009, 10:05:23 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

ldog

Hello and welcome to Gridlockcity.
I expect this isn't going to be an MD in the traditional sense. Indeed I am not altogether sure of the appropriateness of this as an MD.
I will attempt to make it entertaining as well but the main intent is to share information and research data of interest to those who want to learn more about the simulator itself.
This is an attempt to chronicle my experimentation and development within the traffic sim.
Our goal? To learn the deepest darkest innermost secrets from the depths of the traffic simulator.
Abandon all hope of getting to work in 2 commute time units or less ye sims who enter here!

This is a really a companion to ongoing discussions from a thread I started here http://sc4devotion.com/forums/index.php?topic=9179.0 which actually began with discussions here http://sc4devotion.com/forums/index.php?topic=5382.200

I would especially like to thank (in no particular order) Jplumbey, Z, Tarkus, Ripplejet and Catty for the critical information they all provided me in getting started. Mott, Tropod, 7trumpets and others whose names escape me who are no longer even around but who were the pioneers in making the traffic simulator work better than what Maxis provided us with. iLive for making the reader, without which I wouldn't bother with the trouble of hexediting all these files. And last but not least everyone not named who has participated in the discussion.

So first some pictures of our not so fair city:
Regional Network Overview


City Overview


Zone map


These pictures are from the year of our lord, 62 WW(Will Wright :P)

Our population currently sits at 364,157 sims and there are 174,193 commercial and 30,481 industrial jobs in the city.
We are currently the only city in the region called simply "Testhaven".

This is a pretty ugly city. Those of you following along from the beginning will remember originally I wanted to do work off of a smaller scale version of the 3RR map and I did start off doing just that, but it proved too much distraction for the work at hand.

The goal here was to get a city of fairly high population (remember now I don't use the CAM or a ton of custom content currently) within a region fairly representative of how the type I like to play in normally (although I had to go with just making about half of the map water for simplicitys sake) and to do this very quickly (I normally like to develop somewhat slowly and organicly).

From the zone map you can see most of the map is high density, although we still have a few medium density residential blocks left. All industry is HT so our only source of air pollution is from traffic. Education and healthcare services are fairly good, although they are overcapacity. All other services are excellent. We have an income of 70k and a budget of about 80k so the city is not in good shape financially. A side effect of rapid development. I started the city with a network and services that were a bit more extensive than needed, although they have slowly become inadequate. I have been slowly closing the budget gap though in an attempt to make the city sustainable. In the meantime we continue to function thanks to the "click to reset" trainer park.

Not a lot of custom content right now. I wanted to keep the variables down for testing purposes. You will of course see Peg's seaports and marina (mostly de-transit-enabled to not skew results) but that is about it. A lot of vanilla in all it's ugly and repetitive glory.  :thumbsdown:

Anyway, now that we got introductions out of the way, I'm going to end this post and start writing the first test results post which will be the real "meat and potatoes" of this MD.
Fear not, there are pictures!  :thumbsup:

ldog

#1
Ok, since Steve had some pretty shocking results in his Chicago with the stock max commute time of 6, this morning I decided to slap vanilla with .003 PH into Gridlock City and see how it runs.
We ran for 10 years. The following pictures have side by side before and afters. I know that only the volume displays are not accurate, but I feel that it is useful to show not just your congestion map, but also WHAT the underlying networks in place are. For accurate numbers see the graph.

For those of you just joining us for the first time, prior I was running various iterations of traffic simulator L beta, which is just the name I gave my own personal testing version.
The current test of 10 years was run using the stock traffic simulator, with a pathfinding heuristic value of .003 . No other values were changed. The zone and dataview are from the NAM (I did modify them slightly but they really have no bearing on the test).

Noteable differences include my max commute time was 30, max mt commute 20 (stock is 6 and 4). Railbased MT (train, sub, el, mono) capacity was double (6000 vs 3000), oh yeah, street cap I raised to 400 from 100(I don't think there are any streets left in this city). I also had a different CvS curve, intersection/turn effect, commute preferences. All of this accounts for most of the differences seen.






We all know for a fact that the commute time graph is buggy. Interesting though that one of the things that has been generally held to...the x10 commute bug, Maxisfrank said was actually more like x20...which makes sense if you think about it, everyone was assuming commute time still held for the total trip somehow so that is how we get x10 as a rough figure. If you knew (like a Maxis dev) that it was only for the firsthalf of the trip then you would come up with x20. If we take 70 and divide it by 20 we get 3.5 which considering this is the average commute and not the max, it could be highly likely that the sims are commuting 3.5 time units on average (one way). Remember now I don't have any regional commuters so the SHAPE of the graph is accurate, regardless of if the time is meaningless or not. So even though I had an initial max commute of 30, my sims were still only commuting 3 tu on average, and I still got abandonment in spots.

Now I got the usual changes to the city that we all expect by now when you change traffic sims; some figures bounced then leveled out more or less. Some things changed but then I can account for them because of transit type preference differences and network capacity changes. My average commute time actually went up slightly. Consider however that I had a CvS curve that gave us speeds of 1.3 at no congestion, where stock only gives us 1.0 . Also since I cut the subway capacity in half and stock has a higher preference for MT we rode the subway heavily still, the congestion on it was higher. In fact I finally broke 500% station use in the middle of the CBD on bus and got 498% on the subway station there.

These next few shots I thought should be interesting. They are of the above-mentioned stations. The first set show the subways route query. The second the bus.





Pretty amazing huh? What's amazing you say? The fact that these sims are doing a 6 time unit trip on their way to work. 6 is all they get in stock. (I didn't take any pictures of it, but I also went and route queried many of those buildings directly as well) Also interesting, I think this proves Jason right (I was waffling about it the other day) about the mt commute time being how far they will walk, they seem to be indeed walking 4 time units. (I will do another test later with only 2 to confirm) Steve and I have since concluded that MT commute time is an unused value, you can set it to 0 or any other number and it has no effect

We've already proven (I think it is acceptable to everyone involved) that the commute time graph is only affected by the morning commute (see earlier posts).
but what I see here indicates to me that the max commute time only affects the morning commute. Indeed it would seem that although the trip home is pathed out (I did not include pictures of evening commutes but I use a lot of owr so there is no choice but to take an alternate route home for many; also see the commute time graph test earlier) but the time it takes is in fact completely ignored. Now I know we all agreed that the display was buggy, etc, etc. But it was always held to that the game just took the time it took to get to work and doubled it to make the trip length, not that the ride home was free.

sumwonyuno

Now, I'll be following this as closely as the other traffic simulator-related threads.   :thumbsup:

Since the game only considers morning commute, then multiplies the city tile's average by two, it makes sense that the commute time graph is "buggy".

Wait.  There's two commute times?  Maximum and Mass Transit?


The City & County of Honolulu, a Mayor Diary based on Honolulu, Hawai'i.

mark's memory address - I've created a blog!

catty

Quote from: sumwonyuno on November 14, 2009, 12:01:30 PM
Now, I'll be following this as closely as the other traffic simulator-related threads.   :thumbsup:

Ditto

Interesting results so far

:)
I meant," said Ipslore bitterly, "what is there in this world that truly makes living worthwhile?" DEATH thought about it. "CATS," he said eventually, "CATS ARE NICE.

z

#4
Quote from: ldog on November 14, 2009, 11:09:19 AM
Ok, since Steve had some pretty shocking results in his Chicago with the stock max commute time of 6...

This same experiment has since been repeated with a max commute time of 4.  Same results.  An addition to the original post has been made explaining the details.

Quotebut what I see here indicates to me that the max commute time only affects the morning commute. Indeed it would seem that although the trip home is pathed out (I did not include pictures of evening commutes but I use a lot of owr so there is no choice but to take an alternate route home for many; also see the commute time graph test earlier) but the time it takes is in fact completely ignored.

Yes; this has been documented in a number of places.  The simulator doesn't even bother to compute the travel time of the return trip (which saves a lot of time); it just makes sure that one exists.

QuoteNow I know we all agreed that the display was buggy, etc,

With the proper scaling factor in a city with no neighbor connections, it can be made completely accurate.  This can be very useful for testing.

ldog

#5
So our ill-advised foray into the MD world has ended quickly.
I sent a pm to my local friendly site admin asking for rules clarification, and he thought it would be better to give me a project section.
This is really what I would have preferred, so here we are in our new home.

I don't know about the future of this particular thread, but maybe it will stay around a bit in a quasi-MD style.
Much as this is a horrible city in so many ways, it has kinda grown on me.
I of course run other tests in other regions, but every time I think about nuking this one and just building a proper replacement something stops me.
Probably because I know I will start spending too much time trying to build a nice city and not enough time testing.
So our 364,000 unsuspecting sims will get to live a bit longer, and who knows, they may even get some neighbors.

Anyway, I replied to Sumwonyuno (I hope I spelled that right) in a pm, but I will answer here for everyone else's benefit.
What I am showing above is that the trip home is a "free ride". The simulator calculates a route home from work, which may be very different from the route you took to work, but the time does not get counted against you. Neither on the graph, nor under the hood. It does not double the morning commute time or any other calculation therof. It simply ignores it.

Why would Maxis do this you might ask? From a programmers standpoint it makes perfect sense. Remember limited computer resources. It really gains us nothing but it would tie up valuable CPU cycles that could be used for something else. For as good a programmatic explanation of the traffic simulator as we are going to get without having access to the source code, I refer the reader to http://sc4devotion.com/forums/index.php?topic=2763.msg177981#msg177981 . While Steve is speculating on quite a few things, it is all very sound logic and well-educated guesses.

So the engine assumes that since you made it to work, you have to be able it doesn't matter how long it takes you to make it home. We figure out the return path, properly with the pathfinder, and if we can't find a return path home we stop going to work, but they ignore time constraints we completely ignore the time it takes.

Now of course Steve comes in like a Kung-Fu master off the mountain and crushes me, because I was feeling pretty good about myself since I thought I had found something new (but let's not be harsh on Steve, he had a bad morning  ;) ). Of course as I told him in our discussions, somehow I managed to miss this little fact in all the discussions I had with him and Jason, and I'm sure others have as well. So I don't regret running these tests, and I really don't even feel like a complete ass for posting the results. There is a lot of information out there. Some of it is misleading. Sometimes it is good to confirm something for yourself. The more people who do the same experiment and get the same results, the more comfortable we can be in knowing the veracity of those results. So for those of you like me who were unaware of this little fact, there you go, you learned something new. My post stands, I just retract any claims to discovery.

sumwonyuno

Ahh, now I understand that clearly.  No doubling.


The City & County of Honolulu, a Mayor Diary based on Honolulu, Hawai'i.

mark's memory address - I've created a blog!

z

Quote from: ldog on November 15, 2009, 02:39:49 PM
My post stands, I just retract any claims to discovery.

You can take those claims back.  I read through your post a little too quickly, and misinterpreted what you were saying about the commute time.  I thought your main claim was about the evening's commute not being counted for what it was, and somehow missed the fact that you clearly stated that the max commute time was one-way.  That claim is definitely new, and goes against everything I've ever read about that particular property.  But the results you reported certainly seemed to support what you claimed.

Since this is, in your words, "amazing," I decided to do a really simple test myself, to verify that there couldn't be any other factors entering into this.  I built a small town, with residences on one end of a long road, and commercial and industrial businesses on the other.  Cars were the only method of transportation, and the max commute time was 6.  I moved the commercial and industrial down the road to see how far the Sims would travel.  Based on my previous understanding, I wouldn't have expected the Sims to travel more than three minutes to get to their jobs.  But as you found, they're willing to travel the whole six.  And beyond six, all residences get abandoned, as you would expect.

So this understanding of maximum commute time is completely new, at least as far as I know.  I'm sorry I missed the full import of your claims the first time around, and I'd like to offer you a belated congratulations for discovering this.   &apls

b22rian

Quote from: z on November 19, 2009, 11:35:59 PM
You can take those claims back.  I read through your post a little too quickly, and misinterpreted what you were saying about the commute time.  I thought your main claim was about the evening's commute not being counted for what it was, and somehow missed the fact that you clearly stated that the max commute time was one-way.  That claim is definitely new, and goes against everything I've ever read about that particular property.  But the results you reported certainly seemed to support what you claimed.

Since this is, in your words, "amazing," I decided to do a really simple test myself, to verify that there couldn't be any other factors entering into this.  I built a small town, with residences on one end of a long road, and commercial and industrial businesses on the other.  Cars were the only method of transportation, and the max commute time was 6.  I moved the commercial and industrial down the road to see how far the Sims would travel.  Based on my previous understanding, I wouldn't have expected the Sims to travel more than three minutes to get to their jobs.  But as you found, they're willing to travel the whole six.  And beyond six, all residences get abandoned, as you would expect.

So this understanding of maximum commute time is completely new, at least as far as I know.  I'm sorry I missed the full import of your claims the first time around, and I'd like to offer you a belated congratulations for discovering this.   &apls

So your assumption than before Lenny's test was maximum commute time was / 2 . to account for both the
morning and evening commutes, steve ?

Brian

z

Yes, that seemed to be the general understanding.

ldog

Steve, thanks for doublechecking.  :thumbsup:

ldog

#11
So yeah, my response yesterday was a bit on the lame side  &ops
I was feeling a bit under the weather yesterday morning.

I am of course very pleased to make a discovery instead of just reiterating more of the same.
I also want to stress again, what a group effort this all was, all the people named at the top contributed to that discovery.
Catty in fact was quite instrumental as she has been a very apt research assistant.
It was some of the very oldest experiments and posts that she dug up that led me to it.
I am also quite glad that Steve went ahead and verified this in a much more controlled setting so we can be sure.

Also this doesn't mean that everyone should take their existing traffic sim, fire up ilives reader and cut the max commute in half.
There are of course many other variables involved in how your traffic simulator of choice works and many of them have a very strong relationship to their max commute time.

This was an important discovery for anyone starting a new sim "from scratch" but the existing sims, after their theory phase went through much testing and tweaking and so this may be irrelevant to A,B, Z and even some of you with your own unnamed sim. The bottom line is if you are happy with your existing traffic simulator then leave it alone.

I also wanted to make some edits for clarity, but it really irks me when people just delete things to try to "unsay something they obviously said" so I struck out a few items and made edits in italics.

ldog

Actually it appears I am not the first to notice this after all:
http://sc4devotion.com/forums/index.php?topic=4272.msg153585#msg153585
Chris also observed sims making full 6 minute commutes with vanilla back in June of 08.
Why this didn't go much further is beyond me. Chris aka clr1s4d only made 4 posts before he went into lurking (he was last active 4 days ago)