Transit Switch Entry Costs and Station Capacities

[z]

Transit switch entry costs and transit station capacities are two very important properties of virtually every transit building exemplar, yet for many years their optimal settings have been poorly understood.  This post attempts to clarify their usage, so that builders or modders of transit stations can use the correct values for these properties.  These values have been optimized for the current NAM Traffic Simulator ("Simulator Z"), which has been in use since 2009.  The values listed in this post have been adopted by the BSC in the PIM-X tool; the Transit Switch Entry Costs are currently implemented in the tool for stations serving a single travel type.  If you just want the numbers to use for these properties, it's not necessary to read this entire post; the numbers for the Transit Switch Entry Costs are grouped together in this section, and the formulas for calculating station capacities are near the very end of this post.  The intervening material explains why the numbers and formulas are set to what they are.  If you're not sure which number(s) should be used for a particular station, the explanatory material should make it clear.  I'll first discuss the Transit Switch Entry Costs, since this issue is a bit simpler.

Setting the Transit Switch Entry Cost to the proper value is very important.  If it is set too low, then pedestrians and possibly other travel types may cut across the lot, leading to what is commonly known as "shortcutting"; the travel time of the station's main travel type(s) is also distorted if the network travels through the station.  (This last point does not apply to subways, which work differently from everything else.)  If the Transit Switch Entry Cost is set too high, the result can be even worse, as after a certain point Sims may not use the station at all, and in some cases the entire transit line may become abandoned.  The cause of these problems may be very difficult to detect, and for the average user, it may appear that the traffic simulator or some other aspect of the game is malfunctioning.

What exactly is the Transit Switch Entry Cost?  It's the inverse of the speed in kilometers per hour for the referenced travel type.  The game is designed so that it works best when the Transit Switch Entry Cost (TSEC) for a particular travel type allows that travel type to move through a transit-enabled (TE) lot at the same speed that the travel type would travel at on the embedded network.  Based on this, and using the speeds for travel types found in the NAM Traffic Simulator, the proper TSECs for various transit stations are as follows:

Pedestrians and bus stops: .067
Street stations (RTMT): .04
Road and avenue stations (RTMT): .022
One-way Road Stations (RTMT): .015
Bus stations (road passing through): Use value for appropriate RTMT station above
Underground garages: .022
Above-ground parking lots and garages: .067
Highway stations: .0067
Subway stations: .0095
Freight rail stations: .0095
El rail and GLR stations: .0087
Passenger rail stations: .0071
Monorail, HSR, and Bullet Train stations: .0044
Stations adjacent to networks that do not run close together: TSEC of slowest network*
Stations adjacent to multiple networks that intersect or run close together: TSEC of fastest network**

*Note:  "Close together" means no more than two squares between networks.  If one of these networks allows only pedestrian access (i.e., there is no bus stop or car parking on the station), then the TSEC should be 0.067.

**Note:  In the case of adjacent stations, if the station is able to handle more than one network type running next to it, the TSEC should never be higher than the TSEC for passenger rail.

Bus stops have the same TSEC as pedestrians because buses don't actually drive through bus stops (although this is different for RTMT).  However, pedestrians walk through bus stops, so the TSEC is set for their speed.

Diagonal stations use the same TSEC values as their orthogonal counterparts, since in SC4 there is no real diagonal travel; it is all orthogonal.

The above values are incorporated into the current version of the PIM-X tool (SC4 PIM), specifically in the current version of the SC4PIM New Properties XML Update.  However, transit stations that service multiple travel types in addition to pedestrians must have their TSEC set manually.  For such stations, the following rule should be used; this rule covers all situations:

The formula for the transit switch entry cost for any station is 1/speed, where "speed" is the speed of the fastest travel type that is designed to pass completely through the station at normal speed (excluding subways).  If the travel type passes next to the station instead of going through it or if there is no through travel type (such as for stations that are adjacent to their networks or for parking lots), the TSEC should be calculated as described above.

Note that network enabled lots are handled differently from transit enabled lots.  The TSEC of a network enabled lot is 1/speed, where "speed" is the speed of the fastest network passing through the lot.

Some lots contain a mixture of both through networks and networks that contain transit stops.  These lots should be treated as transit enabled lots, and the proper TSEC will be the lowest of the TSEC for the networks that pass through the lot and the TSEC for the transit enabled networks.  The lot capacity will be the total of the standard TE lot capacity plus the network lot capacity as described below.

Note that the formulas for calculating TSECs and capacities are the same whether you are using pass-through lots or terminal lots.


Now on to capacities.  First I'll give the rationale, and then the various capacities and formulae.  If you just want to see the latter, you can just skip to the end of the post.

The way the game is presented to us, transit stations appear to act much like networks.  They have fixed capacities, and when these capacities are exceeded, transit stations appear to suffer their own form of congestion - their "service quality" declines.  Transit stations are even shown in the Traffic Congestion Data View.  When their usage is no more than 100%, they're is shown as green, just like networks, but as their usage passes 100% and their service quality declines, their color goes from green to yellow to orange to red - again, just like networks.  We are led to believe that a transit stations at 300% capacity, which shows up as bright red in the congestion data view, is performing quite poorly compared to one operating at, say, 30% of capacity.

This is all completely false, and Maxis knew it.  A station running at 300% of capacity has the same "service quality" as a station running at 30% of capacity; they operate identically.  Maxis actually could have implemented such a service quality approach fairly easily, by multiplying the Transit Switch Entry Cost by the inverse of the Congestion vs Speed curve.  But as we know, they didn't, and in fact the TSEC is used by Maxis only for toll booths.

If this were the whole story, nothing would need to be done here; the current capacities could be used for transit stations, and though their "service quality" would be displayed inaccurately, no harm would be done.  But it's been know for a number of years by some players that there's a maximum number of Sims that a transit station will process in a day.  However, the numbers reported for this limited varied, although they were always a certain multiple of the station's capacity.  Two years ago, I set out to verify that such a number existed, and if it did, to find out what it was; RippleJet later joined me in these experiments, and Tarkus participated in the thread where the results were presented in depth.

Although I verified that the limit existed, I did so mostly in trivial cases.  RippleJet's experiments were the first to use realistic station capacities.  His results can be found in this post.  Some of my experiments showed slightly different aspects of the problem; one of the more interesting ones can be found in this post.

Between these experiments and others that preceded and followed them, the following facts were determined:  All stations have a fixed capacity that is some multiple of their nominal capacity.  Until this fixed capacity is reached, they suffer no performance degradation; they are completely unaffected by high usage, even when it's well over 100%.  Once the fixed capacity is reached, however, the stations simply stop working for the day.  No more Sims are permitted through the stations via any network.  The stations are effectively broken for the rest of the day.

What is this fixed capacity?  Different experiments have yielded different numbers, as have different in-game situations.  The maximum capacity found by RippleJet's experiments was 400% of the nominal capacity, and although this number has been seen in various other situations since then, no lower maximum has been reported.  Somewhat higher maximums have been reported for specific stations, but no pattern has been observed.  Occasionally people report much higher maximums, but for standard stations, this seems to be due to the fact that the traffic simulator will let as many Sims who want to use the station use it the first time, and only in subsequent runs will it scale back the station usage.  This is in keeping with other known behaviors of the traffic simulator.

If a station breaks at 400% of its nominal capacity, we clearly don't want that to happen; that doesn't correspond to any RL event, and is essentially just a bug.  Therefore, we need to set station capacities high enough so that their usage never exceeds 400% of their nominal capacity.  How high is that?  In other words, what's the maximum amount of traffic that can be expected through a station?

Below is an example of a station in an ordinary city; this city was not designed to test station capacities, but merely provides a good example:

The station is mostly hidden by the query; it is a combination bus and tram station for tram-in-road, and you can see a bit of the tram shelter sticking out just past the number 980.  You'll also notice that for RTMT and NAM stations, "Service quality" has been renamed "Reserve Capacity," since a reduction in this graphic does not imply a reduction of service quality at the station, even in circumstances such as these.

One of the things that's interesting here is that this is no megalopolis; the population here is only around 150,000.  When you figure that about half of that is the city's workforce, and the query includes both commute periods, you see that almost the entire workforce of this city is going through this station.  Looking closely at the picture, it's easy to see why; this station gives access to the rail station, which is the main (and fastest) connection between residences and jobs.

If we were to do a straight extrapolation here, we would come to the conclusion that in order to support cities containing millions of Sims, seven-figure station capacities would be necessary.  But situations such as this don't arise in such cities (fortunately), as their increased density tends to greatly increase the number of ways for Sims to get from their homes to their jobs.  In reality, the station above is the most heavily used station I have ever seen.  I think that we would be safe in saying that no station would need a capacity of more than 200,000; if one did, there would be larger variants available, or multiple stations could be used.

Now we're ready to look at which capacities we should actually use.  There are three basic approaches here.  For the first one, if 200,000 is really a safe capacity for virtually all stations, and the true capacity is at least 400% of the nominal capacity, then we could simply use the following rule for nominal capacities:

All Stations:  50,000

From the point of view of game play, this would work out a lot better than the current capacity system.  The Maxis bus station has a capacity of 1000 and the Maxis subway station has a capacity of 2000, yet it is not at all unusual to see five-figure usage numbers for these types of stations.  This means that stations with capacities set this low, or even in this range, are going to break frequently.  Players may not notice the details of this type of breakage, but they do notice the red in their congestion maps, and undoubtedly the single biggest complaint players have about transit stations is that their capacities are too low.  When a subway line has a capacity of 30,000 (as in the High version of the NAM traffic simulator), and players have erected blocks of buildings where each building houses thousands of Sims, then a subway station every few blocks (which is much more frequent than in real cities) with a capacity of a few thousand just isn't going to cut it; a single tall residence could saturate such a station.  The Maxis transit stations simply were not built with today's custom content in mind, and they and other stations with similar capacities simply do not function well at all in that environment.

On the other hand, a single capacity for every type of transit station simply doesn't look good.  Eye candy is not a bad thing; a BATter's job is to create beautiful eye candy.  So in this case, it looks a lot better if we create numerical eye candy - capacities that are sufficient for each type of station, but which don't get exceeded, so that the station doesn't break.  This also allows us to use capacities lower than 50,000 for many types of stations, because many types of stations will never approach that usage level.

The question at this point then becomes:  What capacities should we use?  In terms of game play, the capacity required for a particular type of station bears very little relationship to the size of the station.  So for this purpose, we just need to know what the minimum capacity should be for a given type of station.  When I first started looking at the issue of station capacities more than two years ago, one guideline I kept in mind was that in RL, stations rarely get congested before the networks they serve.  The reason for this is that it is much easier and cheaper to overbuild a station than to overbuild an entire network.  But what formula should be used for determining these station capacities?

There have been many formulas proposed, but I found that none of them gave satisfactory results.  I think that the reason for this is that all the formulas I saw had station capacities increase linearly with the capacities of the networks that they served, while what I was seeing was that the proper growth curve was more of a logarithmic one.  I found that the formula used in RTMT by Cogeo produced capacities that seemed to come closest to actual usage needs, although being basically a linear formula, it produced numbers that were about right on the high end for complex stations, but were too low on the low end for simple stations.  I found that by doubling the low end numbers, leaving the high end numbers the same, and adjusting the numbers in the middle accordingly, I got capacities that worked extremely well.  As Simulator Z had not been built yet, this was all done using the CAM traffic simulator, which had capacities approximately the same as the High version of the current NAM traffic simulator.  Yet these station capacities still work fine today; the only change I've made is to reduce the bus capacity somewhat, since the potential volume of buses has been limited because buses now contribute to traffic congestion, whereas they didn't in the original Maxis traffic simulator.

The question may be raised at this point as to whether the capacities used for RTMT stations, and inline stations in general, are valid for all stations, since station usage for these two classes of stations includes all through traffic.  I have found the answer to be a clear "Yes."  The reason for this answer is that SC4 traffic does not behave the same as real world traffic; much of the differences can ultimately be pinpointed to the fact that most travel distances in SC4 are much shorter than in the real world.  As a result, for example, instead of taking a train for at least several kilometers, Sims tend to travel much shorter distances.  This means that virtually the entire traffic of a train line may enter or exit at a single stop, and this may happen at multiple stations along the train line.  For such stations, there really is no difference between through traffic and traffic that enters or exits the line at that station.

For buses, the situation is slightly different, but the ratio of cars to buses varies so greatly on any given road that I have found that here too, an RTMT-type bus stop requires the same capacity as a roadside bus stop.

Based on all of this, and on extensive experience with RTMT and traffic simulator experiments, the following base capacities for transit stations are recommended.  Where these transit stations are supported by RTMT, these are the RTMT capacities.

Bus:                      17,000
Subway:                24,000
Freight train:          15,000
Passenger train:      45,000
El rail:                    45,000
Monorail:                45,000
Parking facility:      30,000

The astute reader will notice that all of these capacities are lower than the 50,000 mentioned earlier; the reason for this is that each of these stations handles just a single travel type, whereas the 50,000 is designed for a broader range of stations.

The type of differentiation in the above table certainly seems better than the one-size-fits-all figure of 50,000, but it still makes no differentiation between different sizes of stations, nor does it account for stations that handle multiple travel types.  Originally, while accounting for this in the stations that I use and have modified, I had been doing this on a somewhat ad hoc basis, ending up with something that seemed about right.  In order to be more systematic, I have come up with formulas  that produce numbers very close to those that I have been using; these are the formulas that PIM-X will be using to produce reasonable station capacities for any type of station.  Unlike the TSEC, there is no "exact" correct capacity, so some rounding can be done by the developer.  For example, I had been using a capacity of 50,000 for the Maxis Grand Railroad Station; the formula that follows yields 51,000, but this is one of those cases where rounding off (in this case to 50,000) just looks nicer.

The formulas I have devised for station capacities follow.  For stations with buildings, only the first floor of the building is used in calculating the station's capacity, as additional floors are generally not used for passenger waiting space.  Another point to consider in these formulas is that the bigger the station is, the less distance the Sims have to walk to get to any part of it.  Since the whole station is fairly monolithic as far as the traffic simulator is concerned, bigger stations therefore do serve more Sims, though not by all that much.

Here are the formulas.  Note that in the case of overhanging buildings, all tiles beneath the overhanging building are considered part of the lot for these calculations.

Bus stops and stations:  17,000 for the first tile, plus 1000 for every additional tile on the lot.  Large bus stops may not have buildings as such, which is why the tile count of the lot is used.

Subway and Other Underground Rail Stations:  24,000 for the first tile, plus 2000 for each additional tile in the lot.  This time, we're talking about the lot instead of the building due to the underground nature of these networks.

Passenger train (may include freight): 15,000 per tile for the first three tiles of track served by the station, plus 3000 for each additional tile.  For stations serving multiple tracks, additional tracks are counted at the rate of 5000 for the first tile plus 2000 for each additional tile.  If the tracks are sidings and not through tracks, such as in the Maxis Grand Railroad Station, each tile of siding is counted at 1000 per tile.  Buildings attached to train stations also add 1000 per tile per floor.

Freight train only:  10,000 per tile for the first three tiles of track, plus 2000 per tile after that.  Sidings add another 1000 per tile.  Nothing is added for the size of the lot; freight lots tend to contain all sorts of things, most of which just sit there forever.  Also, the size of the lot is somewhat proportional to the number of tiles of track served, so it's implicitly counted to some extent.

El rail:  Same formula as for passenger rail.

El rail over road:  Same formula as for passenger rail, but with an extra 12,000 added to cover road through traffic.

El rail over avenue:  Same formula as for passenger rail, but with an extra 24,000 added to cover avenue through traffic.

Monorail, HSR, and Bullet Train:  Same formula as for passenger rail..

Parking facilities:  1000 per tile.  For multi-level garages, only the tiles containing the garage count, and their number is multiplied by the number of levels of the garage.  Underground garages have a flat capacity of 50,000, since they don't have any real dimensions.

Network enabled lots: These are lots that allow network traffic to pass through them, but do not permit any transitions with the lot. Therefore, these lots are not actually transit stations, although they are constructed according to the rules used for transit stations. The capacity of a network enabled lot is the sum of all the individual networks (including networks of the same type) entering the lot.

Multiple networks: There are also rules for combining these basic types.  For example, adding a parking facility adds the standard 1000 per tile for the parking lot; if it's a garage, the standard garage rule applies.  The one exception here is a parking lot added to a subway station, as it implies a bigger station.  In this case, it's 2000 per tile of parking lot; for a garage, it's 2000 per tile for the first level, but the standard 1000 per tile for additional levels.  Adding an underground garage to a station adds 15,000 to its capacity.

If the station serves multiple networks, the basic rules are used for the highest capacity network.  After that, for additional rail types, 10,000 is added for the first tile of the first rail line, and 2000 is added for each additional tile.  (For subways, the total length of all the above-ground rail lines in the station is counted.)  For buses, 4000 is added for the first tile of the rail line, and 500 is added for each additional rail tile.  Sidings and buildings are not counted here; they've been counted once, and that's enough.

Finally, always round off the capacities to the nearest thousand, and if they're over 50,000, round them off to the nearest 5000.  The capacity formulas aren't exact, nor do they need to be.

That should cover it.  Although the station capacities produced by these rules are large, they are just large enough to assure that stations won't break under any but the most extreme usage.  They also produce station capacities that rise nicely with station size and layout.  Finally, they happen to produce capacities very close to the ones that have been used all along in many test cities going back several years, and which work very well in a wide variety of situations.

[deadwoods]

Thanks for a concise and handy reference.

One question... How do these capacities, formulas and settings work for transit-enabled lots that aren't stations? For example I have some lots that have bridges over road/ave/rail in the BSC Park set. From a pedestrian perspective these are just eyecandy, but I need to set the Transit Switch Entry Cost and Transit Switch Traffic Capacity values for the road/ave/rail that my eyecandy bridge lots are interrupting. The sims aren't getting into cars/trains, but cars/trains must run through the lots.

[z]

That's an important topic that I didn't cover originally; RippleJet calls these "network enabled" lots.  The easy part is the Transit Switch Entry Cost, which is set in the same way as for transit stations.  The harder part is the capacity, since we don't know what the capacity of the user's traffic simulator is, and even if we did, the user may change it at any time.

Fortunately, these are the very same issues that I encountered with RTMT years ago, and the solution I adopted there seems to work quite well; it's also the basis for the station capacities listed in the main post, so it fits together well with everything else too.  I simply use the capacities of the High traffic simulator.  For those people using the Ultra capacity, this may be a little low, but not significantly so; for those using lower capacity simulators, this may be a bit high, but generally this is not significant, especially since the lot usually covers only a few squares of network at most.  In any case, the difference from the ideal is almost never enough to make any difference at all in the traffic simulation.

The general rule, then, is that for network enabled lots, the total capacity of the lot should be the sum of the capacities of all networks passing through it, excluding subways (which don't interact with this type of lot).  The network capacities should be taken from the High value of the NAM Unified Traffic Simulator; these capacities can be found listed in the first post of the NAM Unified Traffic Simulator and Data View Help thread.

There are some types of network enabled lots where a lower capacity may be appropriate.  For example, some industrial lots have roads going through them that are used primarily or exclusively by freight trucks.  Such lots typically contribute more freight to the road; it's easy to see that this process would cause a slowdown in traffic.  In such lots, using the Medium or even Low capacity could be appropriate.  Alternatively (or even in addition), the Transit Switch Capacity Effect could be set to True for the building exemplar in such lots; this causes the main building to act like an intersection, and cause a reduction in capacity in nearby network tiles the same way the Intersection and Turn Capacity Effect does.  Using the Transit Switch Capacity Effect property has the advantage that the capacity change is more gradual (although typically more severe at its height), whereas simply reducing the capacity affects every network-type tile in the lot.

[dahemac]

Z: Passenger train (may include freight): 10,000 per tile for the first three tiles of track served by the station, plus 2000 for each additional tile.

So, what would you do with Marrast's passenger rail stations? The 1x6 in-track station perhaps should be 10,000 X 3 + 2000 X 3 = 36,000? It is currently 10,000.

It seems to me that a lot of existing transit lots, like SG Bus Stops, discussed also here: http://sc4devotion.com/forums/index.php?topic=12761.0 have tragically low capacities.

[z]

So, what would you do with Marrast's passenger rail stations? The 1x6 in-track station perhaps should be 10,000 X 3 + 2000 X 3 = 36,000? It is currently 10,000.

Yes, 36,000 is the proper capacity for these stations.  I've used these stations and seen that volume of traffic pass through them.

It seems to me that a lot of existing transit lots, like SG Bus Stops, discussed also here: http://sc4devotion.com/forums/index.php?topic=12761.0 have tragically low capacities.

Agreed.  SimGoober was merely using the same capacity as the Maxis bus stop, which for the original game is reasonable.  But with the higher traffic caused by the large populations of many of the custom content buildings, along with the capability of the NAM traffic simulator to move larger volumes of Sims more effectively, such a capacity is now quite inadequate.  For years, the NAM has been releasing traffic simulators with network capacities many times that of the original game; it only makes sense that such networks would require higher capacity transit stations.

[dahemac]

I am going through with Reader and updating the capacities. By your scale the Marrast 2x6 rail station should have a capacity of 51,000. These capacities, they are per month correct?

[z]

I am going through with Reader and updating the capacities. By your scale the Marrast 2x6 rail station should have a capacity of 51,000. These capacities, they are per month correct?

Station and network capacities in SC4 are per day.  They're much higher than RL capacities simply because of the way the game is built.

[dahemac]

For a laugh I looked up the daily RL capacity of Union Station Toronto. 200,000 go through there every day. 

[z]

That's pretty impressive.  So maybe I should change that phrase in my last post to "generally much higher". 

[z]

The Mass Transit Authority (a wholly-owned subsidiary of The NAM Corporation^TM) has recently been hard at work improving transit switches for its stations.  These can be seen in the NITS (NAM Ideal Transit Switches) of NAM 31.2, but recently even further progress has been made, resulting in stations that are able to process previously unattainable levels of traffic.  The full details of how these transit switches work will eventually be added to the first post in this thread.  However, for now it has already been necessary to increase the capacity levels for many stations, as some stations built with the old rules were running as high as 230% of capacity in basic test cities.  In order to avoid usage levels such as this and even higher in actual game cities, the following changes were made:

It was made explicit that all tiles that an overhanging building covers are counted in the capacity calculations.

For passenger trains, the 10,000 per tile for the first three tiles was changed to 15,000, and additional tiles were changed from 2000 to 3000.

All the numbers for freight trains were doubled.

A separate entry was made for El Rail over Road, which still uses the same capacity as El Rail, but then adds an extra 10,000 to account for through road traffic.

Monorail, HSR, and Bullet Train now use the same capacities as passenger rail; there is no longer a 50% premium added.

For stations serving multiple networks, for additional rail types, 10,000 (instead of 9000) is added for the first tile of a rail line, and 1000 (instead of 500) is added for each additional tile.  (For subways, the total length of all the above-ground rail lines in the station is now counted.)

Everything else remains the same.  The new numbers provide a better distribution of station capacities, and help ensure that stations are unlikely to exceed their nominal capacity before their associated networks do.

The first post in this thread has been fully updated with the new numbers; people wishing to construct NAM-compatible stations should use these numbers.

[z]

The first post of this thread has been updated with more accurate values for the TSECs.  The new values are within 4% of the old values, so it's not essential that current stations be updated, as the effect of the new values are minimal.  However, if you are building new stations, please use the new values.