learning from “mini metro”

Mini metro logoI've long wished there were an computer game that would require players to figure out the basic facts of transit network design.  I don't mean complex simulation games like Cities in Motion, Transport Tycoon or (shudderSimCity, which simulate so many things that it's hard to focus on the network element.  I mean a game that is simple but engaging the way chess is, and where the strategy you need to learn happens to also be What City-Makers Need to Understand About Transit (but Often Don't).  

Games are a good way of thinking about real problems (see Jane McGonegal's great book Reality is Broken).   And there are plenty of geeks out there, in city-building professions and advocacy, who'd enjoy learning this way.  I use (non-computer) games in all my courses and stakeholder workshops because things you've played with are things you remember.

Well, now there's a draft of the very game I've been imagining, thanks to Peter and Robert Curry of the New Zealand gameshop Dinosaur Polo Club.   Their game, Mini Metro, is  simple, fun, and (if this is a virtue) addictive.  And if you notice what's going on, you'll learn some sound principles of network design that will serve you well, no matter what your role is in creating the cities of the future.  

This post is not a critique of Peter's and Robert's work.  They eagerly emphasize that they're not transit experts; they're just clever blokes who set out to create a fun game about transit, and who succeeded.  My focus here is twofold.  First, let's notice how accurate a game that set out to be fun turns out to be.  Second, let's think about what steps you'd take to make it a little more accurate, if you wanted to.  

Here's how it works.

You choose one of several world cities, which basically gives you a famous set of water features that will require bridges or tunnels.  Start the game, and little shapes start appearing.  These are potential stations, with demand for travel to other places.  Click on a shape and connect two of them, and you've built a line, where a short train starts running back and forth.  Before long, you have something like this, from a recent session of the Cairo game.  

Mini metro cairo early

(If you don't like the background color, play one of the other cities; I think this one is trying to suggest desert sand.)

Early in the game, you see circles and triangles, and a single square.  Next to each symbol, you'll see little black symbols accumulate.  Those represent someone at that station who wants to go to a station of the specified shape.  For example, in the image above, at the circle station at the south end of my Red Line, above, someone is waiting who wants to go to a triangle.  

The demand-generation rules of Mini Metro are oversimplified, but not disastrously so.  People always want to go to a station of a different symbol.  Most demand is between a circle and a triangle.  You'll note that in the sample above, I've arranged a radial network so that identical symbols (between which there's no demand) are never adjacent on a line, and so that each circle or triangle is connected to a nearby symbol of the other type.    

Both circles and triangles also generate demand to squares; initially there's only one square but gradually more appear.  If you like, the square is downtown, the triangles are local retail/activity centers, and the circles are residential areas.  Because all stations generate demand to squares, and initially there's only one, it's logical to start by building a radial network, like the one above, with the single square at the hub.

So far, the game has presented two obvious oversimplifications.  First, there is no "mixed use," but this is accurate in a way:  Players experience the transport demand problems created by monotonous single-use development patterns.  The residences (circles) are all over here, and the shopping (triangles) are all over there, and a bunch of important things are only downtown (squares), so everybody has to get from here to there, over a distance too far to walk.    So there's lots of need for transport in this city, even from the earliest days!

Second, the society represented by Mini Metro displays some strongly fictional features.  A black triangle by a symbol means someone wants to go to any triangle, not a particular one.   Not only are people content to go to whichever shopping center the next train is going to, they feel the same way about jobs and even about homes!  Nobody  in this city wants to go to their own home+family, but just to whichever is easiest to get to on the next train.  Clearly, we are in a science-fictional society that engages in spontaneous partner-swapping and communal parenting on a citywide scale.

But does this simplification get in the way of the accuracy of the game in conveying how earthbound transit networks are designed?  Only to a point.  It makes the game much easier than real network design is.  In Mini Metro if a circle is connected directly to any triangle, it doesn't need to be connected to others, whereas a real residential area will generate specific demand to each of the retail/activity centers nearby.  

On this point, the game encourages a  lazy design habit that creates problems the real world: focusing  on the large-scale flows of demand to the point that you neglect the fact that people are going from everywhere to everywhere.  I can't tell you how often I've seen someone point to a map and say "these people, they're all going over there!"  What that really means is that above the high level of everywhere-to-everywhere demand, there's some preponderance of demand from this place to that place, but this pattern is rarely as big as it's made to appear.  Some people are just uncomfortable thinking about the reality of everywhere-to-everywhere demand, and need to oversimplify.  Be suspicious when you hear statements of that form. 

Look again at that first screenshot above.  At this stage, I've gotten to nine stations with a totally radial system, and it's working fine, but that's an artifact of the game's oversimplification.  In the real world, someone in that circle at the west end of the Blue Line would want to go to the triangle on the same side of the river — because in a real city, people are going everywhere.  Mini Metro makes that not my problem as a designer, but in a more accurate game it would be; we would see more complex demand patterns, perhaps with many people willing to go to the nearest triangle but some needing to go to specific ones.  

In the upper right of the screen is a timer that counts off days of the week.  Every seven days, you get one "locomotive" (which is to say, another train you can add to an existing line, increasing frequency).  You also get either a new line, a new carriage (which adds to the capacity of an existing train, or a new tunnel or bridge to extend another line across the water body.  I like the Cairo and London games because they don't need a lot of tunnels, which enables me to focus more on network structure.  Mini Metro also includes cities with much more complicated water barriers (Osaka, Hong Kong, Auckland) if you like dealing with those.

The object of the game is to carry as many people as possible to their destinations before any station exceeds its capacity.  In the image above, I've already carried 36 people in this early stage of the game, per the counter just to the left of the clock in the upper right.   If people accumulate at a station faster than trains come to pick them up, the station is marked as critical and you have some time to intervene.  Eventually, though, the game ends with so many stations over capacity that you can't manage to get a train to them all soon enough, and that's the end.  The number of people you've carried by that time is your final score.

This is not, of course, how real transit networks die.   But it is true that capacity is considered a must-solve problem by many transport engineers.   Meanwhile, it's very realistic to score a game on how many people reached their destinations.  That's access, after all, or if you're a businessman it's also fare revenue.

When I say that some engineers think capacity is a must-solve problem, am I implying that it isn't?  Well, in the real world, some capacity problems are so expensive that they are not worth solving, so people stop travelling.  Likewise, if you add a lot of new capacity, you will get more people using it.  These are the phenomena of suppressed and induced demand, respectively.  Unlike the citizens of a Mini Metro city, real people have a range of choices, and choose transit or not based on how it stacks up against alternatives.  (The same things happens with roads, of course.)  So more capacity attracts more people and less capacity pushes them away.  There's a longer term land use impact of this process that can be positive.  If you don't keep meeting a very expensive capacity demand, the development pattern will eventually adjust to one that doesn't require so much capacity — for example, one that fosters generally shorter commutes by putting jobs and appropriate housing closer together.

The Mini Metro world has none of that nuance; instead, it runs on the assumptions of classical transport engineering.  Demand is envisioned as being like floodwater: it's coming at you in a predictable quantity, nothing you do can change that, so you just have to handle it somehow.  So this is another simplification that can mislead.  

Let's take the game above a little further.  Again, here is the radial network that evolved early in the game, with an alternation of circles and triangles on every line.

Mini metro cairo early

 

And a bit later it looked like this:

Mini metro cairo early plus 1

Something interesting has happened on the west side of Cairo.

In the first screenshot, the Green Line west of the river had only one circle to serve, so I routed it through the triangle to end at the circle.  But as more station areas appeared — that is, the city grew due to single-use development — I was presented with a problem.  The original circle was south of the triangle, so I had turned the Green Line south to end there.  The next circle that appeared, however, was northwest of the triangle, and a reasonably direct line could not serve both.  This is the "Be on the Way" principle, probably the single fact about transit that the city-building professions are least likely to understand.   The "be on the way" principle says: "If you want your new development to have great transit service, put it on a direct path with other places that do."  

In this case, I can imagine what happened.  The developer of the first circle, south of the triangle at the west end of the Green Line, took it for granted that the green line would be extended to him, and it was because he was there first.  Meanwhile, the second developer, northwest of the triangle, assumed that the Green Line could be extended to him.  But it couldn't be extended because the triangle was no longer end of the line.  (Mini Metro doesn't allow branching, but if it did that wouldn't really solve the problem because branching divides frequency.)

So the appearance of that one circle northwest of the triangle required me to rearrange the whole westside network structure.  I extended the Blue Line north so that it would cover both southern circles and end at the triangle, making it possible to extend the Green Line to the new circle out to the northwest.  It's never good to have two identical symbols connected, because there's no direct demand between them, but often it's unavoidable.  Here, it's manageable because once the Blue Line extends to a triangle, its trains will pick up triangle-bound people from these two circles in both directions of travel.  This doubles the effective frequency for getting to any triangle, which is all that our flexible citizens seek to do.

Let's notice some other things that have happened by the time of the second screenshot, above.  Apart from the westside problem I just described, I've been able to arrange the rest of the network so that stations visit circles and triangles in alternation.  This is a very efficient use of capacity, because many people ride for just one station; that means we turn over the load quickly, serving many people with few trains.  By contrast, if we have a series of circles with no triangle, each circle's demand adds to that of the previous circle, so we hit the capacity limit of the train sooner, and start leaving people behind.

Is it good that the game's goal is to optimize capacity rather than optimize customer travel time? Real transit agencies care about both, and at low levels of demand (early in the game, say) they're very different.  But they do tend to converge, mathematically, as cities grow.  The key elements of transit travel time are speed and frequency, but these are both elements of capacity as well.  Obviously, too, a capacity failure means that a train leaves the station without you, and that affects your travel time as well.

Less realistic is the game's construction of costing.   In the game, adding a carriage to a train is more expensive than adding another train, where in reality the opposite is true.  In the game, the cost of a carriage is the same as the cost of a tunnel/bridge — which amounts to confusing millions with billions — and it's also the same as the cost of a "new line."  Lines, of course, do not make sense as cost units at all.  Their cost is entirely a function of their length, the vehicles on them, etc.  "Line" is just a unit of nomenclature.  (Once more with feeling: When a transit agency brags about how many lines or routes they run, they are not bragging about the quantity or quality of their service; they are bragging about its complexity.)  Finally, of course, capital and operations costing are conflated, which is not necessarily fatal in a game this simple, but would need to be thought about. 

Now let's skip well ahead in my Cairo game, to see how the network evolves.

Mini metro cairo early plus 2

The city kept adding more station areas — new development nodes — and I've been forced to extend service to them.  In particular, more squares have appeared, signifying multiple regional centers — like the original downtown — that draw demand from longer distances.    Also, many more unique citywide destinations have appeared, such as the cross in the southeast part of town that presumably represents a hospital.  Third, there are more maddening cases where the various demand generators are not conveniently balanced.  West of the river, for example, there are now ten circles, three triangles, and no squares.  That means that it's impossible to set up lines with an ideal alternation of circles and triangles, because there aren't anough triangles per circle.  It also means  everyone from that part of town who's going to a square (or any unique citywide destination) has to cross the river, which means long-distance demand flows that will consume lots of capacity on all the east-west lines.  

That, ultimately, is the problem that will cause this network to fail.  When a station goes over capacity because it's impossible to run enough trains across the river, everyone will yell at the transit agency.  But the problem is a land use problem.  If there were a square on the west side of the river, and the triangles were spread out more widely, the network could be more efficient, which means a lot more people could have accessed their destinations, and gotten on with their lives, without requiring so much expensive transit capacity.

A great thing about Mini Metro, then, is that it accurately conveys the exasperation of transportation planners about transport-ignorant land use planning.  Sometimes, there's nothing but circles in one part of the map.  You can either put them all on one line, which will go over capacity at once, or run a whole bunch of parallel lines to different circles, at huge expense.  Either way is inefficient, and the emotions you are likely to feel about this impossible land use pattern — analogous to many square miles of residential with few destinations, all requiring travel in a single direction at the same time — are definitely putting you in a transport planner's shoes!

What else has happened in the game by this time?  As a result of the city getting larger, the network has not only grown but changed.  Once other squares appeared, it was no longer necessary to run the whole network to the original central square.  Instead, the demand pattern called for a grid, the ideal network structure for a city with balanced everywhere-to-everywhere demand.  Of course, the demand is still not balanced, because of the land use patterns, but as the city has grown it has grown more diverse, with more different kinds of destinations in many areas, so this is moving toward an approximation of everywhere-to-everywhere.  

If you can't see a grid in the screenshot above, look closer.  From top to bottom, the east-west elements are the Grey Line, the Blue Line, and the Orange Line.  From left to right, the north-south elements are the Orange Line, the Green Line, the Red Line, and the Grey Line.  The Grey and Orange serve as both because they are L-shapes.  L-shapes are fine in a grid as long as they main bend of the L occurs on the edge of the city and not in the middle.  

They key requirements of a true grid are observed:  Each north-south or east-west element continues all the way across the grid so that it connects with every perpendicular element.  This means that all trips are possible with a single connection.  This is a good thing for customer travel time, but it is also a good thing for managing station capacity, the ultimate object of the game, because connections are a main cause of station overcrowding.

If you compare the last two images, you'll see that in evolving the grid I've had to change the network substantially, rather than just grow from the orginal radial network.  Earlier we discussed the Green Line on the westside, but now the Green Line doesn't go there at all.  In the earlier radial system this line went north out of downtown and then turned west.  Once another square appeared in the north-central part of the city, it made sense to let the Green Line continue north to that and create a new east-west Grey Line.  In fact, much of the original radial structure has been revised.  

Grids, I should add, aren't always the answer.  The game includes a couple of cities, notably Auckland and Osaka, where there are so many water barriers that the need to conserve tunnels imposes a dramatically different shape.  Auckland, a series of ithsmuses, seems to work best with local circulators and strong axes linking them (the trick is to put all the citywide destinations on the axes, which are then one transfer from almost anywhere).  

Auckland 1

Indeed, the essence of what's fun about this game is that you have to keep revising your notion about what's a good structure.  You don't know where or what the new station dots will be.  You may develop one structure to deal with a poor circle-triangle ratio, as on the west side, but then more triangles will appear and you need to change the structure substantially to take advantage of that.  Playing the game, you notice your attachment to your network ideas, but winning lies in letting go of a network pattern as soon as it no longer fits the growing city.  If there are not too many water barriers, grids are almost always the answer, but what you're seeking is a marriage of the grid principle with the actual geography and land use, and that marriage is always evolving.

The final phase of the game is, of course, the most frustrating.  Demand grows everywhere, and as stations fill up with waiting passengers, little clocks appear at them counting down to the moment when they fail and you lose the game.  A basic strategy is to keep one line in the bank, so that you can create a short, temporary line wherever it's needed as a capacity problem occurs.  If a station is over capacity, and I can see that no train with enough spare room will get there in time, I can build a little two-station line from that station to wherever its people most need to get to.  It runs one trip to evacuate the station, then I remove it and deploy the same line at the next critical station.  (In the Auckland screenshot above, the pale magenta line is one of these temporary relief lines, which I'd just removed.)

We have discovered the most realistic thing about Mini Metro:  If you want to win, think of these "trains" as buses.  

In real rail transit systems, you cannot simply abandon a rail line and build a new one — certainly not just to handle an overcrowding problem.  But to do well in Mini Metro you must revise the network repeatedly, and the last phase of the game you'll deploy lots of one-time-only temporary lines  In fact, for best results, make sure you also have a spare tunnel, so that if you have to get a train quickly to a station on an island, you can build a temporary line to a destination across the water, deleting it after use.  

To a rail engineer, all this is ridiculous, but to a transit network designer, it's the game's most realistic feature.

Build a subway line to run one train once, then tear it out?  No, this is not how rail transit works, but it's very much how buses work, and it's good thing, too.  That's why buses provide a much better sandbox for network design thinking.  When you build powerful networks with buses,  mistakes cost thousands rather than billions, so they're more likely to be repaired.  Real-life transit networks do need to evolve, usually from radial beginnings to more gridlike structures.  That's been theme of many of our recent network redesigns — in Houston and Columbus and even (if you squint) Auckland.  

Of course, this is not a bus vs. rail post; I hope I've established by now that "rail vs bus" is a false and boring dualism.  Rail is needed in many places, but rail mistakes are much bigger than bus mistakes.  When I advise on rail networks, for example, I'm always trying to make sure that planned lines make sense in the context of a much larger city and network of the future.  Grid patterns are safe because they work in so many possible future land use patterns, but I sense danger when a city proposes to build too many radial lines, or proposes short rail fragments that might one day be extended to a useful scale but that turn out to be net obstacles in the meantime (Hello, Toronto!).  

So while it's fun to imagine you're drawing a metro, Mini Metro rewards revision, which means it's about the kind of design that requires the flexibility of buses.  Buses are not flexible forever; well-designed networks do finally settle into successful patterns that become permanent, which means they could safely be converted to rail in the future.   But for now, your rail lines are really bus lines — or, since this is a science-fictional society — a sexy transit tool that's low-infrastructure and hence easy to revise. 

In my recent correspondence with Peter Curry, he mentioned that he was thinking about revising the game to make it more expensive to change structures that are already built.  I hope he doesn't, because that will make it a very different game, one that's about infrastructure instead of networks. Infrastructure is expensive, permanent, rigid, consequential, hard to make decisions about, and therefore best used only when necessary, which it often is.  Networks are full of possibility, are focused on liberating human beings, and for some of us can also be endless fun to think about.  Besides, there are already plenty of infrastructure games.  When it comes, I'm pretty sure that the ideal network planning game — and online teaching tool — will learn a great deal from what Peter and Robert Curry have done in Mini Metro.  

9 Responses to learning from “mini metro”

  1. Mark December 29, 2014 at 8:46 pm #

    FYA you linked to two VERY different versions of SimCity. SimCity (2013) is the terrible version that should rightly be derided, not SimCity 4, which is a veritable masterpiece of complexity and elegance.

  2. lagatta à montréal December 30, 2014 at 2:53 am #

    Clicking on this, I thought it was about the “MiniMetro” in Perugia… But the game is interesting, while flawed as you say.
    I live near the Orange line in Montréal. It is saturated now heading into the city centre, since the extension to the large suburb, Laval. While it is wonderful that so many Lavallois have chosen the métro over commuting by car, it means the line is saturated by the time it reaches Jean-Talon station.

  3. Justin December 30, 2014 at 8:32 am #

    @Mark, in terms of public transit (otherwise it’s a great game), SimCity 4 is just as lousy as SimCityville. The buses and subway act like glorified taxis; simply building a few random stations here and there will result in buses and trains heading to said stations, regardless of geometry or distance. Which is the exact opposite of Mr. Walker’s “Be on the Way” principle.

  4. Emmett January 3, 2015 at 8:49 pm #

    Have you looked at Simutrans?
    http://www.simutrans.com/en/
    Just asking as I haven’t had a chance to look myself.
    I was a huge fan of Locomotion, but it is rail-based and similarly I was out looking for a simulator for meter transit.
    Meantime I’ll go look at Mini Metro.

  5. Irwin January 21, 2015 at 4:56 pm #

    “L-shapes are fine in a grid as long as they main bend of the L occurs on the edge of the city and not in the middle.”
    This is kind of interesting observation. I’m curious what you think of Taipei MRT’s network design that is adding multiple L-turns in the middle of the networks.
    The system went from radial network with 3 lines converging at Taipei Main Station in the early 1990s to this beast (see link). I’m not sure what it is… a hybrid grid?
    http://en.wikipedia.org/wiki/Taipei_Metro

  6. Kari March 24, 2015 at 2:17 pm #

    I love Mini Metro, but one interesting thing about the game (that is not realistic) is the fact that it is a system of transit that is both a train and a bus at the same time. What I mean is that lines cannot cross one another except at stations, which is very train-like, but the lines can be easily moved around, which is very bus-like.
    Does anyone have a suggestion that would make the rider’s transit goals (triangle/square/circle) more realistic than wanting to go to just any triangle, etc…? I’ve thought about this issue, but I don’t have any good ideas yet.

  7. Curtis April 16, 2015 at 11:27 am #

    Zoning could approximate unique destination design. Though it could seriously alter the minimalistic interface design depending upon how it is implemented.
    What I mean by zoning is this: apply some kind of unique pattern/texture/number to land parcels. Apply that same pattern/texture/number to the passenger shapes to indicate their desired destination. The passenger will take the route necessary to arrive at any matching shape located in an area of matching pattern.
    So, polka dotted squares won’t end their journey at just any square; they will go to the square in the polka dotted land zone.
    I realize that this could drastically change the AI, but it would communicate each passenger’s desire to the player without introducing too much visual complexity.
    Implementing zoning could be based purely on geometry, like a predefined grid. Or the zones could be somehow related to the first instances of unique shapes. If Squares really are like ‘downtown’ then the area around the first square might be patterned/textured/numbered as a downtown area. In other words, the first instance of a certain shape would also spawn a surrounding zone. This kind of structure would be much less predictable than a static zone grid.
    Of course, a truly unique solution would simply require a number for each shape. The first square is Square1, the second Square2, etc. but this would quickly serve to inundate the player with information.

  8. S M Sabri Ismail January 27, 2016 at 10:17 pm #

    {{ [Peter Curry] mentioned that he was thinking about revising the game to make it more expensive to change structures that are already built. I hope he doesn’t, because that will make it a very different game, one that’s about infrastructure instead of networks. Infrastructure is expensive, permanent, rigid, consequential, hard to make decisions about … }}
    That is certainly true. In recent version of Mini Metro (current Gamma 4, 22 Dec 2015), there are additional two modes …
    (1) Normal mode – don’t let station overcrowd
    (2) Endless mode – stations do not overcrowd, bulild for efficiency
    (3) Extreme mode – don’t let station overcrowd, tracks are permanent
    … the extreme mode does makes the play harder and make it feel more like building permanent rail network rather than flexible bus network.

Leave a Reply