Hello guys, I’m a level designer, and more specifically in racing games. Founder of my own studio I’ve been working for several years with Kylotonn Games, the french developers behind the rebirth of the WRC licence (WRC5, WRC6 and WRC7). Now specialized in racing, Kylotonn learns and builds knowledge about everything on wheels, from motocycles in Tourist Trophy, to arcade titles like Flatout4 or a more simulation kind with the WRC series. 
This article is about how we've built the WRC7 tracks and how we worked - how we managed  the workflow of WRC7's level design.

1- Understanding before producing

References, references references! We aren’t artist creating realistic assets but we still badly need to search, analyze and understand references. 
In WRC7 we’ve got 13 differents rallies, each one in a different country. They aren’t "closed tracks" like Nurburgring or Imola where a tarmac look and feel like any other, with the same signs, with perfect curves and only one path. Every country has a specific road legislation, with their own roads. In Mexico you’ll have bumpy sand roads with curves and a smooth surface leading to the top of the mountain, but in Germany you’ll be confronted with smooth and narrow tarmac going up and down hills and a lot of corner intersections. Just like artists we have to know and most of all understand why the road has been made this way.

We also share and cross references with artists, because level designers and artists don’t look for the same thing, and every time I ask an artist about a country, the first remark is about the vegetation. We don’t care about knowing the names of trees or grass, but we do care about where and which kind of vegetation grows, because for us vegetation means obstacle for the car and bad readability of the road. A dense vegetation close to a road can also be sign of a bumpy road because of the trees roots growing under, or mean a slippery road due to the mud of the side road coming on the tarmac. Two brains always means more information, no matter if it’s another designer or not; and even if you didn’t learn anything on the country talking with an artist, maybe he did, talking with you.

In Argentina the rally takes place in a rural region of the country, somes of the stages take place at the top of mountains up to 2700 meters of altitude where there is no interest for the locals to build proper tarmac roads because the road basically leads nowhere and is no longer used by car traffic. Some roads have been created one hundred years ago with pickers. There are deformed rocky roads at the top and smooth gravel in the valley. Once you learned that, your design starts to take shape, you are no longer making a track or a road, but recreating a true and credible part of Argentina. 
In the case there isn’t any story to tell, make one. Create and shape one that the player can understand and above all “feel”. The scenario of your track or your stage is very important for the player but for you too. It drives your design and shapes the environment. Maybe the road goes through a perfect desert and you won’t have any clue what to begin with. Maybe there is a simple ghost town on the way, a canyon, a drive in between two sand dunes, a wider road with wet sand on 1 mile. The more you approach an oasis the more there will be secondary roads connecting to the main road with maybe wet sand tire marks. Maybe the oasis is surrounded by very high sand dunes visible from miles away. 
Long range landmarks are important too: having mountains on EAST of your track will allow the player to know the direction is going, he’ll know that at the beginning the mountain was to his right and now they are upfront.
Obviously making a racing game about the Dakar would be harder in level design because there is only few credible tools, where a “sprint race” from the south to the north of Paris would offer a wider panel of tools for your story. 
There is always a story.

Maybe you aren’t working on a realistic game and credibility is none of your concern. From my point of view it still is. Every video game player has come out of his room the day his internet went down to explore the real world, seeing roads, cars, accidents, mountains, vegetation... Even if they don’t, they saw a film recorded in Australia or stared at a photoshoot of a journalist in Poland. We all have knowledge, that’s why when you see the Piz tower you instantly know there is a problem, your eye and brain tell you there is one even before you start analyzing what the problem really is. Same thing for buildings with unconventional architecture; the human eye will recognize a building in 100ms but it will probably take two or three times more if the brains has to interpret it.
I’m not saying you have to make a realistic game, only a consistent one. A five branches fork is credible but a one fork fork looks too far from the concept we are used to know.
This means when you do something not credible enough, too different of the reality, you are disturbing the player, increasing the time needed to read the road ahead or reacting; and while driving in a 380 horsepower car driven by a mad man, a fraction of second is the difference between life and death. 
Well. 
Respawn or stage record at least.

2- Being an artist

My skills in photoshop are...well...particular...and limited. Okay, it’s awful, but it doesn’t mean I can’t imagine what it could look like. Art and level design are very close in racing games and that’s why we have to think about the ones who are going to level build each track. We have to create opportunities for artists to seize. I’m sure each one of us have read, watched or saw enough stuff in our lives to imagine what can be great. Beautiful vista from a mountain, small lakes with a dangerous road connecting them, driving on a barrage… even if it’s empty when we design the tracks, it has to be clear, we need to have a plan to communicate to the rest of the team. Often we put placeholders to replace a gameplay element, but it also can be used to illustrate the graphic ideas. Very often we went to the Art Director asking him what he thought of this vista. Do you like it? Is it representative of the country? How risky is it technically? Is that time consuming? We have to ask ourselves these questions before starting the level design in order to create opportunities.
We’ll see later the idea of “scenario in level design”, and it’s directly linked to think as an artist.
So, even when we started a new track, we used some generic assets previously made: houses, rocks, fences, lakes… artists will replace them pretty quick but it will improve their comprehension and help us test the track in real conditions. It happened that we tested the tracks without any art, not a single tree or a house ; only a road on a heightmap. 100% of the time “micro-tweak” we could have done at this stage were pointless. From my point of view you can’t test a track until you have the road AND what comes next to it. Do the test: make someone drive on a highway alone. Now on the same highway put one truck on his left and another on his right: his lane is still the same size but he will feel trapped, in danger, surrounded by a dangerous environment where he can see the consequences of an eventual mistake. Obvious, right? so do not test your tracks if they are still naked. 
I’m getting carried away...bottom line: It’s not a wasted time to put those hundreds of placeholders. It’s an investment.
Keep in mind: we are creating a game and it happens that we are level designers and yes gameplay must be our main concern but it doesn’t prevent us from artistic, theatrical or hollywood-ish ideas.

3- Temper and uniqueness

Now that we have made some researches we understand the roads and the environment in general, we can compare featured countries between each other’s. Some rallies are really unique like Sweden on the snow or Monte-Carlo on narrow mountain roads, those are easy to design because they have so much temper, because their kind of roads and environments are unique. 
Even for a player fond of real rallying, we are still designing a game and we have to begin making compromises to balance reality with video game. We have to amplify the specificities  of each rally to create a contrast between them, we can’t put too much different ingredients because we want diversity among them. We want the player to feel something different from a rally to another in term of emotion but also in term of challenge.

The global guidelines of the game will be driving the level design. For WRC7 we wanted to have a more realistic approach, to make the player feel what it means to be a rally pilot: dancing on the edge, searching for the limit, pushing beyond reason, exigent, and most of all unforgivable.

Challenge in a track like Finland is all about speed and blind jumping bends, meaning you have two choices: braking on every crest because you can’t see the road behind it, or go flatout and jump at 120 mph in a corner you can’t even see just because your co-driver said so. It’s a unique challenge.
In Corsica you drive on bumpy narrow roads with a lot of turns. You can and want to go very fast because of the tarmac but the bumpy road is unforgivable if you put your tire on the bad spot.
At this point you “only” have to do it all over 13 different rallies. In WRC7 we made the choice with the art team to work first on the more specific rallies first (Finland, Sweden, Deutschland, Mexico…) and then to create the ones which have less gameplay identities (Australia, Sardagna, Spain…) because they can inherit from the previous rallies. For example, Spain is a mixed rally, some part of the stages are on tarmac and other on gravel and mud. One moment you are on a perfectly smooth light grey tarmac going up the hills the next  one you are stuck into third gear between two valleys surrounded by dry vegetation and a combination of sand and mud... So we used what has been done in Germany (smooth tarmac) and what we did in Mexico (technical gravel/sand). Let me be clear: it doesn’t mean we copy/past the level designs, me alive it won’t happen. It means we had the same goals and expectations, the almost-same degree of road deformation, the almost-same roads sizes…

Identifying the temper and the challenge of each rally allowed us to have a clear vision of each country and to set the production priorities. This is very important: the global vision of the level design is very important. Without anticipating it you would be stuck after the 5fth country. You know I love to say stupid things so here another one: When you have to slice the cake you've made you bought, you start to count the number of persons around the table: if there is 8 persons and you make 7 slices, someone is going to be hungry... and the more people there are around the table the more you’ll have to think of how to slice it. And if there is only three? Well go on you are almost certain you can do whatever you want that will works.

The same thing applies for road deformation: some tarmac rallies are smooth while others are very bumpy. We do not try to recreate the real rally saying “Hey look at that, the road in Monte-Carlo is kinda old and sometimes bumpy”. But we do classify Monte-Carlo as an average deformed rally compared to very smooth Germany and the ultra bumpy Corsica... So when we start working on it we know how bad we have to deform the road comparing one rallies to another - instead of what it is in reality. By defining the extreme cases (Spain has the smoothest asphalt you’ll find in WRC7 and Corsica is the worst) you can have a very rational approach of the countries. This is very close to using dichotomy.

This allows you to have countries identities that will naturally lead to your golden rules.

4- Landmarks

We need constraints to be creative and efficient so here are some of WRC7:
Rally must be representative of the country it’s held in 
Number of stages by country
Length of the tracks (in time of gameplay)
Time of production (in man days)

Okay there is nothing original about it, but this is the base of what we can do, what we are “allowed” to do.

In WRC each track has landmarks. 1 to 3, it depends on the track. They can be visual one like a tunnel or a clearing in a forest but they also can be gameplay such as a deadly descent from the summit of a mountain or a big urban area.
Anyway those landmarks have several utilities.

Player's orientation and memory

The first use is to guide the player through the track. Even if there is only one road and he can’t go anywhere else, the player is going from A to B and to help him to remembering the stage we provide landmarks. He knows he will start the race in a forest, a few miles away he’ll cut through a sawmill and then to the road next to a lake before passing the finish line. He’s no longer driving from A to B but always heading from one landmark to the next one, doing so he can pinpoint himself on the layout of the track. “I know there is a touchy bad cambered after the sawmill”, “Oh yes I can go flatout on the lake’s road”. In an FPS or action game you guide the player through the level, and it’s exactly the same thing. It’s mandatory to landmark the tracks for the player to remember, but also to allow him to talk about the game: “Oh yes select the Mexico track, the road in the canyon is awesome” or to the opposite with “The infinite snow descent in Monte Carlo is unforgiven you shouldn’t play it for your first time”.

We are talking about landmarks that have a precise location, but it also can be 2 miles of road on the side of a mountain without any security barrier, or a combo of 5 consecutive jumps. It’s something unique that comes out of the ordinary, enough to be noticed and remembered. The risk with this is to lose credibility because you want to do something that is too noticeable. A specific location landmark is a minor risk for your track’s credibility.
For example we had a problem in Mexico because the tracks are in a desert area where there wasn’t much to use for a landmark. We had to create mountains after the generation of the map by a world machine to have unique situations such as a crest “on the top of the world” or a bridge over a lake between two mountains. Creating those landmarks took us a huge amount of time because they had to be worked a lot and are several miles long. The results are great and they are investments we had to make because the rally didn't have enough obvious landmarks to offer, meaning we had to increase the production line on the other rallies. 

A track shouldn't have too many different sections. In WRC7 we wanted to let the player see, feel and understand the section before passing to another. For example, in Corsica, famous for the sneaky and unforgivable tarmac roads, there is what we call a "goat road": bumpy as hell, barely wider than the player's car and very tortuous, the kind that makes you curse the designers. Those goat roads are not 20 seconds long, because it's too short, there has been for us two uses for them: either it was very short, like 2 seconds of gameplay to link two standard roads, or it was at least 1 minute to let the player understand what the road is about and how he has adapt his driving to stay alive. 

Here is a goat road. As a player you will definitely remember it, and curse the designers. But when you finish the track finally able put the controller down, you realize that you drove in Corsica not in Deutschland, thanks to the original temper and the landmarks. The video also demonstrate the primary/secondary roads discussed just below, helping the player to understand where he is driving.

The level designer's guideline

The second utility is a guideline for the creation of the tracks: basically a track is a start, some landmarks and an end line linked by roads. Meaning we didn’t create the track from the start to the end, but we positioned everything and then we made the roads in between.
First reason is because you can’t have enough ideas to make 8 miles of track. When you’ll arrive to the third mile you’ll think everything look the same, that you’ve repeated the same patterns again and again, giving a monotonous flow without a lot of temper.
On the other hand if you have landmarks you are no longer creating roads to fill gaps between the start and the end, but creating access and exit roads to each landmark.
For example: in order to maximize the effect of entering a clearance in a forest you want to arrive with the road oriented to the center of the clearance: it will create a clean cut in the forest and an explosion of breath and luminosity. So you know you need a straight line before the enter and after the exit of the clearance. As said before, the roads have has meaning, they have been made to link two spots: two villages, two farms, water supplies from the valley to the village in the mountain,... - it helps you understanding the “why” of your roads and giving you more guidelines for the “how to do it”.

Access roads, let's talk a bit about them. Finding the correct shape, the right width with enough bumps and visibility is a challenge but we helped ourselves: primary and secondary roads. Every road isn't the same, it's even more true in rural parts of the countries where rallying takes place. This means you'll have a primary road that is wider, smoother with less turns used as an arteria for heavy traffic and long distance driving. Secondary roads, narrower, bumpier.... are on the other hand alternative roads for local habitants, often they are a dead end. The important between those roads (and you can have a third kind, maybe and fourth and so on) is to make the player feel that he is going somewhere instead of following a mindless track. There is not much things more rewarding than leaving a main road and going down an almost vertical gravel road, only to join again the primary road just like Dom, Brian, Chip, Bo or Luke would have done in their movies. And this happens because the player can understand what he has just done. Those landmarks help creating a consistent level design.

Not a track. A network.

5- RNG in level design

Random Numbers Generator. You know, when you’re lucky and put three critical hits in a row? that’s a lucky RNG.

At Kylotonn we have an game engine based on splines. The principle is simple: we place actor points [dots], and they are linked by splines. Then, the road mesh is created along that spline.
Just like in sculpture (told you, I’m a repressed artist) we start by creating the macro shape of the track with the help of the landmarks: where the straight lines are gonna be, the tricky turns, several hairpins in a row etc… and for this we only need a few spline actor points. But when the general layout is validated we need to enter in the details of the track and to create a lot more of actor points, something like one every 10 meters.
In WRC7 we made the choice to deform the road, to create hostile surfaces to drive onto and to create those roads we needed a lot of actor points on the spline to break the smoothness of a curve, create water dips or on camber roads followed by a jump into a narrow corner opening on a wide straight line. We could use tools to duplicate those points but we avoided to do so because the engine would create and place the points in a perfect way. Cold and mechanical precision when us, as humans, are duplicating the points with an average-ish precision. It creates a natural randomness on the road that we can use and work on. There was hundred of miles of road to built and by duplicating every actor point by hand we created imperfection like a computer couldn’t have. Constraints help being creative; sometimes the duplication will create a natural dip or a very tormented road that we can work on. 

I’m not saying to ship the game with Skynet creating the roads instead of a human being but rather letting it work for you and stimulate your creativity. It’s even more true when you have to make 200+ miles of tracks in a very short amount of time.
We can’t control and think of every inch of a track, so I lean onto RNG to create opportunities and our work is to provocate and seize the right opportunities. The ones that will matter.

6- Changes between WRC6 and WRC7

The 6th opus of WRC was a friendly and easy going rally game: roads were large and smooth just like the side roads ; even the cars dynamics were smooth. That is exactly what we wanted to break in the 7: the too easy part of it and the smoothness of everything.

At first we narrowed the roads. some of them went from 100% to 50%. The side roads were a part of the gameplay and that wasn’t what we wanted: we are the official licence of the World Rally Championship and the only times drivers go on side roads is because they made a mistake or crashed.
By reducing the roads width and prohibiting the player to step outside we created a more realistic and demanding game that makes it worth braking instead of driving flatout onto the road sides. For the record, rally was at first a test of endurance and drivers had to drive thousand of kilometers across several countries with one goal: reach the destination as fast as possible. Well, arriving first was better, but at a time where cars didn't have safety belts, electrical glasses or power steering, reaching the destination alive was a tour de force. Rally is about passing the finish line fighting against the environment and that’s what we wanted to do, not to create an arcade game where you drive at 200mph crashing into every wall until you win.
The goal was to emphasize that one fact: brake and you will lose 1 second, or don’t and take the risk to lose 10 seconds and damaging your car. It’s a simple risk and reward mecanic.

On those two screenshots, you can see the exact same spot between the two games. Yes we re-used some of the content of the previous game, but as you can see it's a different same. The point of view on the road as been improved, the vista on the left as been created and the road is no longer visually mingled with the rocks. Most of all the road surface isn't smooth anymore, it's very lightly broken to provoke the car dynamics. Same spot, same idea, but a different realization.

We changed the car physics. Instead of  having a car that is very smooth to drive we made it aggressive, able to dance on the road with a single push on the joystick. This allowed us to make the level design changes, because the WRC6 cars wouldn't have been efficient on the new tracks. A smooth handling would have been impossible on such tortuous roads. With communication and hundred of car physics iterations later we’ve found the right balance.

One of the other axis of development was to give credibility. Roads have a mean of existence: if there is a corner it goes around something. If there is a bump maybe it’s a root growing under. Basically, instead of designing tracks and then level building them, we created an environment, and built the roads onto it. The mountains do not extend to encompass the road, but the roads sit onto the mountains. And if this road is at 2000 feets In Argentina, then it will be a narrow road but in Poland it is a flat smoothed road along a hill, it can be wider with less dreadful road sides. The danger is the tricky bumpy roads for Argentina when it is the greed for speed of the player in Poland.

Another major change in level design, was to break the pattern of straight line, corner, straight line and so on. By the fact that WRC6 was smooth the curves and corners were too. The only tool we had to create difficulty back then was the corners. But by having a more aggressive and incisive car plus the tormented roads have transform straight lines into a potential danger: potholes, change of trajectory, narrow with deadly side roads, bumps… By the changes made to cars we were able to create more situations, just like a higher difficulty with narrower roads and no side roads did. Placing the car to take the next corner as become a bigger challenge than the corner itself.

The last big point of the manuscript was to get out of the artificial tunnels: we tried in the 6 to keep the player onto the road despite large roads and permissive road sides, and to do so we were outrageously using walls, dense vegetation, fences, security barriers… It was for the artists a big concern because they couldn’t have beautiful vista on the level, or open spaces to breathe. For us as designers it wasn’t the right way to educate the player: we don’t want them to not be able to go outside the road, we want them to be afraid of going outside the road. By the changes about the incisive car, the narrow and tortuous roads and the unforgiving road sides we managed to keep the player on the track without being forced to hold a leash. 
And the artists were ecstatic.
Ecstatic.

7- Bringing up the challenge

If we break down the gameplay of a racing game, arcade or simulation, it’s all about precision and timing and the more you tend to the simulation part, the more the timing becomes important and anticipation starts appearing.

Precision:

For starter, 70% of PC gamers used a pad controller (wheels not included) to play WRC6, a very important statistic when you have to test your precision challenge. To provoke the precision challenge we played on several fields and the first wasn’t even our own: dynamics of the cars. We disabled every single automatic aid that was on the car to allow them to move more, to feel more the road, to stop the car from repositioning itself. We wanted it to feel bumps on the roads: a pothole deviates the car from its course so the player has to use an extra input to keep it on track.
In rally it’s easy: the track is your enemy, it doesn’t want you on its back, and the car is so powerful you can barely drive it. Your job as a player is to stay on it and cross the finish line alive.
It’s a rodeo. 

Now that the car dynamics are more responsive and sensitive, we wanted to create those tortuous roads, potholes, bumps, jumps and everything you wished you never drove onto. We talked about the splines before and why their smoothness wasn’t what we were looking for. Quite the opposite in fact, we wanted to feel the roughness of the gravel under the wheels. Take a look to this onboard camera from rally of Argentina: https://youtu.be/7eCm8eTNcBU?t=14
The road is a nightmare: gravel road, with a lot of rocks, a lot of turns, bordersides slightly collapsed with even bigger rocks... The only solution to reproduce this feeling was to increase our control and precision by adding points on the spline (a lot of them) and deform the entire width of the road. There isn’t two consecutive points aligned. Each of them is a bit more on the left, or below the other, just a few centimeter, because such a deformation at high speed is already a strong bump. And this brings us to the tweak of the difficulty, because a high speed section can’t have as much deformation than a turn. Deformation as to match the roughness compared to others and to the section you are creating. Most of the deformation is invisible to the player, very often the value of the actor points are very small.

Those deformations on the road greatly increase the precision challenge, and the instability of the car, resulting in the decreasing speed, and confidence of the player. During playtests we observed an increase of the player’s inputs: more than twice compared to WRC6.
One important point is that all this deformation has to be subtle: player must not see it, but feel it. Putting 1% camber to the right on your road with a slight dip of 3 inches in the middle can’t be seen by 95% of your players and it’s ungrateful but a bumpy road full of rocks is not about one big jump or a big dip, it’s all about dozen of small bumps and impacts resulting in an overall sensation of instability.

Timing

Since the game is more simulation oriented than arcade oriented, the car does not respond instantly to player’s input, positioning the weight of your car determines the way you’ll take the corner. Two short turns in a raw will force the player to slow down and force him to play with mass transfer. A change of trajectory at a very high speed will either force him to slow down or to take a big risk to wreck his car and to respawn, losing precious time in a section where he should have been at 100 miles per hour.
A light curve or a light turn can be as dangerous than an hairpin, it depends on what there is before and what comes next.

For the player the timing for an hairpin is about 2 things: anticipating the distance he’ll needs to brake and fight it’s own greediness. One of the most ordinary mistake is to brake too late, because beginners and casuals often want to crush the pedal and prefer accelerating than braking. And yes it comes in this order: anticipating and the greed.
But for a quick change of trajectory, it’s the other way around: the greed comes first. If you ask me why I would answer because it’s so cool to make like in the movies, to slalom like in a hollywood movie between two deadly dangers, your car slightly drifting when it comes out with that rush of adrenaline. Also, for the player braking into this change of trajectory while there is a long strait line after, it’s the greed that takes over for most of the players. It's just like a bait.
This greed is more or less strong depending on every player of course, but a “good” driver (in gaming or reality), when in doubt, will always brakes more than accelerate.

Bottom line: It’s not the angle of the curve that determines the difficulty, but the overall situation: speed, width of the road, grip of the road and visibility of the danger.

Visibility

We tend to underestimate the effect of visibility. You can create the best track ever, the best chicane: smooth, sexy in the drift and beautifully decorated if the player can’t see it coming he’ll only crash on it or in the best case will brake too much.
We talked about greed and trajectories, but the visibility will determines the “how much”.
First of all, the player can’t have a trajectory if he can’t see the entire curve. With a road side full of high and dense vegetation, every turn can be an hairpin. Or not. He can adapt neither his speed nor his approach.
It doesn’t mean we have to create a track in a flat desert, it means that visibility is a strong variable to use or not.
El Condor is a famous track in Argentina seen in the video earlier: the visibility, from a pilot point of view is very poor: rocks prevent you from seeing after the turns so you can only rely on your memory and your co-pilot. On the other hand, in Germany intersections are often clear of vegetation and obstacles, meaning you definitely know where you can or can’t put your wheel.
We mostly used the visibility for two purposes: creating vistas and adjusting the difficulty of a rally. A difficult turn can become easy just because you see how it end, because you can anticipate what you’ll have to do and to choose to take a risk by braking late, or cutting inside....

8- In short

If I only had two minutes to tell you about how we’ve done it:

For those who read so far thank you for your time. We are glad here at Kylotonn to share a bit of our experience and feed your thinking.