What’s the focus of this picture? Wires. In my time testing maps I’ve seen a lot that have little or no use of wires, and on the flipside I’ve seen maps that have wires here, there and everywhere. The way official maps employ wires is probably closer to the latter, but they don’t go completely over the top. Wires between telegraph poles, from crane wheels, between building roofs, these are all common spots for wires in the real world and should be in your maps too, they’re something that’s very quick to add to your and easy too. The power a simple wire has to make your map look better is amazing; if you’ve not used wires before try them out. Bung in a move_rope, set it some slack (usually between 90 and 120 looks good) and then give it a keyframerope to connect to. Often the width of the wire is fine at the default value but don’t be afraid to vary it a little. If you’re really feeling daring you could change the material it uses to try a different colour or perhaps rope. I’m not sure if there are any non-default wires in official TF2 though, but when did that ever stop anyone?

Got any mine carts in your map? Instead of leaving them empty, put a small displacement in it with a gravel or even gold texture, this is a dead simple way to make each one look different from the last. Plus, you get the added bonus of it looks like someone actually worked in your map; a map full of completely empty mine carts will just feel weird. Of course you can vary the scale of the texture to make it look like it has bigger or smaller rocks in it. If you’re going to place several next to each other like this shot, make the first displacement then turn texture lock off before you copy it. That way the rocks on the displacement look different for each cart even if the displacement is the same shape. This wasn’t done in this shot and you can see that all three mine carts have the same pattern of rock in them. It doesn’t just stop at mine carts, you can add displacement mounds of dirt/rocks/gold anywhere, like at the end of those tracks in the background of this shot.

Nice bands holding the wall’s support beams together. Taken from close up.

This is the same wall but seen from the opposite side of the playing area then zoomed in. The small bands in the previous shot haven’t just shrunk to a size too small to see, they’ve actually disappeared. This is normally a trait associated with models, when they get far enough away they fade out and your PC doesn’t render them any more, actually this can be done by brushes, and even overlays. It’s a really great way of cutting the amount of stuff your PC has to render, you pick small brushes that you can’t really see from a distance and you make them a func_lod instead of func_detail. That way you have complete control over how far away they should fade out and of course once they’ve faded out they are no longer drawn and aren’t a burden on your system. A good little trick for areas that have a few to many details for good FPS.

Each of these areas that are completely sectioned off from the playing area really help to enlarge spaces. Keep them reasonably small and light them well and they’ll really help convince the player he’s in a bigger environment. Inaccessible balconies are great for this too, as long as the ceiling is high enough, adding  a small balcony with a door either side easily improves a bare room. In the bottom shot there’s another displacement pile of gold, just like the mine carts. Oh yeah, don’t for get to add some glow to those lights, env_sprite, point_spotlight or env_lightglow all work well.

This shows same building viewed from upclose and from the other side of the area. How many props can you see in the first that aren’t in the second? I make it about 26ish (probably more than that though) which means from the far side of the space you can see this building from it still looks much the same but you’re rendering about 6 or 7 thousand less polys than if those models hadn’t faded. The way this is achieved is from the fade distances set in the model’s properties. This is a feature vastly underused in custom maps, many have no fades set at all and in Official maps around 80% of props have a fade distance. Naturally some props are just too big or can be seen too far for a fade distance but the vast majority of props are better with one set.

Ahh yes, the displacement roof, great for abandoned or just aged facilities, not a technique to abuse for any freshly built kind of environment though. Make some of your roofs into displacements, power 2 is enough – then warp them slightly like this one. You can only really see the effect when you’re still but when you’re playing its a small thing that really helps make seeing roof after roof after roof less monotonous. Because displacements are done as a batch process this isn’t as intensive as you might think, whilst I can’t give any really figures on if its better or not it really isn’t the sort of thing you need to worry about when optimising. Displacements are also not cut by the tools/skybox texture so you can safely divide an area across a displacement roof and know that the whole displacement will be visible on both sides of the skybox brush, great for tall buildings.

More coming soon.

Once more into the breach, dear friends

This first shot shows off the 3D skybox being used, notice how as the mountains get further into the distance they become more and more blue? The fog colour that has been used matches the rough colour at the base of the 2D skybox behind it. This holds true for any map, in any game. The fog colour you pick should be roughly the same as the colour of the skybox texture near the horizon. Think about it, when it’s foggy, what colour does everything get as its further away? What colour is the sky, at the horizon? If you answered both those with the same colour then you are correct and understand the principle. Pat yourself on the back.

Here, some simple ways you can block players from leaving the play area. Whilst it is clear that a demoman or soldier could jump over those fences they can’t. And nobody questions it. On Hydro there is a fence that is only knee high and nobody questions it. All that is needed is a visual barrier that divides the playing area (visually interesting) from the out-of-bounds area. The out-of-bounds area is usually quite bland but generally it has just enough key details in it to stop it looking out of place, for example that dirt track, keeps going on through those gates, as does the train and track. That billboard in the background as well. Little things like this keep continuity even though there is an immediate fall off to visually sparse.

Some simple details here. A few tires sunk into the ground along with some foliage, rocks and an overlay on the ground under the plant as well. Most of the ground in Badlands is pretty free from props and they tend to be clumped like these. The tyres can acceptably be sunk into the ground in this near-abandoned desert environment, you wouldn’t want to sink tyres into grass on an alpine map though, soil tends to be a lot firmer than sand. The player will hardly notice these because they don’t have to worry about them. They can run straight over them and are never going to get snagged on them whilst running backwards away from an enemy. It’s always important when putting low props on the ground to make sure that players can run over and around them easily without even thinking about it.

This perfectly illustrates how you should never be afraid to tell the player where to go. You would think that three massive signs all pointing the same way would be too many, but no, it works just fine and players hardly even notice them. Whilst a player is busy taking in lush environments/killing waves of enemies/fleeing from waves of enemies, their subconscious will see the signs and gently push them in the right direction without them really being aware of it.This shot also shows a bit more of how small details are continued in the out-of-bounds areas. And although this area is only blocked off by chainlink fences, don’t be tempted to only use chainlink fences in your map. Don’t forget to vary things a little.

Finally (for now) this image shows where detail is needed. Detail is needed most wherever the player is. That means around doors, around windows they might shoot from, any room with intel/CPs in, spawn rooms… You want to be putting details like junction boxes, clocks, signs, boxes etc where the player will be most often and props clumped around doors are a great example of this.

That’s all for now. Look out for the others coming soon.

This is why the TF2 world is static, with the exception of dynamic gameplay elements. There are only a couple animated props in the game used for detailing, and are not large or upfront in their presence. Smoke trail particle effects do not stand out, but silently add to the environment. Environment objects which do move or change are either direct gameplay elements (capture points, intelligence briefcases, payload carts, dynamic signs) or are environment hazards which affect gameplay (trains, saws) and have loud, clear sound effects to announce their presence. Any other dynamic elements are players, engineer buildings, or projectiles: things pertinent to playing the game.

However, detailing for TF2 does not just end at “keep the world static”. Detail is carefully distributed and allocated in the world for certain reasons, scaling the amount of it to where it is located. I’m going to call this the density of detail. This concept of density of detail can be found in any of Valve’s maps, but to demonstrate, I’m going to use a particular scene from Dustbowl that I think truly embodies it.

Dustbowl stage 3, capture point 1

I’ll break it down into multiple parts. First is the obvious density of detail. What you should notice immediately is that your view is frame by two big, blank walls, with a relatively bland ground, and a blank sky. Already, you can see how detail is centering your view on the capture point, and that is what is key when detailing: drawing the player’s focus to what is important. The big metal wall on the right is not important. Community mappers have the tendency of seeing a big, blank wall like that in their map, and despite it being irrelevant to the player, adding as much detail as possible, using window models, overlays, conduits, and whatever else comes to mind in an effort to make that big blank wall not be a big blank wall, when it actually should be.

The walls have a low density of detail, and then we move on to the capture point area, which has a high density of detail. There’s obviously much more detail on the capture point than on the less important wall. There are doors, props, bright orange cones, and a bright red sign in addition to the capture point hologram, all which bring your attention to the area, which is where you want players to be looking (not distracted by unnecessary detail on the big wall to their right), as well as where you want players to go. The capture point here is not in the middle of the path players take, so they are not forced to walk onto it. But using smart detailing, you can bring players’ attention to it, since it is a high detail area in a low detail environment.

A second point to make here is how density of detail can build story for a map. An area with a high density of detail will look more important than something with less detail, which is in part what leads to attracting player attention. For a capture point, this can answer the question of why a team wants to capture it. If the entire map has a high density of detail, the question which arises is what is so important about the capture point at hand. Using the image from Dustbowl, if the walls on the right and left were as heavily detailed as the building by the capture point, it would make the smaller capture point building look unimportant, deconstructing the story behind it. However, Valve detailed the map wisely, making the capture point building more important than its surroundings. It is loading dock with a garage, with various supplies and boxes scattered around and inside it, with a tall, important looking structure built on top of it. It fills in the reason why the BLU team would want to capture it for itself: it’s an obviously important or strategic building to have. A loading dock with a garage full of supplies is far more important to capture than that big, blank, generic looking industrial building to the right! In addition, there’s a gate to the left of it, so controlling it means controlling traffic into the compound. A large story for the gameplay is built into the map through detail, and it’s all subtle, nothing is forced onto the player. You just need to look at the scene (and have your attention focused by the properly allocated detailing) to understand what is going on.

Now that we’ve covered the higher end of the density of detailing, I want to take a look at the lower end up the scale.

Above Dustbowl stage 2, capture point 2

This screenshot (again of Dustbowl) is taken of a building high on a cliff face, far away from the player. What you should recognize is two things: first, it’s a very simple building can be made of 4 brushes, and second, proper use of textures that simulate depth on a flat surface. The two go together, you can create very simple structures in areas far from the player’s view, and then use textures like the ones shown to give it the appearance of being more complex than it truly is. Often the wood textures with holes are abused by community mappers by being used on surfaces close to the player’s eye, where the fact that they are faking holes is apparent. The wood texture with holes and the vent overlay shown are best used on structures far enough away to not attract the player’s attention, where their looks are not going to be studied and their attempt to look like they have additional depth is debunked. For buildings close to the player’s eye, a wood texture without holes and a vent model would be used instead. Additional detail, such as signs and ground-level props can be used to add more interest to the building, or, actual holes in the building can be made.

It's full of holes!

Summary:

• Place detail where you want players to look
• Do not feel required to make every surface a point of interest
• Detail, if used properly, can bring story to a map
• Only use textures to fake effects far enough away from the player

Like our articles? Join the Nodraw.net Steam Group to instantly be notified of new content!

Height Variation

If there is a single most important factor in determining whether your map is fun and interesting to play, or plain and boring, it is height variation. Changes it height make the map much more entertaining, offers strategic options for various classes, and quickly makes the map layout much more intriguing to a player. Height variation can also serve a function purpose, such as funneling spam, breaking sight lines, or improving optimization.

In Badlands, height variation is a huge part of the map design. A quick look at the mid point reveals a massive number of planes of combat, horizontal levels on which players can fight, as well as slopes and stairs to obliterate the “flat” feeling of a map. It is impossible to look at Badlands and say it is flat. Also notice the use of the bridge to layer levels of play on top of each other; height variation does not have to be merely side-by-side.

Notice how many tiers of gameplay there are

In King of the Hill maps, height variation allows for spam to be funneled away from the capture point, as well as highlighting the point itself. If you look at Viaduct, the point exists on top of a hill (it is King of the Hill after all), with slopes to allow spam to roll off and away from the point. It also breaks the line of sight between the opposing routes from spawn. In Nucleus, the point exists suspended over a massive pit, which fills the same role of eating the spam.

This pit means a death to not just players, but also their spam

Scout Routes

The scout is a class built around rapid, fluid navigation. The scout’s speed and double-jump allow him to go places in ways other classes cannot. This is recognized in many official maps, which contain certain routes that a skilled scout player can use to navigate around or above the battle, staying clear of most spam.

Examples are abundant in competitive maps:

These hoodoos offer an alternate route for scouts

The wooden shack offers a route to the capture point tower

Skill Jumps

These would be the all-class counterpart of the “Scout Route”. Some of the scout jumps can also be used (albeit less effectively) by all classes, such as the jumps to the spire on Badlands. But other affordances are made to offer special tricks skilled or knowledgeable players can use to better or more quickly navigate the map.

One of the most famous “skill jumps” dates back to the early days of the game, in the second stage of Dustbowl. Using the sawhorse, a player can jump up to the ledge above. In Furnace Creek, we used simple props to introduce these skill jumps into the map, offering alternate routes to ledges above that avoid having to walk to a ramp or slope farther away:

A simple prop is also a sneaky path

A subset of the skill jump pertains specifically to rocket and sticky jumps. It is important to realize what value these two features are to gameplay, and how your map should be built to accommodate their existence.

Walls, slopes, and roofs are all “launch pads” for a rocket jump in addition to the floor. Rocket jumps and sticky jumps are used for quick travel for slower classes, heightening the importance of your map to be friendly to these jumps, especially with the introduction of the Gunboats. A quick search on YouTube will reveal plenty of videos demonstrating what these jumps can achieve.

Reasonable Pickup Placement

Health kit and ammo pack placement seems like a simple enough task, but not putting thought into it can have a serious effect on your map. There are a few simple concepts and ideas to keep in mind when placing health and ammunition.

Place ammo where engineers will build. A map has certain sentry spots where engineers will build regardless of the availability of ammunition, however, their opinion of the map, and the fun they have playing it, will rapidly deteriorate when they find there is no metal to use for building and upgrading. This means ammo pack placement needs to be effective and based on map conditions.

In attack/defend maps like Badwater Basin, Gold Rush, or Dustbowl, you will find full ammo packs at the start of the level, necessary for the RED team to build during setup time.

Two full ammo kits located at the start of the map

Health and ammo is less abundant near spawns, as they are unnecessary when a resupply locker is nearby, and can make it easy for the attacking team to spawn camp.

The closer to an area of heavy gameplay, the smaller kits will be. The middle point of Badlands has ammo at its edges, but they are small ammo kits (not full!), while there is medium health and small ammo in the “house” farther away, and off to the side of gameplay.

The bridge sees a lot of action, so the packs are small

Finally, do not place kits in corners or areas where engineers are likely to build: building on top of kits is terrible for everyone involved. Notice how in Badwater Basin the health and ammo is placed away from the optimal locations for engineer buildings:

Engineers build to the left, and the ammo is to the right

It All Adds Up

By keeping these points in mind while constructing your next map, you will add that “something” that is missing. By adding that “something”, you turn your map into a map that is quality and fun, and plays well in TF2.

And when you take all of it to heart, you end up with Badlands. Everywhere you look, you see what I have mentioned above. From the middle point to the final point, you will notice height variation, “scout routes”, skill jumps, and proper pickup placement – which didn’t play small part in winning over the competitive TF2 community. Watch a match on Badlands and see the level of skill which a player can achieve from knowing and practicing the map.

It has depth. Valve maps have a learning curve. They go beyond being a container for the battle, to being something you can learn the ins and outs of. You can be a good player on a good map if you know it well enough.

And it’s not flat.

How many of the things we discussed can you spot?

Did you get what you needed?

This type of gate prevents a player from leaving an area without a key item that they’re going to need. If you’ve just given the player a super gun, or a key card he’s going to need later, you’ll want to make sure he’s got it with him before letting him continue. As with every type of gate there are many ways to ensure the player has what is needed before letting him continue, the most obvious being: getting him to use what he needs in order to get out.

These commonly happen when the player gets a new type of weapon, in the first moments of Episode One you’re stuck in a pit until you get hold of the gravity gun and you can only get out once you figure out (or remember) how to use it. Thus ensuring all players can only progress once a basic mechanic has been learnt. Later on in EP1, when you get your first pistol there is a chain link door blocking your way that has been padlocked shut. A player who failed to see the pistol will try in vain to use the gravgun to blast the padlock open and find it impossible and be forced to search for an alternative. The player who picked up the pistol though will realise quickly that it needs to be shot open. Simple, but 100% successful at ensuring the player has a pistol.

Pistol? Check. No player will ever leave this room if he can't find the pistol. A dead zombie slumped in the corner provides a reminder of how important finding the pistol is.

Hey, you OK?

The emotional/story gate is next up on the importance list. All gates provide a bit of breathing space but this type is there specifically to add much more of a pause in action, typically a door or obstacle that only an NPC character like Alyx or Barney. Just after the stalker car train crash in Episode One, both you and Alyx scramble into a small room, the only exit is blocked by a combine force field. Alyx slumps against a wall and delivers a minute or two of emotional catch up along with a bit of a story reminder just in case you got a bit carried away and forgot that the citadel is about to explode whilst you’re still under it. Both important features of immersion and keeping the player fully absorbed into the universe you’re creating around them.

Alyx's actions constantly remind the player what is going on, She's often a pillar of courage but sometimes even Alyx needs time to come to terms with what is happening. Yes, that was a train crash you were just in, and yes those are people post-mutilation.

Sight seeing

Sometimes you just need to stop and admire the view. Some times you need to beat players with a stick to get them to admire the view you’ve created for them. Fortunately though its often not that difficult. You’re inside the Citadel, everything is falling apart and you’re in a rush to get to the core. You turn and suddenly a huge expanse opens in front of you, do you look? No, of course you don’t, the Citadel is about to explode. But a knee high barrier you can easily jump over is enough of a visual cue to stop you from instantly seeing the way forward, then combine that with Alyx’s ‘Whoa’ and all of a sudden you’re willing to spend a few minutes to look around.

One of Episode Two’s best vistas is blocked off in an even more feeble manner, the Combine infested bridge being your first view of daylight after the antlion tunnels would be an obvious point to stop and admire the country side. In the commentaries though it is reported that players managed to miss the scene between Alyx and the Vortigaunt that unfolded. By placing a few barrels blocking the exit players paused long enough for the scene to start and for them to realise they should stay and listen.

Three barrels is sometimes all it takes.

The tutorial

The first time you encounter something new there will probably be a gate making sure you’ve actually witnessed the introduction. Sometimes, like this example, the introduction actually is the gate. Here the player is introduced to a new enemy in a safe way. It manages to pause between to chunks of action, provide comic relief as well as teaching the player about the zombines that become a key enemy of Episodes One and Two. Tutorial gates can teach the player anything, sometimes an NPC will allow you to leave once you’ve completed X task at least once, like Dr Magnuson when he’s teaching you about how to use his Magnuson devices on striders. Other times the gate opens itself like the zombine gate or the gate might be one you have to open yourself.

A Multi-purpose gate in action, providing a bit of comic relief 'Zombines, get it?', teaching players zombines grab grenades when close as well as providing a breather in between action.

Combining gates

Most gates aren’t one type of gate combining purposes can make the gate far more interesting and help to engage players with NPC characters that they might just be ignoring otherwise. All of the gates I’ve mentioned here are combinations of the key fundamental aims of a gate: Pause in action, lesson, got everything? and  emotional/story developments. They key to a good gate is for the player to not even realise they’re being halted deliberately. With pacing that flows and no abrupt changes Episode One alone sneakily managed to fit in around 35 to 40 gates!

The Soldier

Of course, it might seem that which class is most important is an incredibly debateable topic. After all, each class has quite a devoted following, and affects the game in many significant ways. To single one out as most important might seem a bit reductionist. However, one of the key aspects that sets Team Fortress apart from other games revolves around the absurd movement physics. A game like Counterstrike or Call of Duty, while interested in different heights and the way in which players combat one another in three dimensions, is always stuck with the basic movement rules of the real world. With the emergence of the rocket jump in the original Quake, we see a whole new way of traversing terrain, and a variety of level design opportunities (as well as restrictions) emerge from this new formulation. Team Fortress, emerging from a cobbled together set of Quake mechanics (as well as some new ones) is subject to some of the same absurd treatments of space. And, as the designers are keen to note, one of the things that is most compelling about the TF2 art style is the way in which it enables completely unrealistic height differences. They are, after all, making three dimensional levels, while the common sense world we live in is largely focused on only two.  The class which most frequently and skilfully utilizes this unreal height based geometry is the soldier, and so considerations of height should be at the fore front of your map design.

The soldier is clearly a class which emphasizes height advantage. Its primary weapon offers both the opportunity to acquire said advantage (rocket jumping), but also the fundamental way in which to utilize said advantage (splash damage). We’ll start with some rather basic observations about splash damage, and how level design can provide advantages or weaknesses for players in different positions.

To start with, consider how soldiers treat one another while meeting on perfectly flat terrain. First, because the rocket has travel time both soldiers want to dodge the fire of the other, while at the same time predicting the potential location of the opponent. However, the rockets splash damage and physics bounce add an additional complication. Rather than aiming directly at the opponent, you want to aim at their feet, as splash damage is much easier to hit with than a direct shot. This will also bounce the opponent into the air, potentially throwing off their aim and giving you the opportunity to predict their landing.

Now let’s give one soldier a bit of height, say a small incline or a hill. The soldier on top of the hill still wants to fire at the ground, but it has become a great deal easier, as the amount of ground to target has increased.  Inversely, the opposing soldier has almost no ground to target, and must settle either for a minimal amount of splash at the lip of the hill, or a direct hit, which is even more difficult now given the lip of the hill can block significant portions of the dominant soldier’s body.

The third example pushes the second scenario to the extreme, by positioning one of the soldiers on top of an almost inaccessible height (say, the building at B).  The soldier on the top of the roof has a significant advantage: he only has to step back a tiny amount to be fully concealed, thus preventing all damage. Likewise, a single step forward and he can target the soldier below completely, maximizing splash damage and ensuring safety from reprisal. The soldier on the ground, on the other hand, has little chance of landing a hit, splash or no. The only recourse in this scenario is retreat or a rocket jump to nullify the height.

The most obvious way to implement this sort of gameplay into your map is rather straightforward: if you want to offer an attacking or defending team an advantage for their soldiers, provide them with height to fire down upon the opposing team. This picture from pl_frontier offer an example of switching this advantage from defender to attacker, and how to layer several distinct height sections to provide counters to opposing heights.

However, you shouldn’t think that height is necessarily a question of putting a large structure, set of stairs or tall prop. Even the slightest difference in height can disrupt aim and provide a benefit. Examine the final point of cp_coldfront, where the stairs leading the main combat space net the attacking team a significant advantage, or the slight displacement alterations throughout cp_gravelpit, providing minor adjustments to potential combat.

So we know that height offers soldiers a significant advantage. Is there any way to offset this advantage? The most direct manner to ensure balance between high/low positions is health and ammo. A soldier is not an autonomous force: rocket jumps burn up health, spam burns up ammo, and even a slight fall on a wounded soldier can add up over time. As such, removing health from a high area ensures that it isn’t overpowered (see Valve’s director commentary on Gravelpit’s B roof).

Another way to ensure a fair advantage is to artificially limit the view from above. While the completely exposed roof at gravelpit B is significant, you should note that the small windowed structure in the middle often blocks line of sight, forcing anyone on the roof to expose themselves to get a correct bead on incoming enemies. If the structure was removed, a soldier could pop up and over the lip continuously with little disadvantage, covering almost all of the entrances into B.

A final tactic is to build in splash damage potential from below. As we have seen, soldiers firing up rarely have a chance to fully utilize their splash damage. The target above is protected by the geometry below them. But, if we add a wall or a prop near the height advantage, the soldier below can bounce their rocket damage off the obstacle, making it relatively dangerous for a soldier to take full advantage of the position offered by the height. This can be seen in my map cp_frost, where I removed a rather extraneous hallway, and put a glass wall in its place. The overlooking window provides an interesting sniper/soldier spam position, but allows soldiers lower down to spam back, preventing the position from completely dominating the combat area.

Well, you can find out exactly how much each of these contributes to your map’s size using several programs, I like to use GCFScape for this though, you can find that here

Just open up any .BSP file (either from your maps folder or from one of the valve GCF files located in your steamapps folder)  and navigate to the lumps folder, sort by size and take a look at the biggest lumps

Commonly large lumps (>1MB):

So what do all of these lumps actually contain?

Lump 40 is commonly large for custom maps, the pakfile, this lump contains all of the custom stuff people put into their maps. New textures, models, sounds. This lump can easily hit 50MB and is single handedly the largest contributor to large .bsp file sizes. If you’re aiming for streamlined downloading it will be worth looking at what elements you really need, what you can optimize and what can be done in a different manner (Texture re-colours for example can be done to a limited degree via .vmt proxies)

Lump 8 and 53 contain the lightmap data for every brush face in the map. Compiling only LDR will eliminate lump 53 entirely, often saving ~10MB; however this should only be considered in a dire pinch. The 8 and 53 lumps are often exactly the same size, they’ll start to differ once you start making lights different for HDR.

The key thing to remember with lighting is luxel size, the bigger the luxels on any particular face, the less information stored about it and the smaller the lump. Faces that don’t have any strong change in light should have their lightmap scale increased to reduce the number of luxels. Ceilings, roofs, walls entirely in shade, all are good candidates for lightmap optimization, but that’s for another tutorial.

Lump 34 is very similar to 8/53 in nature: it stores the lighting for displacements, it seems to be much smaller than the normal lighting lumps though. Size is probably tightly proportional to the number of displacement vertices.

Lump 29 contains all the data about your solid brushes and models. Depending on the number of props, displacements and brushes you have, your physics lump can be several MB. Reducing the size of this lump just comes down to making more props non-solid. A tricky one to optimize.

Lumps 7 and 58 contain all the info about the faces after vbsp.exe has spliced them up. There are two versions because of the lighting information stored, all lightmap data is in lumps 8/53 but there is some extra data stored in 7/53. These two shouldn’t get over a few MB at most.

Lump 35 holds game specific data, lists of all the prop_static entities are in here too, they’re not in the entity lump (lump 0) Less props used, smaller game lump. Not often a large lump though, certainly not one to actively try and cut down.

Lastly the original faces lump, 27, can be pretty hefty. By pretty hefty I mean ~1MB. This one contains all the faces of brushes as you see them more like in hammer instead of ingame. It still has all the void-touching faces culled and probably doesn’t contain skip faces either. If this lump is very large you may have some inefficient brushwork going on inside hammer, your faces and faces HDR lumps (7/58) are also going to be larger.

In conclusion:

The lumps you’ll want to concentrate on when optimizing are the pakfile and lighting lumps, cutting out custom content, increasing lightmap scales and removing dynamic lights. The displacement lighting lump is next on the list but generally quite a bit smaller than the lighting lumps; Lower scale displacements, higher lightmap scales.

If you’re interested in the other lumps and more detail, please refer to the Valve Developer Wiki which has a list of all the lumps and more detailed info on the format each.

In this post I’m going to distinguish between two different methods of game design, and I’ll extend those methods to source level design with some examples. The first method is Architectural design (referencing Jonthan Blow’s commentary on narrative in games). This style relies upon a very specific starting point for the design process, perhaps a blueprint whose basic requirements are all but immutable (like a previous level that is being translated from one game to another), or a representational or narrative requirement (a theme or setting that one is trying to capture in your design). Think of a game like HL2: it seems quite likely that the story told via the sucession of levels was arhictected in advance of the level designers work (somewhat in line with the programmers getting the engine up to speed). As such, though the designers had a degree of creative capacity, they are often times composing spaces in reference to the prebaked narrative.

What is emphasized here is that the level is built with regards to the goals or needs of the other object (the previous level, the representation).  The level is built to fit something else, to match up with another creation.

The second style of design is what might be called Emergent design. Emergent design tends to start without a clear goal or idea, no real absolute towards which it must inevitably tend, but rather it allows the experience of designing the level to dictate the formulation of the space. Accidents, intuitions and chance feature heavily in this style of design: one isn’t so much designing a level, as exploring the possibilities of the level as one goes, allowing you think of spilt paint as a goal rather than a hindrance.

An analogy of these two forms which might prove useful is to consider a level as a ball of clay. The first style of design operates by proposing what you want to turn the clay into (a cup, a bowl, etc) and then matching the clay to the proposal. The second style of design offers no proposed target; rather, the artist simply molds the clay into interesting shapes until something happens to catch their eye, thus allowing them to form a coherent level.

Often these styles can be interwoven. Consider the developer’s commentary from HL2: EP2. In one section of the ant lion caverns, a brief story is relayed in which a valve designer randomly opened a wall in order to generate a new space for items. Upon doing so, they realized that the opening gave the player a perfect view of the upcoming path to the Ant Lion Thumper. This vista satisfied some of the general design goals, such as allowing players the opportunity to observe before engaging. It also gives the player feedback along the lines of “come here” or “you’re going the right way”, avoiding unnecessary backtracking or confusion. However, it wasn’t planned out or preconceived; rather, it emerged purely by accident, and just so happened to elegantly line up with Valve’s own design philosophy.

With a game like TF2, we can clearly see how random or unplanned geometry can quickly lead to novel and interesting gameplay spaces. Imagine accidentally creating a small ledge by not properly player clipping a space, only to realize that said ledge gives defenders/attackers a much needed and enticing vantage point. Or, imagine composing a small detail room, separated from the rest of the map where detail work doesn’t hinder play, only to discover that said room perfectly connects two disparate parts of the map together. Likewise, a rock, thrown at random in the middle of a room might just as easily become a rocket (such as with my recent detail contest entry), allowing you to cement the visual nature of a space together and to expand the dimensions of the area in new directions.

It seems clear that at some level one is always architecting spaces together, ensuring enough overall consistency to complete the project. But one should also be sure to allow the very act of the creation enough room to change the blueprint. The very act of making a level should put the blueprint into question, allowing you to bend it as you go.

In almost all TF2 maps, the pickups are located above colored paint patches. Here are a couple of ways you can place the patches:

However, the color of the patches vary heavily between the official TF2 maps. I will show you a table of the results of my research, right after I’ve explained what playerclipped stairs are!

Playerclipping stairs is something that most mappers do. It means that you create a flat brush, covering every step in stairways, all the way from the top to the bottom.
Here is a comparison between a stairway without clipping and a playerclipped stairway:

Now that you know what I’m talking about, I’ll give you the table:

This is what I gathered from this research:

• On Attack/Defend maps, a single color is always used, usually brown or red
• While the ctf maps, ravine and viaduct all use patch colors based on which side of the map they are on, none of the 5cp push maps use this system
• Unified patches are used heavily by Valve
• Playerclipped stairs seems to be the standard, although some mappers at Valve doesn’t seem to use them at all
• The only set of stairs which are not playerclipped in dustbowl, is the ones at Stage3, Control point 1, that was added after the initial release of dustbowl.

Some notes about playerclipping stairs:

There are a couple of reasons why you should playerclip your stairways. The first reason is that stairways without clipping, makes your view vibrate as you walk on them and the player doesn’t move smoothly.
The second reason is that clipped stairways are fully jumpable. When you jump going upwards in a stairway without playerclips, there’s a pretty big chance that you’ll hit the edge of a step, causing your jump to send you straight up, instead of forward up the stairs.

Some notes regarding paint patches:

Paint patches are very useful for identifying the location at which an item is waiting to be spawned. When you are new to a map, this helps you learn the locations of them easier, as well as giving you a visual aid for it when your health is running low.

A few personal notes:

I never use unified paint patches, no matter how close the pickups are to eachother. I believe it’s important to be aware of how many pickups are about to spawn at the patch you are looking at.
Like I said in the paint patch notes, it’s about making the items identifiable and easily found. This is the reason why I also use a different patch color system than any of Valve’s maps.
I place blue patches under my heathpacks and red patches under the ammo packs. Hopefully players will realize this after having played any of my maps, so that when they are on fire waiting on a medpack to spawn, they know that they should stand on the blue patch, and not the red.

I hope you found this little article useful in one way or another!

• IIKKA: The cylinders trick is a good one. Basically, you start with cylinders that are built along one axis and then just rotate the brushes and vertex-edit to get back on the grid. The most laborious part is the joint between a straight and a 45-degree pipe; when you clip the brushes in an angle where they meet, the displacements won’t match. You need to first sew them and then tweak with one-unit steps until it looks good enough. The down-slope is done easily by lowering the bottom vertices; that portion of the pipe is just skewed rather than rotated. Texture alignment is key in making it look good.

2. NODRAW: You have these great one-way windows in the spawns; while this may not be the case here, is there any way to tell VVIS to calculate only out of a portal and not into it? So that while the spawn door is closed, players outside would not render in? I can’t think of any areaportal jiggery to do so…

• IIKKA: Unfortunately area portals are always two-way. This would require new technology.

3. NODRAW: Is there anyway the clips can be taken off the middle two towers’ rooves? These always seem jumpable as the battlements are nearly the same height.

• IIKKA: The towers have differing geometry that gave the BLU team an advantage before their tops were clipped. They were built rather late in the development to replace a different obstacle that you couldn’t jump on (a tall tree on the edge of the cliff) and the clip brush was deemed the best solution to the problem.

4. NODRAW: When working at odd angles, how do you keep track of it all? E.G. the X shaped supports near spawns that are angled on both Z and X axes.

• IIKKA: I’ve built a lot of stuff in odd angles (since dod_avalanche) so I’m just used to it. But there’s a few tips I can give you to get you started.
• Building in a 45-degree angle is easiest when you assume that 96 units diagonal equals 128 units along an axis. This not only lets you stay on a reasonably large grid, but the miter joints between the angled and axis-aligned architecture will be in a simple 2:1 slope. For example when a 16-unit thick straight wall meets an angled one, you just need to move the corner vertices by eight units (or clip in this angle).
• In general, when building in an angle you should establish a large “angled grid”, like an equivalent of 128×128 squares. As an aid, you could make a checkerboard of diagonal squares out of “Skip” material and place it below your level so you can use it to help you align stuff in top-down view (I sometimes do this on graph paper). Start large, and subdivide the grid later when you’re building smaller details. This is useful when building axis-aligned things as well.
• When designing a sloped roof in an angled portion, again stay on as large a grid as you can. Let’s say you want a roof like in the room leading to the battlements in Doublecross. Start by making a slab aligned to your diagonal grid. Then raise the vertices on one side by 64 units for each 128-unit-equivalent grid square that the roof spans. This results in a fairly standard slope and makes it easy to fit beams and such to it.

5. NODRAW: Was this pre, post, or during meet the spy? I.E. the reference made by the spytech warning board.

• IIKKA:It’s never as simple as that. The earliest files I have with elements that lead to this map (the bridge) are from August 2007; but the bulk of the development as a CTF level has been since the summer of 2009 ie. after the Meet the Spy video… It wasn’t really built specifically to house the announcer board, it just fit in perfectly and is appropriate for a CTF map.

DEVELOPMENT HISTORY: DOUBLECROSS

IIKKA:

Back in 2007 when TF2 shipped, I started to develop a capture point concept level with a night time setting, called cp_nightfall. Night levels hadn’t been done in TF2 before and there was some interest in developing this as a new artistic theme to be released in a later update.

I abandoned it a few weeks later and began to work on cp_badlands. Both levels had a similar layout and theme besides the lighting, and we had already invested time and effort into artwork for Badlands so it was an obvious choice. Later I moved over to Left4Dead so Nightfall was left sitting on the shelf. Note the double-deck bridge with railroad tracks on the bottom level (based on Eads Bridge in St. Louis MO, photos below)  and the triple spotlights lighting the BLU logo.

As Left4dead neared completion and I had some idle time between playtests, I began to think about CTF levels. Earlier there had been a tendency to only think of new TF2 levels in terms of capture point gameplay, but it soon became clear that there was an audience for CTF as well as 2fort remained the most popular level. I talked with other level designers to find out what they thought were the strong and weak points of 2fort, and spent some time playing and analyzing ctf_2fort. This scan has some of my notes from this time, measuring various travel distances.

I took Nightfall out of the drawer and started to think of how to use it as a base for a CTF level. Initially I took the flow times from 2fort and placed path intersections at roughly the same distances, although the spaces themselves were dissimilar (e.g. point B on the street outside the base corresponds to the inside of the entryway in 2fort because the paths from the top and bottom of the bridge meet there). I considered a couple of different layouts with the double-deck bridge until I finally came up with the bridges crossing at different heights, which seemed to allow a dramatic vertical feel to the center area and a compact instant-action layout.

I sketched the silhouettes of the two bases with different colored pencils on top of an elevation drawing of what I had built as the base façade. It looks like a mess here, but the final results look pretty much as I imagined.

The inside of the base went through multiple layout designs even before the first playtest. In particular the main path from the front door to the flag changed radically, from a TFC style entrance that splits in two and reconnects, to a “roadway” through the battlement room, to a series of interior rooms and finally to the open back courtyard that exists now. After the last sketch, there were still more changes based on playtest experience: removal of front stairs to spawn floor, making the back entrance to the flag a one-way drop, adding an underground tunnel from the courtyard shed to the sewer, and changes to the architecture between the spawn and the stairwell.

Thanks so much Iikka! Let’s all do our best to put some of these tips into practice.

Sidenote: Compare Iikka’s paintover of the two respective buildings to these doublecross screens taken at identical angles.

Quite accurate! (Via and thanks to Acegikmo)