After a relatively slow winter, I'm pleased to start the new year w/ a bang. Just a month ago, we began a correspondence w/ Mr. Dario Casali, a member of the Valve team who did both Goldrush and Badwater. We drew up a list of questions, and passed them along. http://forums.tf2maps.net/showthread.php?t=4455
As you may have guessed by this point, we received a reply today, and a rather extensive one at that. As a representative of TF2M, the big things I take away are 1. He answered our non-serious questions in good stride. and 2. He specifically asked us if there was anything else we would like. Which means that if you've thought of any new questions, we can sneak them in; http://forums.tf2maps.net/showthread.php?t=4454 Without further ado.
Sorry it took so long to get this done, I took quite a bit of time of recently, what with holidays and heavy snow getting in the way! Let me know what else you’d like.
Dario
Mr. Casali,
Here is what our community has come up with so far. We weren’t sure if this is something we should be sharing on the Steam SDK forum, so right now these inquiries are only from our group.
Your mapping process
The most common thing that people want to know about is your development process. If that’s something you can share, we would like to understand exactly what happens when you are designing a map from the ground up.
I think the first two questions are best answered in the steps outlined below.
Is it possible to simplify this development process into bullet points for general reference?
1) What gameplay type should the map be?
a. Which gameplay type is most popular?
b. Is any gameplay type under-represented?
c. Do we want to create a new gameplay type?
2) Initial map blocking phase. If the gameplay is new, code will be written in this phase.
a. To begin with I build large gray-textured brushes on the 64 grid to build a rough layout of the map.
b. I use models/props_farm/doorframe001a.mdl as a doorway reference and models/props_2fort/window004d.mdl as a window reference.
c. I place doorways as access points, windows as visual portals and use 256 or 320 as the major height variants between ground levels.
d. Once a basic flow is layout is done, I decide where the spawn rooms will be. I usually just have a large brush with models/props_gameplay/door_slide_large_dynamic.mdl on the front where the exit will be.
e. I will place prop_ragdolls at points that I think each class will function within the game. This helps me in several ways. It allows me to get a good sense of scale, and it also allows me to start picturing how the map might be played by each class. I always want each class to have ways to enjoy the map, and using these props at this point in the production process ensures that I don’t leave anyone out.
f. I place prop_dynamics with the model of models/buildables/sentry3.mdl at points that I want to be defence positions. I can visusalise each sentry’s range by giving the prop a fade distance of 1100. When you select the prop, it’ll have a yellow sphere around it giving you a precise picture of how effective the gun will be, and which points it can be out-ranged.
g. I pay particular attention to flow. I like to minimize the number of turns players have to make on their path through the level. The flow through a map should be as simplified and intuitive as it can be. Paths should only be as convoluted as they need to be for vis reasons, or deliberate path elongation for balance reasons.
3) Initial playtesting phase – I like to start collecting playtesting data as soon as the initial blocking phase is complete. The map should have signs, and fully functioning entities before the first playtest. Use as few details as you can in this period, because it allows you to make adjustments to the level very quickly. It also means you can make major adjustments if necessary without throwing out a lot of detail work.
4) Continue to playtest! Once the changes made to the map are getting smaller and smaller, the artist can begin to work on the map. Usually there will be some high-level art concept defined at the beginning of the process, but this is not essential.
5) When the art is done, we playtest continually to find exploits or bugs in the map. Clip brushes are used to smooth catchy edges, and to prevent players from getting to areas they we don’t want them in.
6) Finishing touches are added like cubemaps, soundscapes, HUD messages etc.
7) Rlease!
How many people are usually working on 1 map?
We usually have one level designer and one or two artists working on it.
How closely do you work with the people doing textures and models? And what program do they mainly use for TF2 artwork, and specifically textures?
I work closely with the art guys throughout the process to get the art asset production pipeline working in parallel. Once the assets come online, they are integrated immediately so the artist can have feedback on his work. I believe photoshop is the preferred software for making materials.
What compile settings do you use for your maps? i.e. soft shadows, hdr, etc.
We use VRAD -staticproppolys -staticproplighting –textureshadows –both. Always use VVIS without the –fast option.
What means does Valve utilize to test maps for balance? i.e. How do you figure out the number of seconds to adjust respawn time when a cp is taken?
The way I balanced Badwater Basin was to time the average respawn + travel time of each team to meet at a predetermined point of neutrality. That way I can tell where the threshold of control will be and balance that threshold to guarantee both teams have a somewhat equal chance of winning. One example of this is the final capture point in Badwater. The point of neutrality is just around the corner to the final area. That is the point at which a blue and a red demoman would meet up if they died at the same moment, respawned, and ran toward each other.
Can we see developer-level paintovers of more of your maps, a la the beta Hydro + paintover present in one of your powerpoints?
I’m not familiar with paint overs!
When do you determine a map is finished? Unlike the rest of us, VALVe is able to update maps if an issue occurs. How helpful is this when deciding if a project is done or not?
When a map is free of obvious defects and exploits and has been determined to be fairly balanced, we release it. Upon release we monitor the forums and play the map on public servers and collect a list of post shipping fixes. This usually results in an update of the map a few weeks after the initial release. We don’t like to release too many updates to a map because of the high cost of mandatory downloading TF players have to endure.
General Inquiries
Which of the current maps was most challenging to develop?
It’s always most challenging to develop a map that is a new gameplay type. It takes time to experiment with different ideas as the concept takes shape. It takes a lot of playtesting and a lot of theoretical postulation.
What kind of alternatives were considered to the Alpine environment?
Is it common for level design ideas and layouts to get traded in between different game development teams during the design process? IE: A design idea that turned out not to work so well for Episode 2 having made it's way into Team Fortress or Portal.
This hasn’t directly happened yet to my knowledge. In many cases we take lessons and concepts we have learned from one product and integrate it into another, but this is not often a level design thing. In L4D some aspects of TF were integrated into versus mode – like the 20 seconds respawn time.
Since most displacements are natural cliffs and are relatively simple to understand, can you tell us how you created the satellite dish in Hydro?
This dish was created by building an 8 sided arch with brush edges all at 1:2 grid angles. Once turned into displacements, the whole thing was smoothed using the subdivide command. Because all the edges were on the 1:2 grid, holes could be cut in them where the steps are.
Do you have a favorite game type that you enjoy mapping for or playing?
Personally I love the Payload maps, but then I’m biased since Goldrush and Badwater Basin are the only TF2 map types I’ve made!
How have the ways the community received your maps fulfilled your expectations? How have they surprised you?
I have been very happy with how the maps have been received. There are many Goldrush and Badwater 24/7 servers, like the Q clan servers. I play on those a lot.
What do you think is the coolest idea you've come up with while mapping, and did it end up in the final version?
I really like the concept of the cart that you’ve been pushing through the whole level finally tipping into the big pit and blowing everything up. I think it makes for a very satisfying conclusion! Everything should end with an explosion!
You have shown that you are willing to also include maps with more experimental gametypes like cp_steel in your community map pack. How far would you be willing to take this? Would you include a map with a new prefix and game type if it met your standards for quality?
Absolutely! We are always on the lookout for fun new TF2 maps. If they are new gameplay types, even better!
On behalf of everyone at TF2Maps.net, thank you again for taking time to speak with the community. Everyone here really wants to know how the pros do it, so having some of our questions answered really means a lot us.
As you may have guessed by this point, we received a reply today, and a rather extensive one at that. As a representative of TF2M, the big things I take away are 1. He answered our non-serious questions in good stride. and 2. He specifically asked us if there was anything else we would like. Which means that if you've thought of any new questions, we can sneak them in; http://forums.tf2maps.net/showthread.php?t=4454 Without further ado.
Sorry it took so long to get this done, I took quite a bit of time of recently, what with holidays and heavy snow getting in the way! Let me know what else you’d like.
Dario
Mr. Casali,
Here is what our community has come up with so far. We weren’t sure if this is something we should be sharing on the Steam SDK forum, so right now these inquiries are only from our group.
Your mapping process
The most common thing that people want to know about is your development process. If that’s something you can share, we would like to understand exactly what happens when you are designing a map from the ground up.
I think the first two questions are best answered in the steps outlined below.
Is it possible to simplify this development process into bullet points for general reference?
1) What gameplay type should the map be?
a. Which gameplay type is most popular?
b. Is any gameplay type under-represented?
c. Do we want to create a new gameplay type?
2) Initial map blocking phase. If the gameplay is new, code will be written in this phase.
a. To begin with I build large gray-textured brushes on the 64 grid to build a rough layout of the map.
b. I use models/props_farm/doorframe001a.mdl as a doorway reference and models/props_2fort/window004d.mdl as a window reference.
c. I place doorways as access points, windows as visual portals and use 256 or 320 as the major height variants between ground levels.
d. Once a basic flow is layout is done, I decide where the spawn rooms will be. I usually just have a large brush with models/props_gameplay/door_slide_large_dynamic.mdl on the front where the exit will be.
e. I will place prop_ragdolls at points that I think each class will function within the game. This helps me in several ways. It allows me to get a good sense of scale, and it also allows me to start picturing how the map might be played by each class. I always want each class to have ways to enjoy the map, and using these props at this point in the production process ensures that I don’t leave anyone out.
f. I place prop_dynamics with the model of models/buildables/sentry3.mdl at points that I want to be defence positions. I can visusalise each sentry’s range by giving the prop a fade distance of 1100. When you select the prop, it’ll have a yellow sphere around it giving you a precise picture of how effective the gun will be, and which points it can be out-ranged.
g. I pay particular attention to flow. I like to minimize the number of turns players have to make on their path through the level. The flow through a map should be as simplified and intuitive as it can be. Paths should only be as convoluted as they need to be for vis reasons, or deliberate path elongation for balance reasons.
3) Initial playtesting phase – I like to start collecting playtesting data as soon as the initial blocking phase is complete. The map should have signs, and fully functioning entities before the first playtest. Use as few details as you can in this period, because it allows you to make adjustments to the level very quickly. It also means you can make major adjustments if necessary without throwing out a lot of detail work.
4) Continue to playtest! Once the changes made to the map are getting smaller and smaller, the artist can begin to work on the map. Usually there will be some high-level art concept defined at the beginning of the process, but this is not essential.
5) When the art is done, we playtest continually to find exploits or bugs in the map. Clip brushes are used to smooth catchy edges, and to prevent players from getting to areas they we don’t want them in.
6) Finishing touches are added like cubemaps, soundscapes, HUD messages etc.
7) Rlease!
How many people are usually working on 1 map?
We usually have one level designer and one or two artists working on it.
How closely do you work with the people doing textures and models? And what program do they mainly use for TF2 artwork, and specifically textures?
I work closely with the art guys throughout the process to get the art asset production pipeline working in parallel. Once the assets come online, they are integrated immediately so the artist can have feedback on his work. I believe photoshop is the preferred software for making materials.
What compile settings do you use for your maps? i.e. soft shadows, hdr, etc.
We use VRAD -staticproppolys -staticproplighting –textureshadows –both. Always use VVIS without the –fast option.
What means does Valve utilize to test maps for balance? i.e. How do you figure out the number of seconds to adjust respawn time when a cp is taken?
The way I balanced Badwater Basin was to time the average respawn + travel time of each team to meet at a predetermined point of neutrality. That way I can tell where the threshold of control will be and balance that threshold to guarantee both teams have a somewhat equal chance of winning. One example of this is the final capture point in Badwater. The point of neutrality is just around the corner to the final area. That is the point at which a blue and a red demoman would meet up if they died at the same moment, respawned, and ran toward each other.
Can we see developer-level paintovers of more of your maps, a la the beta Hydro + paintover present in one of your powerpoints?
I’m not familiar with paint overs!
When do you determine a map is finished? Unlike the rest of us, VALVe is able to update maps if an issue occurs. How helpful is this when deciding if a project is done or not?
When a map is free of obvious defects and exploits and has been determined to be fairly balanced, we release it. Upon release we monitor the forums and play the map on public servers and collect a list of post shipping fixes. This usually results in an update of the map a few weeks after the initial release. We don’t like to release too many updates to a map because of the high cost of mandatory downloading TF players have to endure.
General Inquiries
Which of the current maps was most challenging to develop?
It’s always most challenging to develop a map that is a new gameplay type. It takes time to experiment with different ideas as the concept takes shape. It takes a lot of playtesting and a lot of theoretical postulation.
What kind of alternatives were considered to the Alpine environment?
Is it common for level design ideas and layouts to get traded in between different game development teams during the design process? IE: A design idea that turned out not to work so well for Episode 2 having made it's way into Team Fortress or Portal.
This hasn’t directly happened yet to my knowledge. In many cases we take lessons and concepts we have learned from one product and integrate it into another, but this is not often a level design thing. In L4D some aspects of TF were integrated into versus mode – like the 20 seconds respawn time.
Since most displacements are natural cliffs and are relatively simple to understand, can you tell us how you created the satellite dish in Hydro?
This dish was created by building an 8 sided arch with brush edges all at 1:2 grid angles. Once turned into displacements, the whole thing was smoothed using the subdivide command. Because all the edges were on the 1:2 grid, holes could be cut in them where the steps are.
Do you have a favorite game type that you enjoy mapping for or playing?
Personally I love the Payload maps, but then I’m biased since Goldrush and Badwater Basin are the only TF2 map types I’ve made!
How have the ways the community received your maps fulfilled your expectations? How have they surprised you?
I have been very happy with how the maps have been received. There are many Goldrush and Badwater 24/7 servers, like the Q clan servers. I play on those a lot.
What do you think is the coolest idea you've come up with while mapping, and did it end up in the final version?
I really like the concept of the cart that you’ve been pushing through the whole level finally tipping into the big pit and blowing everything up. I think it makes for a very satisfying conclusion! Everything should end with an explosion!
You have shown that you are willing to also include maps with more experimental gametypes like cp_steel in your community map pack. How far would you be willing to take this? Would you include a map with a new prefix and game type if it met your standards for quality?
Absolutely! We are always on the lookout for fun new TF2 maps. If they are new gameplay types, even better!
On behalf of everyone at TF2Maps.net, thank you again for taking time to speak with the community. Everyone here really wants to know how the pros do it, so having some of our questions answered really means a lot us.
