Why building the game is only the beginning

With most of the game built, development on Defense Grid 2 moves into its most complicated phase: adding the finishing touches.

Jump to

Something strange happens during the first few minutes of the meeting.

This meeting starts, as most meetings do, with witty banter and people goofing around, settling into place and wondering if everyone who's been invited has arrived. Then the talk turns to business. Just like any other meeting. But this meeting is about making video games, and it's full of game developers. So the line between where the fun part of the meeting ends and the boring, business part begins is a little harder to spot.

I'm in the big conference room at Hidden Path Entertainment, sitting in a corner, looking at the big TV screen. The art team led by Lex Story is here with Hidden Path co-founder and Defense Grid 2 Executive Producer Jeff Pobst, Lead Designer John Daud and Associate Producer Dacey Willoughby to show off the completed designs for Defense Grid 2's 20 levels.

The meeting begins with Story assuring the assembled other attendants that he has, in fact, studied NASA's plans for experimental railguns, and he jokes about hand sanitizer looking like Ghostbusters ectoplasm. The screwing around, the witty banter and jocularity then stretches on a bit, well into the work. In fact, this meeting will run for over an hour, and the point at which it stops being fun is almost nonexistent.

But that's not what's strange. Jokes and fooling around are normal for this team, and due to their experience, the tomfoolery never seems to seriously stall the process. What's strange is that the very fact of my observing them has influenced their game's development.

Story and his team have been hustling all week to put the finishing touches on these levels in order to show off their coolest new ideas in this meeting. And it's during discussion of that hustle that the strange thing happens: Pobst references a quote from an article I wrote about the making of the game (installment four of this very series, in fact) to make a point about the way Story operates.

Story is pointing out a minor aberration in the appearance of one of the core housings, a "gap" in the geometry between the housing and the ground it sits on. He's explaining that he experimented with the design to try and add detail, but the result didn't quite turn out right. He says he doesn't think anyone will actually notice.

"That's a gap because the geometry [creates] a gap," Story says. "If you're spending that much time [looking at it], you're losing the game."

"Let's get it right. We'll get it right. Yeah, I'll fix that."

Pobst responds: "What was the quote from the last article, Lex? It's not about perfection, it's about getting it right? Something like that?"

Story acquiesces. "Let's get it right. We'll get it right. Yeah, I'll fix that."

And then it's decided.

Story will go back and re-work that art to close that geometry gap. The game will be changed, hopefully for the better, because Pobst was able to quickly press his point using words I wrote after observing his team in action on a completely different day.

It is the first instance, in over a year of visits to Hidden Path, that I've observed the "observer effect" in action. My observation of the process of developing Defense Grid 2 has unintentionally had an impact on the development of Defense Grid 2. It's an outlandish impact on the already complicated process of making a video game — made even more outlandish by the fact I am in the room to observe it happen.

It's a 15-second interaction, but an impactful one. Afterward the team moves on to discuss the remaining game levels, the bits that will be added, the things that will be changed and all of the tasks that must be tacked on to the ever-growing list of jobs left to complete. By the end of this meeting, we will have seen all of the game's 20 levels, and although they are technically "complete," there's still months of work left to do on them.

This is game development.

Dg-2Marketing Director Shannon Gerritzen's white board as it looked in late 2013. Gerritzen uses it to keep track of projected announcement dates, planned content releases and ideas for marketing the game. The board changes frequently.


It's February 2014.

After a week of playtesting the current build of Defense Grid 2, culminating in a round of notes from Hidden Path co-founder and original Defense Grid designer Michael Austin, the Defense Grid 2 team has turned its attention to changes that need to be made to make the game fun.

For the programming team led by Matt "Twig" Johnson, this means pushing tasks aside to jump on the new priorities. For John Daud, this means waiting for those changes to be made so that he can then begin to make the game fun. He's occupying the time in between by addressing other issues that have, until now, fallen into the cracks.

"Now I'm doing some caretaking, some maintenance," he tells me. One of the things he's caretaking is the aliens that are supposed to have shields.

There are aliens in Defense Grid 2 with the power to give shields to aliens that do not have them. This makes the game more challenging and adds a strategic element. The shield-giving aliens can sometimes save the life of an alien near death by shielding it, which can in turn lead to that alien stealing a core and defeating you.

The problem: Right now the shields aren't showing up. Daud investigates.

"[T]he immediate thought is, 'Oh, you're missing a particle effect,'" he says. "So it turns into a rabbit hole."

The deeper Daud digs, the deeper he realizes he has to go. He discovers that the game code includes the proper commands to create a shield on the alien, but that it's calling for the wrong particle effect. Daud changes it to call for the correct effect, plays the game again, and the shield still doesn't show up. So he goes back into the code to investigate further.

"Then it gets deeper and deeper until all of a sudden a couple hours go by, and now I'm dragging Scott [Bodenbender] in," Daud tells me later. Bodenbender is one of the programmers working on Defense Grid 2, mainly on the multiplayer, but helping when and where he's needed. "I drag Twig [Johnson] into it first, and then [Bodenbender] gets dragged into it, and it's just that one thing."

One thing — an alien that should have a shield — that takes what accumulates to half a day and the attention of three developers to investigate. And even after that time, there's still no clear solution. Because making games is complicated.

Daud continues: "And we don't know when it [stopped] ... Because it worked! It worked! Right after PAX [in September 2013] we got this guy working in the game. ... And then we set it aside because there's other stuff to do. And then I add him to a couple missions, and I notice, oh, the shield's not there."

Bodenbender and Daud decide to ignore the alien for now, because they're replacing the particle effects anyway. They'll just add in the proper effect when they have the new one, and tackle the problem — if it still exists — from there. But there's a problem with that plan: an upcoming milestone.

Looking at the calendar, the team decides that now is the time to address things like this shield-generating alien.

"Best foot forward!" Daud says. "'We should fix the easy things. There's gotta be low-hanging fruit we've been putting off. I'll look at this.' And then [the investigation] reveals, 'Oh, that guy stopped working the way we thought he did. Maybe even months ago.'

"So, yeah."


Game development, in other words, is complicated. And Daud wouldn't even be working on this if he wasn't waiting on something else, namely the networking code for Defense Grid 2's multiplayer modes. That's been non-functional since September, and it's severely stalled Daud's design schedule.

I ask Daud what the big obstacle is with the multiplayer.

"I'm not entirely sure what the holdup is," he says. "I know that when we've attempted to play ... I think even right now, I can get a multiplayer game started. My opponent won't have any resources, so there's not much of a game there. But two players can coexist in the game. I can see a cursor moving around. And probably inevitably it'll crash.

"I don't remember. Once gameplay is missing, I kind of just stop paying attention. ... It's had a lot of attention [from others]."

By "attention," Daud means work. And the others, in this case, refers to Twig Johnson.

"I don't remember. Once gameplay is missing, I kind of just stop paying attention."

On the subject of attention

It takes me two days to get 15 minutes of Johnson's time.

In the past I've been able to grab him at his desk and take him away for a chat. This time I'm asked to get in line, and when I finally get him alone it becomes clear why: Every broken thing requires his attention, and there are a lot of them at this stage of development.

Right now, the lion's share of what will make the game fun is on Daud's shoulders, but increasingly, as bugs are revealed and technologies are needing to function perfectly for Daud to finish his own work, the weight of the project as a whole is falling on Johnson's shoulders. And right now his biggest headache is the multiplayer.

The saga of the multiplayer component of Defense Grid 2 starts with the original Defense Grid. That game had multiplayer networking coded in, but it was never finished and never activated. That work was examined when it came time to create Defense Grid 2, to see if it could be fixed and re-implemented. It was determined that starting over would be the better course of action.

That's where a small team including Jessie Bright comes in. This team created an all-new networking infrastructure to work with both Defense Grid 2 and Hidden Path's Windborne. Then Bodenbender took the handoff to get that code working in Defense Grid 2 in time for PAX, which he did, but only partially, and only just enough for the PAX experience to function in a head-to-head matchup on a closed network. Outside of that environment, however, the experience fell apart, and part of that has to do with Steam.

Defense Grid 2 will lean heavily on Steam's native matchmaking technology. Instead of having to create lobbies and other things specifically for Defense Grid 2, Hidden Path is instead tying into the Steam API for such things, allowing the service to do a percentage of the work of getting players in and playing together.

The problem is that even though the Steam technology is robust and effective, it is yet another codebase that must be tied in to Defense Grid 2's engine. And one the team was not necessarily planning for. That process, once begun, revealed some incompatibility in what had already been created for the game. That means: back to work.

I ask Johnson if the current status of the multiplayer is due to having to redo code that was added in as placeholders for PAX, making connections that were ignored for the PAX level or simply troubleshooting things that are complete but somehow not working.

"Uh, both?" He says. "All?"

So its still all coming together.

"All the work that Scott did worked great," Johnson says. "Then, as I added stuff, things became more complex. It's like, 'Oh, we need additional stuff here.' In some cases I ended up changing the way that he initially created stuff, because it wasn't working well with the new stuff that was added.

"That's not entirely uncommon when you're programming. ... There's not always a big drive to code for the unknown. Because it's time. It takes time to make sure it works and write it, and you may never actually end up using it. The down side of that is that when somebody has an idea that falls into the unknown, you may have to go back and change what you already wrote to work with that."

To further complicate the matter, Defense Grid 2 isn't being built to work only with Steam. It will also be released for Xbox One, PlayStation 4 and possibly other formats. Meaning even more code and even further complications that won't be revealed until work has begun on them.

Even when things work perfectly, it's a process that Johnson calls "messy."

I ask him if the process of getting multiplayer working is taking longer than expected.

"It always takes longer than expected," he says.

So if you expect it to take longer, is it still taking longer?

"Yes. Things like multiplayer are things that you plan for, and you plan for them to take longer, and they usually take longer than that."

And part of the reason things like networking take longer than expected is that everything else does too, and that requires re-planning, which involves meetings, which take time.

"I spent lots of time Monday and Tuesday talking to people," says Johnson. "And meetings and working with other people, it's like ... I don't really have time to sit down and get my own stuff done."

What he's too polite to mention is that telling me this is also taking up his time. We wrap up quickly and Johnson gets back to work.

Dg-1Lead Designer John Daud in a meeting with Defense Grid 2 team leads. it is yet another codebase that must be tied in to Defense Grid 2's engine. And one the team was not necessarily planning for.

Nice to have

Back in the level design meeting, if the troubles with networking and item sales have slowed the process of finishing the game, those effects are less evident here. As the assembled game designers examine each of the 20 levels in progress, the sheer volume of effort that's been thrown at the game so far becomes clear. Each level looks phenomenal, and more art elements and technological tricks are being added every day.

Daud walks the team through the design decisions affecting the layout of each level and Story outlines how and why each rate element has been created. Individual artists and developers step in frequently to defend or explain their own contributions. The result is like playing an entire game with a commentary track of a dozen or so creators, each explaining the thousands of creative decisions made in the course of creating it.

A bridge is debated here, sword-like entrance structures there, layouts of levels and the detail of complexity thrust upon the player elsewhere. Throughout, at each level, time is spent deciding where the elusive "key" or "boat" will be docked. A once-controversial design element that will whisk the player from level to level has been made official and is being worked into the design for each level. Yet I still haven't seen it. It and its animations will be a late addition.

The core housing with the gap that got Story in trouble with Pobst early on still isn't off the table. Another artist asks Story if it's going to eventually spin, like a turbine. Story says that it will.

"Eventually I want it to spin nice and slow," Story says. "The cores don't spin. It spins."

"Excellent," says Pobst.

"And then, eventually what I want to have is, when all the cores are gone those two slabs slam together," says Story, "they go, whoosh! The only thing that's keeping it apart is that containment, the energy of that core."

Willoughby cuts him off, concerned that "getting it right" has feature-creeped toward "overdoing it."

"Wait, wait," Willoughby says. "You want them to close when there's no cores?"

Pobst walks back his excitement. He says the added feature is on the "nice to have" list. But you can tell his heart isn't in it. It may be officially "nice to have," but he's expecting it will appear in the game.

A discussion ensues on the ramifications of adding the visual effect. If the core housing slams closed when all of the cores are stolen by aliens, it will have to re-open if the player kills those aliens and the cores are allowed to come back. But if the player doesn't ... then the cores leave the level and the player loses.

And that's when things get exciting. When the player loses, according to this impromptu design plan, the core housing explodes. It's a big "emotional response."

"Let's say the last core leaves," says Story. He's leaning in. There's movement, action and explosions. Doesn't matter that it hasn't been fully designed yet, this is what he lives for. "It's going down the road. Then there's another core that's really close to it. We give it a certain amount of time. It's not going to slam shut. It'll come shut after so many seconds."

Pobst, perhaps sensing the damage his own excitement has caused, continues to walk it back.

"Let's table this feature and discussion on the feature and you guys can design it," Pobst says.

"Then we can see if we have time for it." says Willoughby.

The discussion is then tabled and talk turns to the proper way for the art lead to insult his producer. The word "little" is ruled right out. He turns his attention to her heritage instead, and there's laughter.

As the meeting crosses the one-hour mark, debate over one of Daud's most complex levels, level 16, set in the top end of a planet's giant space elevator, turns serious. It's a monster of a map: dual elevator columns shooting down toward a massive planet below, surrounded by multiple spiraling pathways with multiple entrances, a variety of moving set pieces and an alien "warp point" that moves about the map causing aliens to appear from practically anywhere. It is designed to be challenging, but looking at it, Pobst wonders if it might be impossible.

"So just going into this, I love it, I think it's great," Pobst says. "It would not surprise me that when we focus-test this one, we're going to have to pull it back. [Players are] not going to be pissed, but they're going to be confused. Let's just assume we're coming back to revisit this one. That's my warning. That's all. Let's go on."

And then comes Level 17 — the bottom of the space elevator — and it's even more complex. The vertical space elevator that projects downward from the center of the previous level intersects in the exact middle of this one, terminating in a block of buildings directly in the center of the map. The aliens will follow pathways in and around and through those buildings, emerging from portals on all sides in unpredictable patterns, their exact paths hidden by the buildings overtop.

"It's crazy in there," Story warns. "You do not want to look inside there. ... Go ahead and look inside there."

Daud mouses over the buildings and the camera "clips" inside, revealing a labyrinthine maze of pathways.

Pobst, astonished, says, "They go left-right, and then they ... "

"It's crazy town, dude," says Story.

"The fact that you just came from one [level] that was really worrisome," Pobst says cautiously, "and then you go to this one ... that's really worrisome ... John, I'm worried we've lost our users at this point."

Pobst suggests removing the tops from the buildings to at least give players a fighting chance.

"So we're going to rip it all off?" says Story.

"I'm suggesting that," says Pobst.

"What do we do with our elevator?" says Daud.

Pobst suggests moving the large building cluster to the background and having the elevators terminate there, and not in the center of the map.

"We still want the elevators to come down," Pobst says. "And they can still go down into that thing. I just think that if they're not above our playing surface, that will be more interesting. I think you've said this. You've had a fear that you're not going to be hard enough for the player. I have no such fear.

"Try [the change] and let's see. This is the biggest level change we're making at this point."

"It's an interesting challenge, though," says Willoughby. And by "interesting" she means it will take time that they might not have.

Daud doesn't hesitate: "We can do it."

And it's decided. The "nice to haves" might have to wait.

"It's crazy town, dude."

Dg-3Every Monday the Defense Grid 2 team gathers in its corner of the studio to discuss who's doing what and why. It's a quick 10-20 minute meeting, but it reminds everyone what the goals are and how much work is getting done by other members of the team.

Temp magenta

The last level is where things get silly again.

It's a level set inside a giant crystal structure, with huge crystal shards hanging down from the ceiling. And it's purple. A very light, but very saturated purple. It's like looking at a battlefield in Candy Land.

"Wow," says Pobst.

"That's so beautiful," says Willoughby.

"Every time a mob dies it explodes into confetti," says artist Mark Forrer.

Pobst suggests the crystal art is effective, but cautions the purple color might be "strong."

"Don't worry," says Story. "Tuneage is happening. This is the grey box."

"Grey box," meaning the level is drawn with placeholder textures, not with final art. By "tuneage," he means he's going to alter the color of the light. The purple is temporary.

The artists, while waiting for final direction on exactly what color of light to go with, decided to use, instead of the usual grey (as in grey box), purple.

"So this really is temp magenta, then," says Pobst. And there is agreement.

At some point the the level will be redrawn in a new color, but it will most assuredly not be "temp magenta." The meeting winds down and the game makers get back to work.

The last time I visited Hidden Path, there were two Defense Grid 2 levels complete. Now there are 20. The tasks remaining are becoming simultaneously smaller in scope and more complex in execution: debugging multiplayer, tracking down missing effects, tuning fun and colors and identifying and eradicating problems that inevitably pop up whenever multiple complex systems are turned on and exercised for the first time, in concert.

The game, for all intents and purposes, is built, but it's only after the building is done that the thousands of missing details that will make the game feel complete are revealed. Now they have to be added. And the clock is still ticking. Babykayak

Continue to part eight of this series here.

This story is part of a series covering the development of Defense Grid 2. To read previous installments, please visit the Making of Defense Grid 2 page. This series will continue into mid-2014, the projected launch date for Defense Grid 2.

Images: Hidden Path Entertainment, Polygon
It's like looking at a battlefield in Candy Land.