Reversing the sunk cost fallacy: Devs recount what shouldn't have been cut

Toward the end of 2017, we ran a story on sunk cost fallacy in game development.

It featured around a dozen developers describing the price they paid for ignoring red flags and continuing to work on bad features or entire projects that were set for failure simply because they'd invested so much time, money, or energy into them.

A short time after the article went live, we got a suggestion from Axiom Verge developer Tom Happ to do a follow-up piece looking at the issue from the opposite direction: what about good games or game features that got cut when perhaps they shouldn't have been?

Like sunk cost fallacy, this is an easy blunder to make. Game development requires juggling so many disparate-yet-interdependent elements that it can be hard to keep track of which ones are the lost causes and which ones should be prioritized as important. But as many respondents noted when I put this question to them, it's much harder to actually gauge whether you've cut something that you shouldn't have.

Most cuts are made well before completion — when a game and its systems are still taking shape — and their removal is rarely an independent event without consequence both good and bad on other parts of a project. To extend the juggling analogy, if you let one ball drop then it may be easier to keep the others under control. But it may also knock other balls off course on the way down.

And as veteran designer Noah Falstein notes: "The problem with regretting cutting features is that often the best parts of a game are unexpected to the developers, or at least can have a surprising impact. Conversely, a lot of things that a designer might think will be fantastic turn out to be just adequate. Game design is still as much of an art as a science, and as such success can be inherently difficult to predict."

All the same, as you're about to see, there's plenty we can learn from the thought exercise. Even that difficulty to predict the impact of design decisions and feature cuts, for instance, can itself weigh on a dev's mind.

"There was this one quest we cut from The Witcher 3," says CD Projekt RED's lead quest designer Mateusz Tomaszkiewicz. "Players were supposed to infiltrate the Wild Hunt in order to eliminate one of its generals, Caranthir. In the process, players would learn more about Wild Hunt's motivations, and have a chance to interact more with them. On first glance it sounded cool and beneficial, and from time to time I'm wondering was cutting it a good decision, but it generated a lot of design issues."

With the quest included, they'd have had to worry about ways to prevent the player killing everyone there and then. And whether or not to make new animations for moving around in heavy plate armor. And a host of other problems that didn't seem to go away after a few iterations on the quest's design.

Pivot pains

Happ has one of the more concrete examples of a cut that perhaps shouldn't have happened. He worked as an engineer at Petroglyph Games during development of cancelled MOBA End of Nations, which he remembers went through a few design variations over a long development cycle at publisher Trion Worlds' behest before initially settling on a massively multiplayer real-time strategy format. Happ says the game was finished in this format in late 2012, and that it was the best RTS he'd ever played.

"I couldn't wait for them to make it live so that I could play it with the world," he continues. "But once we were at the end of our task list there was a week or two where we were all like, 'It was supposed to go live on day X, why is Trion holding back?' We had moved into a new office in anticipation of switching to a live support role for EON while working on other games. The day of the move, we found out the game was cancelled."

Worse, Petroglyph had to make big staff cuts to deal with the financial fallout of this unexpected change. Meanwhile, Trion moved development in-house and turned End of Nations into a MOBA, stripping out nearly all of the RTS systems they'd spent years honing, only to then cancel that version the following year.

The suddenness of these events and the lack of forewarning that Trion was unhappy with the game took a toll on Happ's psyche. "To this day (just past the five year anniversary) I still feel traumatized by it," he says. "You always hear about trauma in regards to things like gun violence or sexual abuse but I think layoffs should count as trauma, too, and I wasn't even one of the ones laid off.  It was just like my whole world changed overnight."

Question producer/designer Michael Kelly drew various lessons from the ill-considered decisions to pivot that he witnessed during his years in triple-A development at 2K Games.

"Pivoting a video game is like turning an ocean liner," he explains. "You can't pivot on a dime in AAA with large teams and/or multiple studios. From these experiences, now I'm a believer in scoping early and often — down to the very minimum core of a game if I have to. The features which ultimately ship should have ample time to get polished rather than making lots of non-essential bells and whistles which stretch the team too thin."

He cites The Bureau: XCOM Declassified as an example. Its development was led at various points by different teams with different visions, each of which left ghostly impressions on the design, art, and code as the next team refactored the old work for the new direction because they didn't have the resources to start anew. "We had production workflows for art and code that could not be removed even though the features they were designed to support were no longer in the game," says Kelly.

When you pivot on a game's direction, Kelly explains, you not only create inefficient workflows but also harm momentum — as many features must be redone and some tools built for the old job don't fit the new job. "And you now have less money and time to finish," says Kelly. "Or to break even on your investment, you will now have to sell 30% more units than your original target. If you had just shipped the initial design, in that same time you would have a chance to pull in revenue at this point." (And as a point of career advice, Kelly adds that it's better to have another shipped game on your resumé than a gap of a year or two without a title credit — or better to have 10 years experience with multiple credits rather than zero or one, as in the case of the poor souls who worked on Duke Nukem Forever for its entire development.)

In many projects it doesn't even matter if the new direction might be better than the old one. Pivots can and sometimes do succeed, with final games that have a clear identity and cohesive design, but more often developers find that the legacy decisions of old come back to haunt them. Kelly concludes: "I think lots of developers who spend a few years on a pivot ask themselves, 'What if we had just cut the things that didn't work, doubled-down on the core loop, and got it out the door rather than spending another year (and lots of money) trying to reconfigure this into the new design that is also unproven?'"

External factors

There are lots of reasons for cutting features from a game during development. Some of them relate to the (lack of) quality or suitability of said features for the project, but many of the factors that drive cuts are external. Starcraft: Ghost, for instance, was reportedly cancelled in large part because World of Warcraft was all-consuming at Blizzard for a few years.

Medal of Honor: Airborne lost more than a year of development time to an EA request to switch from the Quake III engine to Renderware 4.0 (only to later switch to Unreal 3 after 16 months of struggling to get the incomplete Renderware 4.0 engine to do what they needed). Dead Space 3 also had issues caused by EA requesting changes, with much of the narrative depth and complexity it was criticized for lacking cut to fit in more microtransactions and marketing bullet points, while a promising Core Design-developed Tomb Raider 10th Anniversary remake was cancelled despite an impressive demo build because publisher Eidos/SCi changed its mind and gave the remake to Crystal Dynamics, which had just completed Tomb Raider Legend.

Unravel creative director Martin Sahlin describes one occasion where an entire control method was cut from a game he worked on just prior to launch, at the behest of the marketing department. "Maybe marketing was right," he says. "I don't know, but it felt like gutting the entire game."

More rarely, good-quality features or entire games can fall afoul of pessimistic business practices. Veteran artist and illustrator Mark Ferrari once told me about a cancelled X-Men TV controller game he worked on in the 1990s. They had a tiny budget and production schedule, but between them Ferrari and the rest of the team figured out how to make a good game by using a number of palette manipulation tricks. One art file, for instance, would display a valley filled with trees and fluffy clouds using one palette and a gloomy cityscape with bridges and freeways with a different palette.

"We sent it off expecting them to call amazed at what an astonishingly awesome product we'd made for so little money in so little time," Ferrari recalled. "And what we got was a really angry call to the head of our division from the head of that project at this big company, saying 'We did not ask for this, and you have really screwed us over.'"

The client company was angry because if they were to release this actually-good 8-bit TV controller game they would render the rest of the line of games — already available in stores — unmarketable. It was too good to release.

Happ remembers a less severe example of a feature being cut from a game for business and marketing reasons. "I was a technical artist on Tiger Woods PGA Tour 2008 and 2009," he recalls. For the 2009 installment, one of his colleagues decided to try bringing the dynamic hair system from the Battlefield franchise over, to replace some of the awful helmet-like polygons that served as hair in their golf game.

"He got two or three hairstyles working with this kind of hair," continues Happ, "and it looked much better. But those higher up didn't want to use it unless all the hair options could be replaced with this new system."

Tiger Woods PGA Tour 2009 hair in action

The dev team didn't have the resources to make this transition within one year-long cycle, so they cut the hair that was already made. The most recent entry in the series, Rory McIlroy PGA Tour 15, still retains the old helmet hair tech — likely for the same reason.

Contractual obligations

Fun Bits co-founder and CEO Chris Millar's entire career is dotted with great multiplayer modes that should have been. After he left Blizzard in 2002, Millar went from game to game — almost without exception — seeing fantastic multiplayer modes take shape only to be killed or undermined at the expense of inferior single-player experiences.

First there was Goblin Commander, an early attempt at real-time strategy on consoles that he worked on at Jaleco. "We got [online] multiplayer working on PC [and] it was working on Xbox," says Millar. "At the time, the platform manufacturers were like, 'Everyone either has to be all the same or everyone has to have something unique.' But if we released multiplayer on Xbox, then PlayStation and Gamecube would not let us through certification if we had a feature that was better than their platform."

In order to release on all three platforms they had no choice: cut it. Goblin Commander released on all platforms with only single player and two-player split-screen modes, to the chagrin of reviewers and Millar believes to the game's ultimate success.

Next he moved on to Lionhead, where he first helped wrap up Fable development then landed on the 80-strong team building god game/RTS hybrid Black and White 2. It was an ambitious game, and Millar arrived just in time to fall in love with the city building and multiplayer modes — he describes the former as an "insanely cool" chill-out Farmville-esque experience that predated Farmville and the latter as the most fun mode in the game for its monster vs town and god fighting god madness — before the realities of contractual obligations, budget restrictions, and scant time remaining forced some tough decisions.

Facing cancellation if they missed their deadline, they left out an open-ended free-form city-building mode and cut the multiplayer — which, while more fun and unique, could not be left as the only game mode.

Lesson learned, Millar later looked at PSN and decided that, since anyone who buys a game on PSN must have network connectivity (or otherwise they couldn't have downloaded the game), he could now go multiplayer first. That led to Fat Princess, a critical darling and cult hit 32-player action-RTS, which somewhat ironically ended up with a single-player mode shoehorned in at the last minute at Sony's behest.

Millar was to face his cancelled multiplayer curse one more time, however, with a PS4 sequel that was prototyped as multiplayer-first only to swap to a co-op campaign with a planned multiplayer mode add-on (that was finished and ready to go) because Sony wanted a different direction. It tanked on release as fans and critics alike struggled to make sense of the new direction. Millar holds no ill-will toward Sony, who he calls a "great partner", but that decision to de-emphasize multiplayer in favor of a campaign mode took its toll — "they canceled our future contract discussions," says Millar. "And we had to shrink the company down."

For Millar, who's now once again trying multiplayer-first development with Squids from Space, these experiences show that it's critical to focus on what's most engaging for your audience before you even start to tinker with other elements and game modes. For him and the companies he's worked for, that's online multiplayer. If they can get the core product right, then that gives them the best possible chance to get their core audience, and then Fun Bits hope to leverage this core product in other directions only after they have a strong foundation that's fun and has some traction in the gaming community.

No cut is harmless

While it can be hard to predict in the moment which features will go on to make or break a game, or in hindsight which features might have elevated the design to a higher level, it's important to have the courage to kill things that you aren't confident in as a team — even if it means disappointing certain individuals or a portion of the game's audience.

Bruce Shelley has one such example of a big company meeting during Age of Empires III development. It was about an elaborate formations system. "It was like 51 keep it and 49 forget it, and we had to make a call," he recalls. "The decision was that was not a strong enough majority — we had to be like 80 percent keep it, not 50 percent keep it or 51." With only half the team still excited about the feature, it had to go.

It's also important to remember that no cut is harmless or free from overhead. "Cutting something means that you have to do a number of changes in other places," says Tomaszkiewicz, "so yes, you'll gain some time in production that you'd need to finish it. But you'll have to allocate time for adjusting the remaining content to that change."

And Sahlin points out that cutting a feature during development, for whatever reason, that perhaps could have been something special need not be the end of it. "It's actually quite nice to go back and review stuff we've cut every now and then and imagine what it could've been," he says. "Sometimes we find things to salvage, too."