Reactive game dev: How Serious Sam 4 became The Talos Principle

“We’ve only made Serious Sams, but now we’re here to talk about something completely different,” said Croteam CTO Alen Ladavac today as he kicked off the GDC 2015 Independent Games Summit with a talk on how the Serious Sam studio taught itself a new way to make games as it embarked on its first puzzle game: The Talos Principle.

The Talos Principle started out as Serious Sam 4. But as Croteam began experimenting with new mechanics, especially a “Jammer” mechanic, they realized they had more ideas for fresh mechanics than a Serious Sam game could encompass.

So they reacted by deciding to put Serious Sam aside and make a new game, a puzzle game that the team saw great potential in but couldn't quite predict whether or not it would be satisfying to play.

“It’s hard to predict which forces influence ‘fun’”, said Ladavec. “So you have to be reactive in your design. This is what we learned.”

Ladavec compared game development to the business of bridge construction, noting that a game falls apart when developers don’t account for unexpected faults, just like when civil engineers fail (very rarely!)  to account for an unexpected weather condition -- like when Washington's Tacoma Narrows suspension bridge collapsed four months after it opened in 1940 due to aeroelastic flutter.

“The moments when your feedback tells you your game is completely wrong are what I call 'Tacoma Narrows' moments, but you shouldn’t fear these moments, added Ladavec. “You should welcome them.”

So Croteam came to GDC this year to advocate for game design as a series of prototyping processes, pursued by developers with open minds willing to be reactive and radically change design -- even when they're late in the game.

What went right

The studio's approach to building The Talos Principle seems straightforward, but flexible. 

Internally, Croteam designers built puzzles for their new game and then tested each other’s puzzles in turn. That took care of fixing basic flaws, and then each puzzle would be played by the entire team to verify whether it was satisfying to solve. 

At this point Croteam also worked out a new way to collect statistics on how difficult each player found each puzzle, which (when collated into a giant spreadsheet) made it easier for the developers to tweak the puzzles and lay them out in a proper order.

After the game was rejiggered in accord with internal play testing, Croteam took it external -- and immediately felt pressured to react to a host of new issues.

“A game that had seemed complete was showing a lot more problems,” said Ladavec, noting that it took a long time for new players to get through the first world of the Talos Principle alpha and many felt bored.

“We removed a lot of redundant puzzles,” said Ladavac. “It was a great deal of work, but very worth it, because the players had a lot more fun." 

They also removed a level completely (Rome) and extensively reworked the game’s Sigil system, as well as opening up the level design of Talos Principle to give players room to release stress and think as they explore.

Croteam CCO Davor Hunski jumped in to add lots of smaller examples of how Croteam substantially reworked game mechanics like the user interface, the in-game auto-connector items and the exploding mines.

He also encouraged other developers to take their games out to public-facing events like E3 and PAX, noting that taking The Talos Principle to events where lots and lots of people could play the game in a short period proved invaluable in the team’s iterative design process. 

Hunski also spoke positively about Croteam's experience launching a freely-available public alpha of The Talos Principle prior to launch, as it served the dual purpose of drumming up press for the game and reaping player feedback to help the team refine each puzzle’s design.

“There were a lot of very small fixes that really changed the entire game, not just the demo,” added Ladavec. He described the process as somewhat exhilarating, excitably noting that Croteam saw forum threads generating over 1,000 replies a day as the public began digging into the Talos Principle public alpha.

Bring in a bot

Croteam also developed an AI bot that could do automated testing. What a human tester needed ten hours to do, a bot could do in four hours…of game time, which Croteam could then speed up even further.

“It’s unexplainable how good this bot is,” said Ladavec. “We would need to put eight full-time testers to test the game all day for a year” to equal the QA work Croteam’s automated tester tackled, said Ladavec, which might well have bankrupted development.

The bot’s auto-generated test results were also much easier for the developers to work with than feedback from human players, because an engineer could simply highlight a reported bug, skip to the exact moment when the AI tester experienced the issue and recreate it from the bot’s perspective. 

“He reported more than a thousand bugs like, ‘I’m stuck here, this door won’t open’ and then the developer could just pop right in and fix it,” said Ladavec. "It made it far easier to react quickly."

Hunski added that Croteam also wound up significantly changing The Talos Principle to accommodate the divine narrative that was added in late during development, adding music, easter eggs and more. 

Despite the extra work involved, Hunski encourages fellow developers to be receptive to late-game additions and significant “value-added” content like an interesting narrative or hidden secrets because “it really brings very much to the table,” potentially reinforcing your game around an unforeseen weakness.

"To enable writers to do their jobs properly and easily, we game them tools to quickly test their dialogue in-game,” said Hunski. “We then had to make huge changes to support the narrative our guys were delivering, and it ended up being much, much bigger than we planned, but we thought it was okay because it brought something fun to the table.”

What went wrong

While they're satisfied with the game's performance, Croteam says they didn’t have enough time to make necessary changes before launch.

Hunski thinks the game’s “Gates of Eternity” narrative element should have been easier to see, and that more puzzles should have been trimmed with rebalanced Sigil puzzles. He also noted that Croteam isn’t satisfied with the game’s hint system, and should have put some more work into it before release.

The game’s localization was also a thorny issue for Croteam. “Localization always ends up at the end of our projects, and it’s never scheduled right,” said Ladavec. Localized versions of Talos Principle were initially rushed to meet a deadline that wound up being extended, wasting effort. Then the team slipped up again by waiting until the last minute to do localization, in order to try and minimize the need for revisions.

However, the team also learned to appreciate working with individuals instead of dedicated agencies.

“We know this guy who localized it into Spanish, and this girl localized it into German,” said Ladavec. “Those localizations were much better than those we outsourced to an agency,” because the individuals were more passionate about the project and were more likely to ask specific questions about meaning and intent. 

“Another thing that went wrong for us was the effort tee put into launching this companion app, Sigils of Elohim,” said Hunski. “We thought it would have a huge player base," but in the end Croteam's Steam playerbase dwarfed the number of people playing Sigils on mobile.

Finally, Hunski admitted that the first world of The Talos Principle was probably the wrong one — given another chance, Croteam would have picked something other than a (somewhat generic) medieval setting. 

Together, he and Ladavec then reminded game makers to be reactive -- to voraciously gather feedback from your players and your fellow developers, and be open to making huge changes in response -- even if it's expensive or changes your vision.

“If we hadn’t done all this, we wouldn’t ever have started making Talos Principle,” said Ladavec. “It turned out to be a game we enjoyed, and now we’ll be relying on reactive development for future titles.”