One of our most anticipated games of 2017 was For Honor. It had vikings vs knights vs samurai. It had a new, melee swordfighting system. It had synergetic systems stolen from fighting games. And at the preview events, For Honor was a lot of fun.
That’s the problem with the controlled environment of a preview event. For Honor as a pre-launch product, with everyone playing in the same room, was fine. We could all enjoy the actual design of the game. But For Honor in the wild was a different story entirely.
At the recent Game Developer’s Conference in San Francisco, Ubisoft was open about some of the mistakes made on the game — mainly, the network architecture that rendered the game unplayable for so many people.
It must be one of the hardest holes to dig yourself out of as a development team. Game director Damien Kieken and creative director Roman Oriola went through their strategies for retaining players while working on a massive fix to For Honor‘s multiplayer.
A Promising Beta
Ubisoft was clearly proud of its progress during the For Honor beta. Hundreds of improvements were made to the game, it brought on six million players in the public window, and became fourth in all time fighting games.
These betas were held for different reasons. In February of 2016, there was a genuine server load test. September of 2016 was an open beta marketing exercise, to get the word out about the game. By December 2016, the full build was ready.
But even at these early stages, signs of a network architecture that wasn’t up to the job were surfacing. The open beta had players desynchronising all over the place, instantly kicked out of the match as soon as it would start.
The result was something like this:
Ubisoft was aware of the issue, and had about a month before launch to come up with a fix. Its answer was to have For Honor populate your match with bots as soon as you desynchronised from the match.
All of a sudden, everyone in your game would appear to have left, and bots would join in their place. It would often happen quite early in the match. The hasty “fix” at least allowed players to get to the end of a match to earn some persistent rewards like money and gear. But it wasn’t multiplayer, and it wasn’t fun.
On Steam, the reviews were brutal.
Balance and Exploit Fixing
A few days after launch, the developers were having a tough time against players. Not just versus their complaints in real life, but actually playing the game against them.
That’s one of those cute anecdotes that developers love to tell, giving credit to their skillful communities. And certainly it’s not unusual to be bypassed by fighting game pros very quickly. Nonetheless, dedicated For Honor players might roll their eyes at the comment. Had the developers been aware of a very obvious exploit, the competitive scene could have had a much better start than it did.
The first major For Honor tournament, with $10,000 USD on the line, highlighted an exploit in which players could start an attack animation while locked on, then terminate the lock, and their attack would become unblockable.
The competition devolved into who could exploit the best. The winner cheesed his way to victory, which was slightly acknowledged on-stage by the hosts, and For Honor was embarrassed in the competitive scene.
Servers With a Smile
Here, we get to the main crux of it.
For Honor tried to create a hybrid network architecture which wasn’t quite your usual peer-to-peer system. Instead, it was a “meshed P2P” system. There was no host, and everyone was sending info to everyone.
The developers called this a “deterministic simulation.” There was no randomness in the game session. Each client only had to send the inputs of each player, not their location or animation, which would lower the bandwidth needed.
“We thought we had the magic formula, and we had a year and a half to test it,” said game director Damien Kieken. “But after launch, we started seeing a destability in 4v4 matches and didn’t know what was going on.”
It worked better in the 1v1 mode, and somewhat better in the 2v2 mode — as the system was closer a system to what fighting games use. But as soon as you tried the main 4v4 mode, it would putter and die.
For some, it was alright. But for a large portion of the playerbase, it was unplayable. You wouldn’t even get close to the end of the match, which meant no persistent rewards.
“There was a glass ceiling. With the meshed P2P system, the weakest link impacts everybody in the session,” said Kieken.
It was clear that a change in architecture was needed. After many public statements about how the system should work “in theory,” it was clear that in practice it wasn’t doing its job. But switching to dedicated servers meant eight months of work to change the whole online structure of the game.
The challenge became to keep the playerbase interested, and continue fixing bugs, while completely remaking the online portion.
Operation ‘Save For Honor‘
After launch there were different expectations on the team. There were constant emergencies, frequent milestones to hit, lots of interruptions, and there had to be a high amount of quality assurance to avoid regression.
This had an impact on the team’s morale and fatigue. “By doing the same job they were doing before, they had less results than before,” said Kieken.
It probably also didn’t help to see dozens of news stories about people not being able to play For Honor.
In March of 2017, the 1.04 update was released, which brought six new maps, 15,000 bug fixes, and overall half of the launch content. In addition to making sure lots of metrics were in place to track how often abilities were used, as well as win percentages, more qualitative balance measures were taken.
All designers were in a Discord channel with the game’s top players. Regular private balancing tests were held to make sure the game was also balanced for those not on the top tier.
But with all of the activity, it was important to preserve the team. To that end, Ubisoft removed public deadlines. They were only internal for features and improvements, and there was flexibility for emergencies and player requests.
“For the players, that way you’re only doing what you say you’re doing,” said Kieken. “We had to be very transparent and show early progress. Otherwise they think you’re on vacation and you’ve just taken their money.”
The team hosted weekly livestreams to acknowledge problems, explain the strategy, and share priorities. With no public deadlines, it was all the more important to show progress and iterate with players.
“We spent 1-2 hours a day trying to be reachable online,” said Kieken.
Through new seasons, big streams, new content, and free weekends, Ubisoft hoped to recapture inactive players and press, and lure new players in. It was important to fund all of this post-launch development with in-game revenue, but certain things like maps, modes, and events were always free. Even the Season Pass was free — buying it would only guarantee earlier access to the content.
Ubisoft also decided on a single currency, earned by playing and available to buy. Through that and the freely available content, it never split the For Honor playerbase. Weekly content, a changing meta, and introducing more game modes gave people a reason to keep logging in and increased retention.
After the Battle
Recently, For Honor brought out its dedicated servers. But in the process, Kieken and Oriola learned a lot about post-launch development.
“Our job was evolving from gamemaker to entertainer,” said Kieken. “By breaking that routine, and introducing surprises. We made lots of trailers and images to promote the events.”
They witnessed 20% more playtime during the special events, in which alterations were made such as using skeleton models for grunt NPCs during Halloween. It was also fun for the team to work on.
The entertainer mindset moved outside of the game and into the weekly show as well. People eventually tuned in not just for info on the game, but purely to be entertained. It’s accumulated over five million Twitch views so far.
“It’s like running a marathon, but you remove the finish line. It can get really hard at some points,” said the developers. “If we could go back in time, we’d say it’s important to start preserving your team earlier, not when you launch the game. And the sooner you can work with your community, the better.”
All’s well that ends well. For Honor turned around a catastrophic launch and, one year later, is a very decent, playable game. No one was ever doubting the game’s design — it was always the technical issues getting in the way. Now that there are so many quality games coming out, it probably won’t ever reach that same level of interest. But it’s great to see its original potential fully realised.