Jeff Kaplan and Michael Heiberg are debating the relative popularity of ice cream flavors.

“I always thought chocolate was the flavor everybody would like most,” says Kaplan, Overwatch’s game director.

Heiberg, a principal game designer, shakes his head. “Statistically, vanilla is more popular.”

“All right, so—statistically, vanilla is the flavor,” says Kaplan. “People have a propensity toward certain ice cream flavors, just like they do toward roles in Overwatch.”

They’re trying to explain how they came up with the Triple Damage (one tank, three damage, two support) team composition set to be tested in Overwatch’s new Experimental mode. It’s a significant change, and the team isn’t sure how it’s going to work out, but that’s the beauty of Experimental mode: allowing players across all platforms to test the dev team’s wildest ideas without leaving the client. 

Kaplan and Heiberg laid out the full story behind Experimental mode, plus a full breakdown of what Triple Damage means for the game.

What prompted the team to create Experimental mode?

Kaplan: When we make balance changes, they have to go to the PTR [Public Test Region]. While they’re on the PTR, we’re also going through the console certification process for Sony, Microsoft, and now Nintendo. On average, that takes a week—and that doesn’t include testing or deploying, so it can be a long process. To change the damage of McCree’s gun by one point takes longer than players think.

Then, from a balancing standpoint, we also want to live with those changes for a while before making our next round. So, it ended up putting us on an almost bi-monthly balance schedule, just by the nature of the process.

If you asked any designer on the Overwatch team, “what are the top 10 features you want?” one of the things they’d say would be, “The ability to immediately change things and be much more responsive to balance requests.”

When did the opportunity to create that feature arise?

Kaplan: There were two engineers that were most proactive in the Experimental idea. One was Derek Mulder, a cutting-edge gameplay engineer, and the other was Phil Orwig, our mastermind network engineer. Derek proposed the idea of, hey, imagine if we had a version of the PTR in the game that everybody could play.  

This excited us because players wouldn’t have to download a separate client, and all our rewards systems, like leveling your account and earning Loot Boxes, would still work. Plus, console players could be a part of it for the first time.

It was one of the easiest design decisions ever. It never works this way—where somebody has an idea and everybody else is like “That’s absolutely what we should do!”—but that’s pretty much how Experimental mode happened.

The first change you’ll be testing is a Triple Damage—1 tank, 3 damage, 2 support—team composition. Where did that idea come from?

Heiberg: When we rolled our Role Queue with 2 damage, 2 support, and 2 tank players per team, we ended up with a lot more damage players in queue than tanks and supports. This created longer queue times, and one of the things we considered was, what if we tried to make the team composition closer to the actual ratio of players by role? If that did work out, people would be waiting less, playing more. So, it behooved us to actually try it.

How did your internal playtests go? Any interesting observations?

Heiberg: When you start having more damage and fewer tanks, the amount of lethality in the game changes. The amount of pressure on the people playing tanks—and the people trying to keep the tanks alive—changes.

Kaplan: With three damage players, it opened team compositions that we weren’t really seeing in the current game. We’ve seen compositions with two flanking characters, like Genji and Tracer, supported by aggressive tanks. We’ve also seen compositions with two snipers, like Widowmaker and Hanzo, with a more defensive playstyle. Now you might end up with two snipers but also have a flanker. So the game really plays differently. It’s a lot more chaotic. A lot more action-based.

It polarized the dev team. People tended to either hate it or be intrigued by it. Most people were cautiously optimistic, like, “Oh yeah, this could be really interesting.” But then we had people staunchly opposed to it who said, “This is so different, it puts too much pressure on the tank player, the game feels too lethal.”

We’re running experiments to see if we can balance our way out of those issues. Like, with the game feeling too lethal: certainly, the number of tanks that are present can be an influencing factor, but we could just as easily look at everybody’s damage and hit points and make adjustments there.

Can you talk about some of the specific changes you’ve considered?

Kaplan: The biggest question was: What happens to Roadhog, D.Va, and Zarya?

When we created our tanks, we didn’t think of any as off-tanks. In fact, we had anticipated people playing with only one tank. That’s much of how early Overwatch was played. If you ever jumped into Quick Play before Role Queue, you know that having one tank was a luxury.

But the way that players settled into 2-2-2, Zarya, D.Va, and Roadhog are exclusively considered off-tanks. As part of this experiment, we’ve re-tuned those three to be main tanks—including giving Roadhog a new ability. Now when he uses Take a Breather, he gets a gas cloud around him that mitigates damage for everybody in the area and heals him.

And all those changes will be in Experimental mode? Not just—oh, we changed the number of roles, but also an overhaul of the entire roster?

Kaplan: Some heroes will be completely re-balanced, but those changes only pertain to Triple Damage. If we stay with 2-2-2, we are gonna jettison those changes.

Heiberg: That Roadhog would be broken in 2-2-2.

Kaplan: You’d see him in every match.

How do you hope Triple Damage will affect the average Overwatch player?

Kaplan: This is new territory for us. When we put stuff on the PTR, those are typically changes we’re confident in and likely to push live. This is more like us sharing an experiment with our 50 million closest friends.

We’ll be looking at two things. One: player feedback and gut feelings. That’s super valuable to us. And second is the effect on queue times. The goal here is not to change the meta—it’s to reduce queue times.

After Triple Damage, are there any upcoming changes you can discuss?

Kaplan: After Triple Damage, we will probably focus on balance changes in Experimental. At least until we cook up our next wacky experiment.

What are some of the reasons you feel players queue up for damage more often than tanks?

Kaplan: There are a lot of different reasons. One is simply that there are more damage heroes to choose from than tanks right now. I also think the reward mechanisms in the game are skewed toward damage. If you think about all the visceral feedback you get from a kill—we light up the kill feed, you get an elimination message, there’s a very satisfying sound—we don’t do quite the same level of diligence when it comes to things like how much damage did you block, or did you save somebody from dying by healing them at the right time. And that’s on us. That’s something that we need to fix over time.

That said, there’s a lot to be attracted to in the tank role. Wrecking Ball is one of the most skillful and creative heroes in the game. The diversity in playstyles if you look at Rein versus Sigma versus Orisa—none of them play the same, even though they’re all barrier tanks. I think that they’re unique and fun to play. But I think fewer players feel like they’re up to the task of playing a tank. It’s a high-pressure position. The team counts on you

Is one of your goals to get more people to play tank?

Kaplan: Not necessarily. This is how I’ve been thinking about it: imagine we’re an ice cream store, and we have three flavors of ice cream. We have chocolate, vanilla and strawberry, and you have to line up for all three flavors separately. So, imagine the vanilla line is way, way longer than the chocolate or the strawberry line. I feel like it’s the wrong philosophy to ask, “What can we do to convince the vanilla people to like strawberry more?” It makes more sense to say, “We need more vanilla ice cream!” 

That’s a great analogy.

Heiberg [scrolling through his phone]: Actually, the most popular flavor—it looks like it might be chocolate after all. Depends on the source.

Kaplan: Well, we know it’s not strawberry. Strawberry is the tank.