One of my notable tasks on War Robots was developing a content release plan. It least that is what I was asked to do. What I actually did there was createing a whole content production strategy, design framework, and new tools for GD and CM.

It started with a single thought: people are different. They want different things. So probably the content we release should be different as well... And that changed the project forever. If in the beginning we had a simple CoD-on-mechs type of game, eventually it transformed into a sort of hero shooter. Along with many other radical changes.

So, people are different. Different people create different market niches. What are those?
By that time we had 40-50 different mechs released. Inside we had some idea of what types of those mechs were, what was the purpose of each mech. But I wanted to verify our idea.

So first thing I went to community management dept and asked them to do a player survey. With the question "What mech categories do you think are present in our game and how you would distribute mechs between those categories". This technique is present in traditional web UX. Like when they do an online shop they ask focus groups to distribute goods between categories. Because common people have common sense, unlike us specialists.

Sometime after CM returned to me with survey results. And surprisingly most of the respondents had quite a similar picture of the game. They named the same categories and their ways of distributing mechs were similar as well. That was a pretty solid first lead.

So we had like following categories of robots:

• Glass cannon/fighter
• Brawler/tank
• Sniper
• Support
• Saboteur

And some lists of robots within those categories. I needed to check what's common between robots in the category.

The common thing was their stats. I checked key gameplay metrics for robots - win rate, lifetime, and damage done. And within a class robots had quite similar stat profiles.

• Glass canons - explosive robots. Short lifetime, high damage, average win rate
• Brwalers - opposite. High lifetime, low damage
• Snipers - high damage, high lifetime, low win rate
• Saboteurs - no damage, short lifetime, high win rate
• Supports average lifetime and win rate, low damage

While damage and lifetime are more or less obvious parameters in statistics and class design, win rate picture here is interesting. Some robots win more often with no damage done. And some lose with a lot of damage. That's because win condition in the game is capturing control points. So light saboteur robots can leverage mobility to win more without participating in combat and dying fast. Snipers dish out a lot of damage but do not interact with control points and don't move the team to victory.

Gameplay patterns were clear in raw numbers. And what's more important, players were able to feel those numbers. Even when designers had no classification and no idea about this meta, no UI hints, players were able to correctly categorize the content.

This is not only my work. I did an initial check by looking at numbers, but then analytics ran a proper cluster analysis and offered some valuable corrections to my model. Later they took over this project and expanded it a lot.

Being able to reliably put mechs into categories we became able to analyze both categories as a whole and individual content.

Discovery 1) Now we could estimate sales in the market niche. Glass cannons sell pretty well, snipers not so well. Tanks and saboteurs in the middle.

Basically, this is a guide to what content to produce with what frequency. over a year 3 glass cannons, 2 tanks, 2 saboteurs, 1 sniper, 1 support. At this point, we can consider the initial task solved. This is the content release plan. But there's much more.

Discovery 2) It explained the sales of some content. For example, there was mech Bulwark. A tank that got sold much worse than we expected from him. All because by its stat profile, it turned out to be a sniper. By sniper benchmarks it performed pretty well. Just snipers as a niche is much weaker than other classes. And that explained A LOT.

That was a turning point for content design on War Robots. We developed a set of markers and benchmarks for different class designs. Long story short, after that there were no flops like Bulwark. Intention and implementation always matched.

Consequentially having this knowledge and my previously developed balance approaches changed our live ops paradigm. We started to use a matrix similar to R6 Siege, but adjusted by robot classes. We counted presence in meta, efficiency, and sales and introduced necessary changes when needed. That wouldn't be possible without sorting mechs by classes, because it would be a mess. We would confuse a good sniper for a weak robot because of the low win rate. But within War Robots gameplay a sniper would never have a high win rate.

This project introduced a lot of changes to the game and how we saw it.

1. At start we designed individual content. Like we were making a generic shooter. While in fact we were a class-based/hero shooter. A proper understanding of our game focused our design efforts and really boosted pipelines. Design quality rose along with player satisfaction with content (based on surveys).
2. Proper estimates of content niches and niche rotation through the year gave us tools to predict profits, build optimal release plans and rotate content in meta optimally.
3. This led us to developing new analytics tools and live ops approach, that are still in use after 6 years or so.
4. Transition to class-based shooter gave a new tool to community management. Before we were saying "Check this new robot out"! We didn't have much to say about the robot. Now we were saying "This is a new saboteur. This sneaky guy deals not so much damage but is irreplaceable at capturing points"! We could formulate a message, target it to the right audience, and put the right emotion into it.