How do you balance real-time responsiveness and fairness in multiplayer games?

Lag and latency can be a barrier to immersion in gaming. Latency is especially an issue with cloud gaming, which many dub as the future of our medium. However, latency MUST be tackled for cloud gaming to truly take off, especially for multiplayer games (the top games on console and PC). We are beholden to physics (and Internet infrastructure).

I almost always get my head exploded when I think about how extremely fast we can transfer information around the entire globe. But that's a double edged sword when it comes to multiplayer games. When you give players extremely fast response, they expect nothing less from other games as well. Designing your game around a point where you can allow your developers for tiny pockets to adjust, interpolate, re-order things around is a big plus here. Communicating this to your players as well usually helps. Rocket League is a great example for this.

Most games use a Server-Client architecture. Prediction is anticipation, and Interpolation fills in the gaps, but one crucial part that every developer must keep in mind is: Authority. The server is the only one who should validate data. Depending on how well structured the package load is, this can be faster or slower, opting for chunks or more real-time updates. Highly depends on the gameplay. Turn based games gain nothing from constant updates and can soften the blow by buffering most of the changes. Server-side validation allows for a synchronized common gameplay state between all clients, and ensures no client can make unsupervised changes. These often open up the door to hacks and exploits.

As pointed out by Fritz, server-side validation is key to having authority. This authority is needed to prevent cheating and have a functioning game economy (resulting in revenue for the studio). That's why P2P is barely used by games with LiveOps. What you need to remember is that network latency represents over 60% of the overall cause of "perceived lag" for players. This latency is typically not under the studio's not the player's control. That's why, getting the right visibility on this aspect can help apply surgical changes to your game in order to have a significant impact where it makes the most sense (versus the shotgun approach).

Prediction has a big friend: Determinism. That word only changes your entire infrastructure and tech stack when developing multiplayer. You can do things in completely different ways and techniques depending on whether you have determinism or not. But certainly helps if you have it. Being able to trigger a physics action on both machines and having the exact same result saves millions of messages that try to synchronize every step of the movement in order to match both clients.

Authority these days usually ends up being a server or a server farm, validating messages between clients. Depending on whether your setup is one-to-one, one-to-many or many-to-many and how fast you need your clients to sync, authority can be very hard to tackle or can be a breeze. Going all the way to running an entire game instance on a server to host all clients in to having just a simple Node.js script confirming calculations of physics, authority needs to be at the core of your multiplayer tech design.

I have seen a pretty cool compromise when it came down to a "who shot who" situation where two shots were almost fired at the same split second. It was a 1-1 style game so the game stopped for both players, a coin flip is done and players abide by the result. Funny little gimmick and a tiny-mini-game to increase the excitement. Of course the trade-off was stopping the fast paced action and reducing the heat but it was definitely a fun compromise.

Testing must be huge for multiplayer. I always suggest starting QA right at the moment of starting development just to be able to cover the incredible amount of possible user cases. Relying on internet connection changes many things. It's always great to have an app that can deliberately hinder your connection so that you can test things in different scenarios.

Balancing is not about issues such as bugs, glitches, and player experience problems. It focuses on how about enjoyable and fair gameplay is. Games with more breadth than depth usually pile character over character, each one with similar mechanics but different base values. Balancing these values and the possible outcomes (sometimes as emergent gameplay) is a daunting but rewarding task. This ensures the game remains “true” and doesn’t punish player choice. Now, in the real world: some characters and mechanics are broken, and they’re left that way due to popularity. It’s an entirely different subject, but one more point of view to consider!