Skill-based matchmaking, explained by the engineer who built Call of Duty's rating system

There's a system running in the background of every multiplayer game that quietly decides whether you have a good night or a bad one. Most players never see it, most of the people arguing about it online are wrong about how it works — and, as our guest points out, plenty of the studios that depend on it don't understand it much better.

Charlie Olsen would know. He wrote the matchmaking rating (MMR) system that still runs, essentially unchanged, inside Call of Duty. He left Activision in 2020 and founded Invokation, which now sells matchmaking to live games like The Finals and Splitgate. We brought the player-side questions; our co-host Lewis brought the white papers. Charlie didn't dodge any of them.

Skill-based matchmaking doesn't distribute skill

Ask players what SBMM does and they'll say it balances skill. Charlie's framing is sharper: the thing being handed out isn't skill — it's the chance to dominate a lobby. The honest objective is to give everyone a shot at performing well, instead of getting stomped match after match.

Why blowouts actually feel bad

Activision's 2024 white paper found that blowouts correlate with players reporting less fun. Charlie's read is that the blowout is mostly a proxy: what players really care about is their individual performance. Get stomped as a team and you probably had a forgettable game yourself. Play an even match and you've got a real shot at being the best player in the lobby — which is the feeling people are actually chasing.

The churn cascade

Loosen the skill settings and the bottom of the player base leaves first. Then the next tier up becomes the new bottom, and the same thing happens to them. It's how a single tuning change — invisible on a two-week dashboard — can quietly hollow out an entire community over a season. Nearly every major game that has studied it sees the same cascade.

Engagement is not fun

The KPIs studios optimize — quit rates, returning players, kills per minute — are short-term, two-week-measurable proxies for what they actually want: long-term retention and revenue. Charlie's warning is that engagement and fun are not the same thing. Optimizing the proxy while ignoring mounting player sentiment can compound into a problem you won't see until it's already cost you the base.

You can buy matchmaking now

For most of the genre's history, every studio built its own matchmaking. That's changing. A quiet vendor market is opening up — Invokation among them — letting studios buy a system most teams underinvest in and get wrong. For a live-service game where match quality is retention, that's a real shift in the build-versus-buy math.

Listen to the full conversation with Charlie Olsen on the Player Driven podcast.