If you're building a ranked ladder for your gaming community, the first real decision you'll make is which rating system to use. Elo, Glicko-2, and TrueSkill all answer the same question — "how good is this player relative to everyone else?" — but they make very different trade-offs along the way.
This guide breaks down how each system actually works, where each one shines, and how to pick the right one for the size and pace of your community. No PhD required.
The Core Problem Every Rating System Solves
Every rating system is trying to do two things at once:
- Estimate skill from a series of wins and losses against opponents of varying strength.
- Update that estimate efficiently as new results come in, without overreacting to a single lucky game or underreacting to a genuine improvement.
The differences between Elo, Glicko, and TrueSkill come down to how much extra information they track about each player and how confident they are in the current estimate. That confidence — usually called uncertainty or deviation — is the single biggest dividing line between the three.
Elo: The Baseline Everyone Knows
Elo is the system most people picture when they hear "ranked." It's the one chess has used since the 1960s, and it's the foundation under most community ladders. If you want the full mechanical breakdown, see our deep dive on what an Elo rating actually is.
The short version: every player has a single number. Before a match, the system computes an expected result from the rating gap. After the match, it nudges each rating toward reality:
New Rating = Old Rating + K × (Actual Score − Expected Score)
The K-factor controls how aggressively ratings move. That's the only dial Elo gives you, and it's a blunt one — a K-factor that's responsive enough for a brand-new player is too jumpy for a 500-game veteran.
Strengths
- Dead simple to explain. Players can see exactly why they gained or lost points.
- Transparent and predictable — no hidden variables.
- Battle-tested across six decades and dozens of domains.
Weaknesses
- A single number can't express how sure the system is. A new player and a veteran with the same rating are treated identically, even though one estimate is a wild guess and the other is rock-solid.
- No native concept of inactivity. A player who hasn't played in a year keeps the rating they had, even though their real skill may have drifted.
- Team and free-for-all support has to be bolted on with averaging tricks.
Elo is the right default for most communities — especially ones that play 1v1 or fixed-team formats and value transparency over statistical sophistication.
Glicko-2: Elo With a Confidence Interval
Glicko (and its successor Glicko-2) was created by Mark Glickman specifically to fix Elo's biggest blind spot: it doesn't know how reliable its own estimate is. Glicko adds two extra numbers to every player:
- Rating Deviation (RD) — how uncertain the system is about the rating. A high RD means "we're not sure"; a low RD means "we've seen enough games to trust this."
- Volatility (σ) — how erratic the player's results have been. A player whose performance swings wildly gets a higher volatility, which lets their rating move faster.
The key behavior: RD grows when a player is inactive and shrinks when they play. So a returning player's rating can move quickly to catch up to their real skill, while a grinder's rating stays stable because the system is confident. This is why Chess.com and Lichess both run on Glicko variants.
Strengths
- Handles inactivity gracefully — exactly the problem Elo ignores.
- Faster, fairer convergence for new and returning players.
- Still relatively transparent — RD is an intuitive "confidence" number.
Weaknesses
- More complex to explain to players ("why did my rating jump 60 points?").
- Designed around rating periods (batches of games), which doesn't map perfectly onto a community that records matches one at a time.
- Like Elo, team support needs adaptation.
Glicko-2 is a strong fit for communities with intermittent activity — players who come and go, seasonal ladders, or anywhere a stale rating would be a real problem.
TrueSkill: Built for Team Matchmaking
TrueSkill is Microsoft's Bayesian system, designed from the ground up for Xbox Live matchmaking. Instead of one number, it models each player's skill as a bell curve with a mean (μ) — the estimated skill — and a standard deviation (σ) — the uncertainty around it. Your displayed rating is usually a conservative estimate like μ − 3σ, so new players start low and climb as the system gets confident.
The headline feature is native team support. TrueSkill can take the result of an 8-player, 4v4 match and correctly distribute credit, factoring in each player's uncertainty — something Elo and Glicko can only approximate by averaging.
Strengths
- First-class handling of team games and arbitrary team sizes.
- Fast convergence — it often nails a player's skill in far fewer games than Elo.
- Models individual uncertainty even inside a team result.
Weaknesses
- The most complex of the three. The math is genuinely hard to explain, and the displayed number (
μ − 3σ) is unintuitive. - Patented in some implementations; the open variants vary in quality.
- Overkill for 1v1-heavy communities where Elo would do fine.
TrueSkill is the right pick when your community is predominantly team-based and you care more about matchmaking quality than about players understanding their exact rating math.
Side-by-Side Comparison
| Elo | Glicko-2 | TrueSkill | |
|---|---|---|---|
| Numbers tracked per player | 1 (rating) | 3 (rating, RD, volatility) | 2 (μ, σ) |
| Models uncertainty? | No | Yes (RD) | Yes (σ) |
| Handles inactivity? | No | Yes | Partially |
| Native team support? | No (averaged) | No (averaged) | Yes |
| Convergence speed | Moderate | Fast | Fastest |
| Easy to explain? | Very | Moderate | Hard |
| Best for | 1v1 / fixed teams, transparency | Intermittent activity | Team matchmaking |
How to Actually Choose
You don't need to overthink this. Use these rules of thumb:
- Small or transparent community, players want to understand their rank? Use Elo with a K-factor around 32. Add placement matches so new players don't appear on the leaderboard with one game played.
- Players come and go, or you run seasons? Lean toward Glicko-style behavior — or use Elo with a higher K-factor for low-game-count players plus a rating decay policy that nudges inactive players down.
- Mostly team games and you care about match balance? TrueSkill-style systems give the best matchmaking, but for most Discord communities, Elo with team averaging plus skill-based matchmaking queues gets you 90% of the benefit with a fraction of the confusion.
For the overwhelming majority of gaming communities, Elo is the correct starting point. It's transparent, well-understood, and flexible enough to cover 1v1, fixed teams, and free-for-all once you add the standard extensions. You can always layer on uncertainty-aware behavior — placement matches, decay, format-specific ratings — without throwing out the simple core that players already trust.
Doing This in Practice
You almost certainly shouldn't implement any of these from scratch. The math is the easy part; the hard part is recording matches reliably, handling disputes, updating a live leaderboard, and assigning roles based on rank.
Team Up is a Discord bot that runs an Elo system for you with the practical extensions that matter — adjustable K-factor, placement matches, rating floors, multiple rating types (global, per-format, per-character), and tier roles that update automatically. You can stand up a ladder in a few minutes; see the setup guide to get started.
Want to feel out how Elo math responds to different rating gaps and K-factors before committing? Plug numbers into the free Elo Calculator.
Frequently Asked Questions
Is Glicko better than Elo?
Glicko is more sophisticated than Elo, not strictly better. It handles uncertainty and inactivity that Elo ignores, which matters for communities with irregular activity. But Elo's simplicity and transparency are real advantages — players trust a system they can understand. For most communities, well-configured Elo (with placement matches and rating decay) closes most of the gap.
Does Team Up use Elo, Glicko, or TrueSkill?
Team Up is built on the Elo system, with the practical extensions communities actually need: adjustable K-factor, placement matches, rating floors, free-for-all support, and rating decay for inactive players. This gives you Elo's transparency while covering the inactivity and convergence problems that push some communities toward Glicko.
What rating system do League of Legends and Valorant use?
Both use internal matchmaking ratings (MMR) that are proprietary and not publicly documented. They're broadly Elo-inspired but customized heavily. The visible ranks (Gold, Diamond, etc.) are tiers layered on top of the hidden MMR, not the raw rating itself.
How many games before a rating is accurate?
With Elo and a K-factor of 32, ratings typically stabilize after 15–30 games. TrueSkill often converges faster (10–15 games) because it models uncertainty directly. This is exactly why placement matches exist — they hide a rating until the system has seen enough games to trust it.
Further reading:
- What Is Elo Rating? How Competitive Ranking Works in Gaming — the mechanics behind Elo
- Elo Rating Calculator Guide — every calculator parameter explained
- Tier Roles & Rank Decay — keep a leaderboard fair and active
- Rating Types — global, per-format, and per-character ratings
- Elo Calculator — test rating math with your own numbers