Trofeo· The Trofeo team
Elo vs TrueSkill vs Glicko-2: which rating system fits your game?
Three rating systems, three different problems they solve. A plain-language guide to picking the right one for ping pong, foosball, chess, or whatever your office actually plays.
Elo is the famous one. It runs chess federations, it's a verb ("he's underrated, his Elo will climb"), and it's almost certainly what you picture when you think "leaderboard rating." But Elo makes two quiet assumptions: that every game is 1v1, and that your rating is equally trustworthy whether you played 200 matches or two.
Most offices don't play only 1v1. Foosball is doubles. Mario Kart is four players on a couch. Chess is full of draws. The moment your games break those assumptions, you want a system built for them - and two of Elo's descendants are. Here's how to choose.
Elo: the 1v1 workhorse
Elo gives every player one number (new players start at 1200) and moves points between winner and loser after each match, sized by how surprising the result was. Beat someone you were supposed to beat and you gain a little; upset someone 400 points above you and you gain a lot. The points the winner gains, the loser loses - it's zero-sum, self-correcting, and it converges to a meaningful number after 10–15 games.
It's the right call when your game is one-on-one and results are win/loss. Ping pong singles, pool, darts, 1v1 FIFA - Elo is hard to beat for these. It's simple enough that everyone trusts it, and "I beat the person above me, I went up" is exactly the story you want. (If you want the mechanics, here's Elo explained with office ping pong.)
Use it for: ping pong, pool, darts, 1v1 FIFA - any clean 1v1.
TrueSkill: built for teams and free-for-alls
TrueSkill was built at Microsoft to matchmake Xbox players, and it solves the problem Elo can't touch: rating individuals from team results. It models each player not as a single number but as a range - a skill estimate plus an uncertainty band - and after a 2v2 match it nudges all four players, using who partnered whom and how strong everyone was.
That's what you want for foosball and padel, where the eternal question is "who's actually carrying?" TrueSkill answers it over time: win with a weak partner against a strong pair and your rating climbs faster than if you'd cruised alongside the office champion. It also handles free-for-alls - record the full finishing order of a Mario Kart race and everyone's rating updates based on who they beat and who beat them.
Use it for: foosball, padel, 2v2 anything, and multiplayer races like Mario Kart.
For the full picture - how credit gets split across a team, and why rotating partners helps - see how TrueSkill rates team games.
Glicko-2: Elo with a confidence dial
Glicko-2 is Elo's modern successor, and it's the system behind the major online chess sites. It keeps Elo's core idea and adds one crucial thing: a reliability measure (called RD, rating deviation). A new or returning player's rating is treated as uncertain, so it moves fast; once they've played a steady stream of games, the rating locks in and stops swinging on every result. Sit out for a month and the uncertainty widens again - the system admits it's no longer sure where you stand.
It also treats draws as first-class results, which matters enormously for chess, where a draw against a stronger player is a small triumph and the rating should reflect that. Plain Elo can score draws, but Glicko-2's uncertainty-aware updates make them land more honestly.
Use it for: chess, and any 1v1 game where draws are common and rating trust matters.
We go deeper on the confidence band and proper draw handling in Glicko-2 and the problem of draws.
The one-table summary
| System | Best for | Handles draws | Team play | Tracks uncertainty |
|---|---|---|---|---|
| Elo | Clean 1v1 | Yes (basic) | No | No |
| TrueSkill | Teams & free-for-alls | Yes | Yes | Yes |
| Glicko-2 | Chess-style 1v1 | Yes (first-class) | No | Yes |
Pick by game
If you just want the answer for what's on your office table:
- Ping pong, pool, darts, 1v1 FIFA → Elo. Simple, fast, trusted.
- Foosball, padel, 2v2 co-op → TrueSkill. It untangles individual skill from rotating partners.
- Mario Kart and other races → TrueSkill, using the full finishing order - finishing 2nd of 8 should beat winning a duel.
- Chess → Glicko-2. Proper draws, proper handling of rusty returners.
Trofeo lets you set the system per game, so your ping pong ladder can run Elo while foosball runs TrueSkill and the chess club runs Glicko-2 - all in the same workspace, all posting results to the same Slack channel.
The honest caveats
No rating system is magic, and knowing the edges keeps you from over-trusting the number:
- None of them model form. The system knows your rating, not that you slept four hours or that someone swapped the ping pong balls. Day-to-day variance is real and it's supposed to be - it's why favorites lose sometimes.
- Small samples are sketches, not portraits. A rating off six games is a guess. Glicko-2 and TrueSkill are honest about this (their uncertainty bands stay wide early); Elo just looks confident sooner than it should. Treat early-season numbers loosely.
- Switching systems mid-league resets the story. Each system speaks its own scale - TrueSkill's numbers aren't Elo's. Pick per game before you start, not three weeks in.
The one-sentence version
Elo if it's 1v1 and simple, TrueSkill if there are teams or a finishing order, Glicko-2 if it's chess-shaped and full of draws - and the only real mistake is forcing one system onto a game it was never built for.
Pick the right engine per game, and the leaderboard stops being a tally and starts telling the truth.