A rating is an estimate, not a measurement
A ranked ladder tries to answer a question nobody can observe directly: how good is this player? You cannot read skill off a player the way you read height off a ruler. All a ladder sees is match results. A skill rating is therefore an estimate built from those results, and every design choice in the system is about how to turn a pile of wins and losses into a single number that is useful.
Seen this way, a rating is a summary of evidence. Each result is a small piece of information about the players involved, and the system keeps a running total. That running total is what the ladder displays, sometimes directly as a number and sometimes hidden behind tiers, divisions and badges.
The core loop: predict, observe, adjust
Almost every rating system works on the same loop.
- Before a match, use the two sides’ current ratings to predict how likely each is to win.
- Play the match and record the result.
- Compare the result to the prediction and move the ratings by an amount that depends on how surprising the result was.
The third step is what makes the system self-correcting. If a strong player beats a weak one, that was expected, so the ratings barely change. If the weak player wins, the result carried more information, so the ratings move further. Over many games, ratings drift toward values that predict results well.
This also means the number is only meaningful relative to other players. A rating of 1500 has no absolute meaning; it matters only because of how it compares with the ratings around it. Simple schemes such as Elo ratings use a fixed formula for the prediction and the adjustment. Others add extra information, such as rating uncertainty, so the system can say not just how good a player looks but how sure it is.
What the number can tell you
Used sensibly, a rating is good at a few things.
- Ordering players. On a well-populated ladder, a higher rating usually means a player has performed better against comparable opposition.
- Predicting results. The central test of a rating system is whether its predicted win chances roughly match what happens. A good system is calibrated: among games where it predicts a 70 percent chance, the favourite wins about 70 percent of the time, within the usual noise.
- Setting up fairer games. Ratings let matchmaking pair players who should have competitive games, which is the main practical reason ladders exist.
What the number cannot tell you
The limits matter as much as the uses.
It reflects results, not reasons. A rating cannot tell you whether someone wins through mechanics, game knowledge, communication or simply playing a strong character or role. Two players with the same number can be strong in entirely different ways. Understanding why results happen takes tools like replay review rather than a ladder position.
It is noisy over short stretches. A handful of games can mislead, because luck plays some part in most games. Anyone judging a change in their own level should keep sample size in mind: a ten-game run says much less than a hundred-game run. The broader problem of separating luck from skill applies directly to reading ladders, and games with more randomness make ratings noisier.
It depends on who you play. A rating is built from games against a particular pool. Move between regions, queues or game modes and the number may not carry over. Ratings can also vary in meaning between seasons, since resets and population changes shift the whole ladder, as covered under seasonal resets.
It assumes a single skill. One number treats a player as having one level. In reality, skill can depend on matchup, map, role or character. A player might be well suited to one style of play and weaker against another, and a single number averages that away. Related ideas show up in how head-to-head records can differ from overall standings, since a player may consistently beat someone who outranks them.
Team games blur individual contribution. When the result belongs to a group, the system must guess how much credit each player deserves. Most simply assume a share based on the team’s total. That is workable over many games but unreliable for any single match, and it is the subject of team ratings in its own right.
Ratings versus ranks and tiers
Many games show a tier or a division rather than the raw number. Tiers are bands on the underlying scale, often with promotion and demotion rules layered on top. They are easier to read and more motivating, but they hide detail: two players at the top and bottom of the same tier may differ more than two players on either side of a boundary.
Some ladders also keep a hidden rating for matchmaking separate from the visible rank. The two can disagree for a while, which is one reason a displayed rank can feel out of step with how games are going.
Where ratings go wrong
A few recurring problems are worth knowing about, because they explain many complaints about ladders.
- New players have no history, so their first ratings are guesses, handled through placement matches.
- Experienced players on new accounts can distort a bracket until the system catches up, which is the smurf accounts problem.
- The whole ladder can drift upward or downward without anyone changing, the effect described as rank inflation.
- Short-run swings are normal and often say little about real change; rating volatility covers how to read them.
- Faster matches usually mean looser matches, because of the queue time tradeoffs every matchmaker faces.
How to read your own rating
A few habits help. Treat the number as an estimate with a margin around it rather than a verdict. Compare over long stretches rather than single sessions. Remember that a rating going up can mean you improved, but it can also mean you were lucky, played a favourable pool of opponents, or that the ladder itself shifted. And since results and decisions are different things, it helps to judge your play by process, not by the last result alone.
None of this makes ratings useless. They remain the best simple tool ladders have for sorting a large population into reasonably even games. The sensible stance is to use them for what they do well and not to ask them to settle questions they were never built to answer.