The starting problem
A new account has no results, so a ladder has no evidence about it. Every rating system needs some answer to the question of where to put a player who has never played. The options are to start everyone at the same default and let results sort things out, or to run a short block of games, often called placement matches, specifically to estimate a starting level before showing a rank.
Option one: the default start
The simplest approach is to give every new player the same rating, usually near the middle of the scale. This is easy and transparent. The downside is that the default is wrong for almost everyone. A beginner placed at the middle will lose a lot before reaching their level, and an experienced player will win a lot, in both cases dragging other people’s games off balance in the meantime. Under plain Elo ratings, this correction can take many games, especially with a small K-factor.
Option two: placement matches
Placement matches try to fix this by compressing the correction into a fixed number of games, commonly somewhere around five to ten, though it varies widely by game. During placements, the system treats the player’s rating as very uncertain and moves it by large steps. At the end, it assigns a visible rank.
The technical idea is that the player begins with high uncertainty. Systems that track a confidence value, as described in the discussion of rating uncertainty, allow large movements early on and smaller ones later. Even without explicit uncertainty tracking, a big temporary K-factor achieves a similar effect.
Some games also use hidden information to give the first guess a head start, such as:
- The player’s rank from a previous season or a related mode.
- An account level or hours played, as a rough proxy for experience.
- A rating carried over from another region or platform.
- Results against opponents of known strength, since beating strong players is more informative than beating weak ones.
How reliable is the guess?
Honest answer: only moderately. The reasons are statistical.
Few games give a wide margin. A single result is mostly a win or a loss, which carries limited information. With around ten games, the number of wins you would expect from a given skill level naturally varies a good deal from one run to the next, simply through chance. Two players of identical skill can easily finish placements with noticeably different records. This is the same issue behind any small sample size problem, and it applies to ratings as much as to anything else.
Opponent quality is uncertain too. Placements often match new players against others with uncertain ratings, so the evidence for each result is weaker than it looks. A win against another newcomer tells you less than a win against an established player.
Form on the day matters. Fatigue, a bad stretch, learning a new character or role, or playing while still getting used to controls can all lower early results. Some players also perform differently once the stakes feel real. The effects of pressure are part of the mental game, and placements can feel high-stakes.
Team games add noise. In team modes, a placement record depends on teammates the player cannot choose, so it reflects the group as much as the individual.
The upshot is that placements usually get a player into a rough neighbourhood, close enough to start playing reasonably matched games, but the exact starting position is often off by some amount in either direction.
Why starting conservative is common
Many systems deliberately place new players below where their results suggest. There are a few reasons.
- Placing someone too high produces many losses and a discouraging first experience, while placing them slightly low produces wins that feel good.
- A new player who is placed too low will climb fast, because the system is still uncertain and moves them in large steps.
- Conservative placement protects higher brackets from being filled with players who have only had a lucky ten games.
The cost is that strong newcomers spend their first stretch beating opponents who are below them, until the ladder corrects. This is one source of the mismatch problems that smurf accounts exploit, since a skilled player on a fresh account will typically be placed low and then climb quickly.
Correcting a wrong start
A good rating system expects placement to be imperfect and builds in a way to fix it. After placements, uncertainty usually stays elevated for a while, so ratings keep moving quickly for a number of games. Players who were placed too low rise fast; those placed too high fall back. Within a few dozen games, the effect of the initial guess is usually much reduced, although the exact timing depends on the system.
What this means for players
A few practical points follow.
- Treat your first rank as provisional. It is a starting estimate, not a verdict on your ability.
- Do not over-read the placement record. A weak run of ten games does not mean you are weak, and a strong one does not mean you are strong.
- Expect movement afterwards. Rapid rises or drops in the first weeks usually reflect the system correcting itself.
- Judge progress over more games. Long-run results tell you much more than the opening block, and a steady log of your results, as suggested by stat tracking, makes the pattern easier to see.
Placement matches are a reasonable compromise between speed and accuracy. They are a quick way to get a new player into roughly the right games, not a precise measurement, and the ladder’s real assessment comes from the games that follow.