What drift means
A rank is meant to describe how good a player is compared with others. Inflation is when the same level of play ends up with a higher label over time; deflation is the reverse. Nobody has to improve or decline for it to happen. The labels themselves slide, and an apparent increase in the number of high-ranked players may reflect the system rather than the population.
This matters because a rank is only useful if it means roughly the same thing from week to week. When it drifts, comparing yourself with last month, last season or other players becomes unreliable.
The conserved-points case
Start with the clean example. In a standard Elo ratings system with a closed group of players, every point a winner gains is a point a loser loses. The total of all ratings is constant, and so is the average. In that setting, no inflation is possible: if some players climb, others must fall by the same total.
Drift appears when this balance is broken. Most real ladders break it in several ways.
Common causes of inflation
Gains larger than losses. Some games award more for a win than they subtract for a loss, to feel rewarding or to nudge players upward. Each match then creates points rather than just moving them, so the average rating rises over time. This can be deliberate: designers sometimes prefer a ladder where most people feel they are making progress.
Loss protection. Features that stop a rank dropping, such as demotion shields or a buffer below a division boundary, mean losses are partly forgiven. Wins still count in full, so the total tilts upward.
Bonuses. Extra points for streaks, for the first win of the day or for playing at certain times add rating that does not come from another player’s loss.
Players who leave. People who quit tend to stop at different points than they started. If many players leave while sitting at a low rating, their points leave with them, and the remaining pool’s average may shift. In particular, when weaker players quit more readily than stronger ones, the average of the remaining pool rises, even though no individual has improved. The effect depends on who leaves, so it can run in either direction.
New players entering low. Newcomers who start at a default and then climb bring points in as they improve, which pushes the pool upward if they gain more than they were given at the start. If new players generally start above their true level, the opposite happens: they lose points to established players and deflate the lower end of the ladder.
Resets that do not preserve the mean. At the end of a season, many games compress ratings toward a central value. Depending on how the compression is done, the average may shift. This overlaps with the topic of seasonal resets, but the point here is simply that the reset rule itself can add or remove points from the pool.
Common causes of deflation
Deflation is less often intended but arises in the mirror image of the above.
- Losses larger than gains, for example in systems with decay or a point tax meant to keep players active.
- Inactivity decay, where rating drops for players who do not play. This removes points from the pool and can pull the average down, although the players it affects are mostly the ones not competing.
- Strong players dominating a thin pool, where losers keep falling to opponents who then never lose rating because few players can challenge them.
- A sudden inflow of strong players through smurf accounts or a popular new release, which takes points from existing players without their skill changing.
Why true skill can stay the same
Underlying skill is separate from the rating labels. A rating is only a position on a scale defined by the other players. If the whole population gains or loses 100 points, everyone’s relative standing is unchanged, and no one is better or worse than before. A player moving from “Gold” to “Platinum” without improving is mostly a statement about where the boundaries sit relative to the pool.
One exception is genuine change in the population’s skill, such as the community collectively learning a game, which is not drift but real improvement, though it can look the same from the outside. Telling the two apart is hard, because both show up as the same label moving up.
Why designers sometimes accept it
A perfectly stable ladder can feel stagnant. If every win is exactly offset by someone’s loss, the average player spends most of the time near where they started. Designers worried about motivation sometimes allow gentle inflation so that progress feels achievable. The downside is that high ranks lose their meaning as more players reach them, and the top of the ladder can become crowded.
Others hold the mean fixed on purpose, using approaches that renormalise the distribution each season or assign ranks by percentile, for example “top 5 percent” rather than a fixed rating value. Percentile ranks are immune to drift in the labels, since they are defined by position, though they bring their own issues, such as someone dropping in rank because others improved.
Reading ranks sensibly
A few habits protect you from misreading drift.
- Compare within the same season, where the scale is consistent, more than across seasons.
- Look at percentile or distribution information where the game provides it, which describes standing more stably than the label.
- Be cautious about celebrating or lamenting a number change on its own. Without knowing what the pool did, you cannot say whether you moved.
- Watch for rating volatility around boundaries, where small changes cross a tier threshold and look bigger than they are.
Drift is not necessarily a flaw. It follows from decisions about incentives, new-player handling and resets. The useful skill is knowing it exists, so a change in your label is not automatically read as a change in your play.