← osu! guides

osu! star rating explained: how map difficulty is measured

Star rating is osu!'s automatic estimate of how difficult a beatmap is, shown as a number of stars and recalculated by the game whenever a mod changes the map. For osu!standard it is built from separate skill measurements — aim, speed, and a reading component covering visual density, plus a flashlight component when the Flashlight mod is active — which are combined and then compressed through a cube-root transform. Because pp also depends on accuracy, combo, misses and length, a higher star rating does not automatically mean more pp. Star rating is a model rather than a fact, and it is periodically reworked; the most recent osu! update landed on 3 July 2026.

What star rating measures, and what it does not

Star rating is a single number the osu! client computes from a beatmap's hit objects and difficulty settings. It estimates how hard the map is to play well, and it is the main difficulty input to the pp formula. Every ranked map has one, and it is recomputed for the modded version of the map whenever you apply a mod that alters the map.

What it is not: a measure of how hard the map is for you specifically, a measure of how hard it is to pass, or a prediction of how much pp you will earn. Star rating describes the beatmap. Passing difficulty and pp both depend heavily on the player. A 6★ map made of long, dense streams and a 6★ map built from a handful of enormous jumps demand entirely different skills, and most players will find one far harder than the other despite the identical figure.

Difficulty names and star bands

The osu! wiki assigns each star range a difficulty classification, which drives how a difficulty is presented in the beatmap listing.

Two caveats matter more than the bands themselves. A mapper's difficulty name is free text and need not match the band at all — a difficulty called "Insane" can sit anywhere on the scale, and older maps frequently do, because naming conventions drifted over the years while the bands did not. And mods move a map across bands: a 3.8★ Hard becomes an Insane or an Expert once Double Time is applied, because the rating is recalculated on the sped-up map rather than carried over from the original.

The bands are therefore best read as a rough shelf label for browsing, not as a statement about what a specific map will demand from you. Two maps in the same band, or even at an identical rating, can be separated by an enormous gap in how playable they feel.

The skill components behind the number

The osu!standard calculator does not treat difficulty as one quantity. It runs several skill evaluators over the map's object sequence:

• Aim — cursor movement demands. It is evaluated twice, once counting slider movement and once with slider travel stripped out, so the calculator can tell how much of a map's aim difficulty depends on sliders.
• Speed — tapping demands, driven by how closely notes are spaced in time.
• Reading — how hard the map is to process visually, scaling with note density and with pattern angles. This is where Hidden is now accounted for.
• Flashlight — created only when the Flashlight mod is active, and folded together with reading into a combined cognition value.

Each evaluator works on strain: it walks the map in order, accumulates difficulty as objects arrive quickly, and lets that value decay through gaps. The output therefore reflects a map's hardest sustained sections rather than its average across the whole length.

From skill values to a single star figure

The per-skill values are converted into performance-space numbers and then combined with a p-norm rather than a plain sum. A p-norm sits between "take the largest" and "add everything up": a map that is extreme in one skill lands close to that skill's value on its own, while a map strong in two skills lands above either but below their total. That is how a pure-jump map and a balanced technical map can reach the same star rating by completely different routes.

The combined value is then passed through a cube root to produce the star figure. That cube root is why star rating compresses badly at the top end. The gap from 4★ to 5★ and the gap from 8★ to 9★ are both one star, but in the underlying difficulty space the second is far larger. Treating stars as a linear scale of effort will mislead you at every level above roughly 6★.

How mods change star rating

Mods that change the map change the star rating, because the calculator is simply run again on the modded map.

• Double Time and Nightcore — playback at 1.5x compresses every interval in the map, raising both aim and speed values. This is usually the largest star rating jump any single mod produces.
• Half Time — 0.75x playback, the same mechanism running backwards.
• Hard Rock — circle size up 30%, and approach rate, overall difficulty and HP drain up 40% each. Smaller circles raise aim difficulty; the faster approach raises reading difficulty.
• Easy — halves those same four stats, lowering the rating.
• Flashlight — adds a flashlight component that does not otherwise exist, so it raises the rating on every map, and more on long ones.
• Hidden — since the July 2026 update it feeds the Reading skill and therefore moves the star rating. Under the previous model it did not.

The mods that leave star rating untouched

Several mods change your play without changing the map's measured difficulty at all. No Fail, Sudden Death, Perfect and Spun Out leave star rating exactly where it was. They alter your risk, your failure condition, or who spins the spinners — none of which is a property of the beatmap's object sequence.

This matters when reading a leaderboard. A play listed as NFHDDT sits on a map whose star rating is the HDDT rating; No Fail contributes nothing to that number and instead takes up to 10% off the pp. Spun Out likewise does not lower the star rating of a spinner-heavy map — it applies its reduction at the pp stage.

The practical rule is straightforward. If a mod changes what the objects are, how fast they arrive, or what you can see of them, it changes star rating. If it only changes what happens when you fail, or automates part of your input, it does not.

Why star rating and pp diverge

Star rating is one input to pp. Your accuracy, max combo, miss count and the map's object count are the others. Two consequences follow.

First, a clean play on a lower-rated map routinely beats a broken play on a higher-rated one. pp climbs steeply with accuracy near the top of the range and is penalised for misses, so a full combo at 99% on a 6.5★ map can outvalue a 96% run with several misses at 7.2★. Chasing star rating alone is a reliable way to accumulate plays that never enter your top 100.

Second, star rating says nothing about your profile total. Your pp is a weighted sum of your top 100, with the n-th best play weighted at 0.95 to the power (n−1). A 500pp play landing at rank 10 contributes 500 × 0.6302 ≈ 315pp of weighted value, and every play below it shifts down a place and loses 5% of its own weight, so the net gain is smaller still.

Bonus pp, and why volume is not difficulty

A separate term is added on top of the weighted top-100 sum: bonus pp, awarded for the number of ranked beatmaps you have a score on. The osu! wiki gives it as 416.6667 × (1 − 0.995 to the power of min(N, 1000)), where N is that map count. It is capped at 1,000 maps, worth 413.894 bonus pp.

The curve is steep early and almost flat later: the first hundred maps are worth far more per map than the ninth hundred, and past 1,000 maps the term stops growing entirely. This is the one part of your pp total that responds to breadth rather than difficulty, and it is why play count and play time are volume metrics rather than skill metrics.

It also explains a common source of confusion. A newer player who spreads their play across many different maps can appear to gain pp without setting anything impressive, because the bonus term is still climbing. An established player with thousands of maps played has exhausted it completely, so every pp they gain from here must come from displacing a play in their top 100.

Where star rating is a poor guide

Star rating is a model fitted to how people actually play, and it has well-known weak spots.

• Patterns the evaluators under-read. Awkward rhythms, irregular sliders and heavy technical mapping are widely reported by players as rating below how hard they feel, because the strain evaluators are built around spacing and timing rather than coordination cost.
• Low approach rate. Maps that are difficult purely because objects arrive with little warning were under-rated for years. The Reading skill introduced in July 2026 exists partly to address exactly this.
• Length and stamina. The rating is dominated by a map's hardest sustained sections, so a two-minute map and an eight-minute map with the same peak can rate similarly while differing enormously in stamina cost.
• It is not stable over time. Reworks reprice existing plays. Nothing about anyone's play changed on 3 July 2026, but the star rating and pp value of a great many plays did.

What changed in the July 2026 rework

The update of 3 July 2026 revised star rating and pp. For osu!standard the headline changes were:

• A new Reading skill replaced the previous static approach-rate and Hidden bonuses, calculating note density at each object and scaling reading difficulty from it, with angle-based adjustments on top.
• Speed moved from geometric to harmonic summation, so every note contributes to the final speed value and no separate length bonus is required.
• Aim gained a separate flow evaluator that distinguishes snap aim from flow aim using velocity and angle change rather than distance alone.
• Speed now scales off tapping deviation rather than raw accuracy, correcting inequalities on high overall-difficulty scores.
• Repetitive acute-angled jumps were nerfed.

Using star rating on your own top plays

Star rating becomes personally useful once you can see which ratings actually produce your top plays. Start a TrackMyOsu session, play a set of maps inside a narrow star band, then stop it. The saved session lists every play that entered your top 100 while it was running, alongside your pp and rank at both ends, so you can see whether the 6.5★ hour or the 7★ hour actually paid.

Sessions are kept permanently, so the comparison sharpens over months rather than resetting each time you look. The dashboard refreshes every 60 seconds during a live session, new medals are recorded alongside new top plays, and your public profile at /u/<username> carries the same history. The pp-gains leaderboard and Hall of Fame show the same measurements across other tracked players, which is a useful check on whether your own star-band habits are unusual.

By carrot12_. Last updated 2026-08-12.