35 Reference Tables
Computed from the osu! wiki formulas for the stable client; lazer differs in places. 46,47,48
Hit windows by OD
How far off, early or late, a hit can be and still earn each judgement.
| OD | 300 | 100 | 50 |
|---|---|---|---|
| 5 | ±50 ms | ±100 ms | ±150 ms |
| 7 | ±38 ms | ±84 ms | ±130 ms |
| 8 | ±32 ms | ±76 ms | ±120 ms |
| 9 | ±26 ms | ±68 ms | ±110 ms |
| 10 | ±20 ms | ±60 ms | ±100 ms |
Stable counts a hit only if the error is below the rounded window, so each is effectively half a millisecond narrower (±19.5 ms at OD 10). Hard Rock multiplies OD by 1.4 (maximum 10). Double Time keeps the listed OD, but every window is a third shorter in real time. 46
How long objects stay visible (AR)
| Approach rate | Visible for | With Double Time |
|---|---|---|
| AR 5 | 1,200 ms | 800 ms |
| AR 7 | 900 ms | 600 ms |
| AR 8 | 750 ms | 500 ms |
| AR 9 | 600 ms | 400 ms |
| AR 10 | 450 ms | 300 ms |
Objects fade in over the first two-thirds of this time. Hard Rock multiplies AR by 1.4 (maximum 10); Easy halves it. 47
Mods at a glance
| Mod | What changes | Score multiplier |
|---|---|---|
| Easy | CS, AR, OD, and HP halved | 0.50× |
| No Fail | You can't fail; pp reduced 2% per miss, capped at 10% | 0.50× |
| Half Time | 0.75× speed; hit windows and visible time a third longer | 0.30× |
| Hidden | No approach circles; objects fade out after appearing | 1.06× |
| Hard Rock | CS × 1.3; AR, OD, HP × 1.4 (max 10); map mirrored top to bottom | 1.06× |
| Double Time | 1.5× speed; hit windows and visible time a third shorter | 1.12× |
Lazer uses different multipliers. Relax taps for you and Autopilot aims for you; neither is ranked. 60,62
What mods do to a setting
| Setting | With Hard Rock | With Double Time |
|---|---|---|
| AR 7.2 | AR 10 (7.2 × 1.4, capped) | Visible time of about AR 8.5 |
| AR 8 | AR 10 (11.2, capped) | Effectively AR 9.67 |
| AR 9 | AR 10 (12.6, capped) | Effectively AR 10.33 |
| CS 4 | CS 5.2 (radius 31.1 instead of 36.5) | Unchanged |
| OD 8 | OD 10 (11.2, capped) | 300 window shrinks from ±32 to about ±21 ms in real time |
Circle size and radius
| CS | Radius (osu!pixels) | Note |
|---|---|---|
| 3 | 41.0 | Common on easier maps |
| 4 | 36.5 | The de facto default |
| 5 | 32.0 | Where precision starts to bite |
| 5.2 | 31.1 | CS 4 with Hard Rock |
| 6 | 27.5 | Precision maps |
| 7.8 | 19.5 | CS 6 with Hard Rock: about the smallest seen in tournaments |
| 8 | 18.6 | Half the radius, a quarter of the area, of CS 4 |
Stream speed
| BPM (1/4 streams) | Notes per second | Per finger alternating | Gap between notes |
|---|---|---|---|
| 140 | 9.3 | 4.7 | 107 ms |
| 150 | 10.0 | 5.0 | 100 ms |
| 160 | 10.7 | 5.3 | 94 ms |
| 170 | 11.3 | 5.7 | 88 ms |
| 180 | 12.0 | 6.0 | 83 ms |
| 190 | 12.7 | 6.3 | 79 ms |
| 200 | 13.3 | 6.7 | 75 ms |
| 220 | 14.7 | 7.3 | 68 ms |
| 240 | 16.0 | 8.0 | 63 ms |
| 260 | 17.3 | 8.7 | 58 ms |
| 270 | 18.0 | 9.0 | 56 ms |
A single finger tops out around 5 to 7 taps per second, which is why alternation is not optional above about 150 BPM. 116
Accuracy targets
How many 100s each accuracy allows, assuming no 50s or misses: 63
| Accuracy | 100s per 100 objects |
|---|---|
| 99% | 1.5 |
| 98% | 3 |
| 97% | 4.5 |
| 95% | 7.5 |
| 90% | 15 |
A 50 costs 1.25 times as much accuracy as a 100, and a miss 1.5 times as much.
Worked example: why centering your hits matters
A UR of 100 means a standard deviation of 10 ms. Treating hits as normally distributed (an assumption osu! developers have used when estimating UR), here is how often hits would fall outside the ±32 ms window for a 300 at OD 8: 64
| UR | Outside the window if centered | If your average hit is 10 ms early |
|---|---|---|
| 80 | 0.006% | 0.30% |
| 100 | 0.14% | 1.39% |
| 150 | 3.29% | 7.38% |
At UR 100, a 10 ms early lean produces about ten times as many non-300s as being centered. That gap is what offset, or retraining your timing, closes.
How your top plays are weighted
Each score in your top 100 counts about 5% less than the one above it. 59
| Position | Weight | Meaning |
|---|---|---|
| 1st | 100% | Your best play |
| 10th | 63% | Still most of its value |
| 30th | 23% | The top 30 hold about 79% of the total |
| 50th | 8% | Raises your standard, not your rank |
| 85th | 1.3% | Worth playing for, not for pp |
Polling rates and input delay
Polling adds, on average, half the report interval.
| Report rate | Interval | Average delay | Worst case |
|---|---|---|---|
| 125 Hz | 8.0 ms | 4.0 ms | 8.0 ms |
| 133 Hz (stock on several Wacom tablets) | 7.5 ms | 3.8 ms | 7.5 ms |
| 200 Hz | 5.0 ms | 2.5 ms | 5.0 ms |
| 300 Hz | 3.3 ms | 1.7 ms | 3.3 ms |
| 390 Hz | 2.6 ms | 1.3 ms | 2.6 ms |
| 700 Hz | 1.4 ms | 0.7 ms | 1.4 ms |
| 1,000 Hz | 1.0 ms | 0.5 ms | 1.0 ms |
| 8,000 Hz | 0.12 ms | 0.06 ms | 0.12 ms |
For scale: a 480 Hz monitor draws a frame every 2.1 ms, and a player at UR 100 scatters hits with a 10 ms standard deviation. Fix a 125 or 133 Hz device; past a few hundred Hz you are optimizing below what you can perceive. 33