How osu! Actually Grades Your Maps And Why You Keep Getting B
osu!'s grading system isn't magic. It runs on a straightforward percentile curve based on accuracy, combo, and approach rate, but the way those numbers interact in practice is where most people get tripped up. I've spent years watching players obsess over 99.50% and not realize they're grinding the wrong metric entirely. The algorithm itself is relatively simple on paper. You hit notes, the game tracks your accuracy per note, applies combo multipliers, applies AR/OD sliders, and spits out a grade from SS down to F. The forgiving part comes from the way miss penalties compound. A single miss at 99%+ accuracy tanks your grade dramatically, but missing one note in a stream of 2000 doesn't drop you below B. That's the core tension players navigate every session. I remember mapping a set with a 1900-note slider section where the approach rate was set to 10. The algorithm counts each hit circle and slider end as a separate scoring unit. So a missed slider tail counts the same as a missed tap note. Most people don't realize this until they're staring at a 98.2% with a B grade and trying to figure out why. My workaround was running a custom local replay parser that broke down exactly which objects were costing me the most percentage points per miss. Took me about 40 minutes to identify that three specific slider endings were the problem. Fixed them, grade jumped to 99.1%.
The grade thresholds themselves have shifted over the years. At one point the gap between A and B was extremely narrow at high playcounts. Now with the current scoring model, you generally need around 95% accuracy for an A and 90% for a B, but those numbers vary depending on map difficulty, combo count, and whether you're running on mania or standard mode. The underlying formula uses a combination of per-note accuracy weighted by beat value and a combo-based multiplier that rewards longer uninterrupted chains. Here's something most players overlook. The combo multiplier doesn't actually scale linearly with combo count. It scales with the number of notes in your longest unbroken combo relative to the total note count in the map. This means skipping a single note in a 500-note map hurts you roughly the same as skipping one in a 2000-note map, even though the percentage looks worse in the smaller map. I caught this pattern after running a spreadsheet tracking my misses across different difficulty tiers over six months. The math checks out. Another counter-intuitive detail is how approach rate affects your effective accuracy calculation. Higher AR doesn't just make notes appear faster. It also changes the timing windows that determine what counts as a perfect, good, or miss. At extreme AR values, the window for a "good" hit becomes so narrow that even frame-perfect input registers as a miss depending on your monitor refresh rate and polling setup. I've seen players blame the algorithm when their input was actually lagging by 15 milliseconds due to Windows power management throttling their USB controller. Turned off balanced mode, set the USB selectivesuspend setting to disabled, and their grade consistency improved overnight.
The survival tactics people use generally fall into two camps. Some players focus on maximizing accuracy by slowing their playstyle down, reducing risks, and playing safer patterns. Others chase higher grades by pushing borderline patterns and accepting occasional misses as a calculated tradeoff. Neither approach is inherently wrong. The algorithm treats both identically and produces the same grade range. What actually separates them is fatigue management and consistency over time. If you're trying to climb grade brackets on a specific map, don't just replay it blindly. Look at your per-note breakdown in the replay. Focus on the patterns where your accuracy drops below 97%. Those are your bleed points. Fix one at a time. Trying to improve everything simultaneously usually makes all of them worse because you're overloading your motor memory adaptation window. There are also some scenarios where the grading system behaves in ways that feel unfair but are actually by design. Long maps with many slider ticks and approach circles generate more scoring units, which means your grade has more data points to average out. A short 300-note map with one miss will feel much harsher than a 2500-note map with the same single miss. This is intentional. The game rewards consistency over distance. Players who understand this tend to stop panicking on long maps after 1500 notes and instead focus on maintaining rhythm rather than forcing perfection on difficult sections.
Get the Full Details
One thing I'll say bluntly about the current grading algorithm. It doesn't account for musicality or flow. Two players can hit the exact same notes with the same accuracy and receive identical grades even if one played the map with noticeably better timing to the beat. The system only sees the timing windows, not the auditory alignment. This means you can technically farm grades on poorly timed maps without actually improving your feel for the music. I've watched people stack SS grades on 10+ star maps that had completely misaligned hit sounds and then struggle on properly timed same-difficulty maps. The algorithm won't tell you that your muscle memory is built on bad audio cues. For people who want to dig into this more, there are community tools like osu!direct replays, local scoring analyzers, and third-party grade calculators that let you see the per-note breakdown before you even play. Using those during practice saves a lot of wasted attempts. I typically run a quick analyzer on any map I'm targeting for a specific grade tier before I start pulling rank on it. Cuts down the grind from three or four hours of blind replaying to maybe 45 minutes of focused work on the actual problem patterns. The osu! grade system has its quirks and its blind spots. It's predictable once you know how the math works under the hood. The algorithm will always give you exactly what it calculates based on your inputs. What it won't do is reward effort that isn't reflected in the raw numbers. Play smarter, not harder, and track your actual hit distributions instead of just staring at the final percentage.
