The Sharpness Score Algorithm: How We Built a Personal Cognitive Baseline

Why a Personal Baseline Changes Everything

Most cognitive apps give you a score, a rank, or a badge. You finished in the top 30%. You scored 850 points. You're on a 14-day streak. These numbers feel motivating, but they tell you almost nothing about your brain's actual performance trajectory.

The problem is that population-relative scores flatten individual variation into a meaningless average. A percentile ranking tells you where you stand against other users — but other users aren't you. They have different ages, different education levels, different baseline processing speeds. Comparing your Tuesday morning arithmetic performance to a global pool of anonymous strangers generates a number, but not a useful one.

The Sharpness Score takes a fundamentally different approach. Instead of ranking you against a population, it compares you to yourself — specifically, to your own rolling performance baseline. A Sharpness Score of +5.2% means you solved problems 5.2% faster than your recent norm. A score of −8.1% means you were measurably slower. The reference point is always your own history, which makes the number immediately actionable.

The Rolling Baseline: How It Works

At its core, the algorithm maintains a rolling window of your recent performance data, calculated separately for each of the four arithmetic operations: addition, subtraction, multiplication, and division. This per-operation separation matters because your brain doesn't handle all four equally. Most people have faster retrieval times for addition facts than division facts, and the gap can be substantial. Lumping them together would mask meaningful variation.

For each operation, the baseline is computed from a rolling window of your recent sessions. The algorithm uses a hybrid of median and mean calculations, with outlier filtering to prevent a single unusually fast or slow session from distorting the baseline. If you had a terrible night's sleep and your multiplication speed dropped 30% one morning, that single data point won't drag your baseline down — it'll show up as a temporary dip in your Sharpness Score, which is exactly what you want to see.

The baseline adapts as you improve. If your arithmetic speed increases steadily over six weeks of daily practice, the rolling window shifts upward with you. This means a Sharpness Score of 0% — right at your baseline — represents a progressively higher level of absolute performance over time. You're not chasing an arbitrary target. The bar rises with you.

The best personal metric isn't one that compares you to the world. It's one that compares you to yourself yesterday, last week, and last month.

Per-Operation Independence

Separating baselines by operation isn't just a technical detail — it reflects how working memory and mental acuity actually function. The cognitive processes involved in retrieving 7 + 8 are different from those involved in computing 84 ÷ 7. Addition and multiplication facts are primarily stored as declarative memory (you've memorized that 7 + 8 = 15). Division and multi-step subtraction engage more procedural processing and place heavier demands on working memory because you're often holding intermediate results while executing a sequence of steps.

By tracking each operation independently, the Sharpness Score can reveal patterns that a blended score would hide. You might discover that your addition and multiplication speed is stable across the week, but your division performance drops measurably on Monday mornings — a signal that your working memory is more affected by weekend disruptions than your fact retrieval. That kind of specificity is what makes the data useful for self-experimentation and pattern discovery.

What Gets Filtered Out

Raw response times are noisy. A phone notification mid-problem, a sneeze, a momentary distraction — these create outlier data points that don't reflect your actual cognitive speed. The algorithm handles this through statistical outlier detection, identifying and down-weighting response times that fall far outside the expected range for your current difficulty level and operation type.

Accuracy matters too. A fast response time on an incorrect answer doesn't indicate sharpness — it indicates a speed-accuracy tradeoff that went the wrong way. The algorithm weights correct responses more heavily, so you can't game the score by rushing through problems and guessing. This mirrors how cognitive scientists approach psychometric measurement: speed is only meaningful when coupled with accuracy.

The Assessment: Bootstrapping Your Baseline

A personal baseline requires data, and data requires sessions. This creates a cold-start problem: on your first day, there's no baseline to compare against. The initial assessment — a six-minute guided session that walks you through problems at progressively calibrated difficulty levels — solves this by generating enough data points across all four operations to establish a preliminary baseline from day one.

The assessment also determines your starting difficulty level. Rather than asking users to self-select (most people either overestimate or underestimate their arithmetic ability), the assessment observes your actual response times and accuracy rates, then places you at the difficulty level where you're challenged but not overwhelmed. This calibration means your subsequent daily sessions generate maximally informative data — you're always working near the edge of your ability, where meaningful variation shows up.

Why This Design Is Deliberate

Every design choice in the Sharpness Score algorithm traces back to a single principle: measurement, not gamification. There are no streaks to protect, no points to accumulate, no leaderboards to climb. The score goes up when you're sharp and down when you're not, and it tells you the truth either way.

This is a deliberate departure from the brain training model, where scores are engineered to trend upward over time to keep you engaged. Lumosity's scoring system, for example, was designed around level progression — you "improve" partly because the system gradually reveals harder content, creating a sense of advancement that may or may not reflect actual cognitive change. The Sharpness Score has no such inflation mechanism. If your processing speed plateaus after four weeks, the score will reflect that honestly.

The tradeoff is clear: the Sharpness Score is less satisfying for casual users who want to see a number go up every day. But for the kind of person who tracks their HRV, logs their sleep, and actually wants to know whether their cognitive performance is changing — it's the metric that earns trust by telling the truth.

What You Can Do with Honest Data

An honest cognitive metric enables something that gamified scores never can: real self-experimentation. When you know that your Sharpness Score isn't inflated by level progression or engagement design, you can start asking genuine questions. Does that new sleep protocol actually affect your morning processing speed? Is your division performance trending upward after three weeks of daily sessions, or has it plateaued? Are you measurably sharper on days you exercise versus days you don't?

These questions only have meaningful answers when the measurement system isn't trying to make you feel good. A metric that always goes up is great for retention but useless for learning. The Sharpness Score was designed for people who'd rather know the truth about their brain than be reassured by a number that flatters them.

That's the difference between a score designed to keep you using an app and a score designed to keep you honest about what's happening inside your head. The algorithm behind the Sharpness Score is built for the second kind of person.