How Scoring Works
Know Ball grades one skill: how close your order came to the order reality produced. You never predict points — you line players up, the games line them up too, and your score measures how well the two lists agree.
Your order · actual finish
One week, three rankers
94%
Saw it coming
52%
Mixed week
9%
Had it backwards
The closer your order tracks what actually happened, the higher you land in the field.
Locking follows kickoff.
Your first entry is due before kickoff. After that, players freeze as their games start while unstarted players remain editable.
Submit before kickoff
Your first weekly entry must be submitted before any ranked player's game starts.
Players lock individually
When a player's game kicks off, that player stays fixed in your rankings.
Unlocked players can still move
You can keep updating the players who have not kicked off, but only around the locked slots.
78% means you beat 78% of the field.
Your score is a place among everyone who played, not a percent of picks you got right. Some NFL weeks are chaos and every raw score drops; other weeks go to script and every raw score rises. Grading against the field is what keeps a wild week and a chalk week equally fair.
Everyone who played this week
Better →
You — 78%
beat 78% of the field
Each bar is a ranker's week. Your percentile is simply where your week landed among all of them.
Under the hood, we grade your order.
Here is one WR week. Level lines mean you saw it coming. The two calls that hurt are both visible: the bust you ranked second, and the breakout you never ranked — who still enters your grading, as if you had ranked him below everyone you listed. Nothing can be hidden and nothing disappears.
Your order
Actual finish
Built so the grade is fair.
Four design choices protect the score from luck, blowups, and cherry-picking.
The industry's player pool
You are graded on your ranked players plus the week's actual top finishers — the same pool construction and cutoffs FantasyPros uses to grade its experts.
Getting the top right counts most
Real production is not evenly spaced: the gap between the week's best player and its 5th-best is huge, while the 24th and 28th are nearly interchangeable. The grading scale spaces rank slots the same way — ranking the week's #1 WR down at #15 costs you real points, while shuffling your mid-20s barely moves anything.
1
5
10
15
20
One monster game can't warp a week
When a player goes nuclear for 50, nobody saw the exact number coming — and in a points-math system that one game would swamp everyone's score. Here, each performance is graded by where it lands against the last 16 weeks at the position, so a 50-point day and a 35-point day both count as an elite day. If you had him ranked high, you were right; you didn't need to call the fireworks.
Skill is the order, not the exact points
Nobody can predict whether a good day means 15 points or 25 — that part is noise. What repeats week after week is judgment about who outscores whom. Grading the order rewards that judgment instead of lucky guesses at exact totals, which is why the same sharp rankers keep landing near the top.
Overall works like a decathlon.
Your six position scores combine into one complete week, and that complete week is what gets placed against the field. Overall only exists when you submitted all six positions — no entering your best events and skipping the rest. On the leaderboard, Avg % is the multi-week version of the same idea: your typical week against the field.
Weekly results
Position percentiles
Overall score
70%
The six raw scores combine first; the combined week is then placed against the field. That is why 70% is not the average of the six percentages — each card sits on its own curve, and Overall sits on the combined one.
Technical note+
For each position, the evaluation pool is the union of your top-N ranked players and the week's top-N actual scorers. Your ranks map to an approximately normal projection vector via an inverse normal (probit) transform; actual half-PPR points map to a result vector by ranking each performance against a rolling 16-week positional history, then applying the same probit remap. The raw score is the Pearson correlation between the two vectors (Spearman is stored as a robustness check). Weekly percentiles standardize raw scores across all rankers in the week (z-scores mapped through the normal CDF), and every result is stored under a methodology version so the algorithm can evolve without rewriting history.