How the numbers work
Nothing here is a black box. The rankings and the trade analyzer run on one small engine, and this page is the whole recipe.
1. Counting categories
For points, threes, rebounds, assists, steals, blocks and turnovers, a player's score is his per-game average measured against the pool:
z = (player average − pool average) / pool standard deviationTurnovers are flipped so that fewer is better. A score of +1.0 means one standard deviation better than an average rostered player.
2. Percentages
A 60% shooter on three attempts matters far less than a 55% shooter on fifteen. So for FG% and FT% we score impact, which weighs the percentage by volume:
impact = makes − pool% × attemptsImpact is then turned into a z-score the same way as the other categories. Because it adds up across players, it also adds up correctly across a trade.
3. The pool
"Average" should mean the players a league actually rosters, not every player in the NBA. The pool is the top N players by total score (156 by default, about 13 teams of 12). Because the pool depends on the scores and the scores depend on the pool, the engine repeats the step until the pool stops changing. You can change N on any page.
4. Total and punting
A player's total is the sum of his category scores. Tap a category to punt it: it is removed from the total, the pool is rebuilt without it, and the board re-ranks.
5. Roster spots in trades
A 2-for-1 trade leaves one spot to fill, and a 1-for-2 forces a drop. With roster spots on, the difference is filled with replacement-level players: the average of players ranked just outside the pool.
What this leaves out
- Week-to-week volatility. A noisy category like threes is less reliable in a weekly matchup than its z-score suggests.
- Recent form, minutes changes and injuries. The numbers are season per-game averages.
- Your own roster. The trade analyzer scores players, not fit.
- Games remaining and schedule strength.
These are the next things to build. Each will be added as an option, with its formula shown here.