nba methodology
NBA Methodology
The rating measures one thing: how good a team really is right now. Everything else — projected wins, seeding, title probability — falls out of simulating the season from that rating.
Start with a single number for each team's strength. Let every game nudge it. Adjust it when a roster changes over the summer. Then play the whole season out thousands of times and count what happens. That's the entire idea — and each piece exists because it earned its place against results the model hadn't seen.
engine
Power ratings
Every team carries a single strength rating. After each game the winner's rating goes up and the loser's comes down — and the size of the move depends on the margin, because winning by twenty says more than winning by two. Strong teams that beat weak teams barely move; upsets move the needle hard. The rating doesn't reset in October either: a team ends one season where it deserves to and carries most of that into the next, so a genuine contender isn't treated like a coin flip on opening night. On the ratings page we show that strength as a league percentile, with the average team sitting at the 50th — it's easier to read "this is a top-quarter team" than a raw number.
key levers
What controls the projection
Projected wins
Once every team has a rating, the schedule is known, so we simulate the full 82 games — thousands of times over. Each simulated season gives one win total; average them and you get the projected wins we show as the projected win total. The runs also disagree with each other, and that disagreement is the honest uncertainty: the "± N" next to each total is how wide the middle 90% of outcomes runs. A team can be projected for 48 wins with a tight band or a loose one, and the band tells you which.
Title probability
The same simulation keeps going past game 82. We seed the bracket from each simulated season, play out the postseason series by series, and crown a champion. Do that enough times and a team's championship probability is simply the share of runs it won it all. Seeding chances, a top-four finish, making the playoffs — all of it is just counting how often each thing happened across every simulated season.
Off-Season Moves
Ratings shouldn't pretend last year's roster is this year's roster. When a team adds or loses real talent over the summer, we shift its rating before the season starts to reflect the new group. On the ratings page that shift is shown as a percentile jump — a team that got meaningfully better moves up the league order, one that got worse slides down — so you can see at a glance who won and lost the offseason.
Team methodology
A team's projection is not a pile of individual player forecasts added together. We rate the team as a unit — because how a roster wins games is more than the sum of its box scores, and role, fit, and how minutes actually get distributed matter. Player pages exist for the individual view; the team rating is what drives wins, seeding, and title chances.
faq
Common questions
What do the percentiles on a player card mean?
Each point on the Scoring and Role shapes is a player's place among everyone we project for 20 or more minutes a night — about 250 of the roughly 570 players on the board. That gate keeps the regulars, so the middle of it is a real rotation player, not the average of everyone on a roster: the median player in it is projected for about 12.6 points against 7.5 across the whole board. Percentiles round, so a 100th can be the second-best, not only the best.
What are the Resolve metrics on a player card?
Six of Resolve's own advanced metrics, always in the same order: three on offense (drive value, and-1 rate, clutch scoring) and three on defense (rim protection, defensive events, three-point defense). Each is placed among the players who have a value for that metric, and that group differs by metric — from about 110 to 310 players. Hover a point for a summary; click it for the full ranking.
Why do some numbers hold from one season to the next and others move to the middle?
What a player earns — minutes, the share of his team's shots he takes, how often he shoots threes — tends to repeat from one season to the next. What he converts — shooting percentages — barely persists at the same sample size and is pulled toward the league average from whichever side the player sits on. The projection leans on the first and discounts the second, which is why a player's shooting forecast can sit a little below or above last season even when his role is unchanged.
What does the Power Rating percentile actually mean?
It's where a team ranks in overall strength, with the league average pinned at the 50th percentile. A team in the 72nd percentile is stronger than roughly the bottom three-quarters of the league. We show a percentile instead of the raw rating because it's easier to read at a glance.
Why is projected wins shown as a single number with a ± instead of a range?
The single number is the projected win total — the average of every simulated season. The ± is the honest uncertainty around it, covering the middle 90% of outcomes. One number tells you the expectation; the ± tells you how confident to be in it.
How can a title probability be so low even for the best team?
Because the playoffs are a gauntlet and any single series can go the other way. Even the strongest team has to win four rounds against good opponents, and when you simulate that thousands of times, the best team in the league usually still loses more runs than it wins. Low title chances for a favorite aren't a bug — that's how hard winning a championship is.
Do the ratings update during the season?
Yes. The rating is designed to move after every game, so as real results come in a team's strength tracks toward what it's actually shown on the floor. The preseason page is the starting point; the rating doesn't stand still once games are played.
Why not just trust last season's record?
Records don't carry margin, strength of schedule, or roster changes. A team can win 45 games against a soft schedule and be weaker than a 42-win team that played everyone tough. The rating accounts for who you beat and by how much, then adjusts for who's actually on the roster now — which is why our order sometimes disagrees with last year's standings.
apex framework
For the platform-wide methodology framework — pre-registration policy, data philosophy, bias controls, and honesty notes — see the apex methodology page.