The short version
RQI stands for Roster Quality Index. It answers one question: given exactly these players, how good is this team?
It works from the bottom up. Every player on a roster is priced individually. Those prices are combined into a team number. That team number becomes a projected season — a record, a projected NET ranking, tournament odds.
Because the team number is built out of player numbers, it exists the moment a roster exists. It does not wait for games.
Four terms worth knowing
- Per-40. A stat restated as "per 40 minutes on the floor," so a twelve-minute player and a thirty-two-minute player can be compared fairly.
- Effective field goal percentage. Shooting accuracy, with three-pointers credited for the extra point they're worth.
- Efficiency points. The currency everything ends up in — net points per 100 possessions. The Division I average is zero. A team at +1 is one point per 100 possessions better than average.
- Possession. One trip with the ball. Counting per possession instead of per game is what lets fast teams and slow teams share a single scale.
One player becomes one number
Start with the job description. Every position has a baseline: what an average Division I guard, forward, or center produces. A player's starting score is how far he sits from that baseline — on scoring, on shooting, on passing — with each gap weighted by how much it actually moves offense.
Then price the context. A stat line earned against the best defenses in the country is not the same stat line earned against a soft schedule. Conference strength and schedule difficulty adjust the score up or down.
Transfers get special treatment. Rather than taking last season's numbers at face value, an incoming transfer is priced through a model trained on how transfers have historically translated when they change levels. A player moving up gets projected for the league he's arriving in, not the one he left.
Then give him his share of the floor. A basketball game contains exactly 200 player-minutes — five positions, forty minutes each. That budget gets allocated across the roster. A player's contribution is his score scaled by his share of those minutes, so a starter and a deep reserve are weighted the way they'll actually be used.
Then trust it in proportion to the evidence. The fewer minutes behind a player's numbers, the less those numbers are allowed to swing the result. A large scoring rate produced in very little playing time is treated with appropriate suspicion until there is enough of it to believe.
Players become a team
Offense. The roster's individual contributions combine into a team offensive projection, expressed in points per 100 possessions. A completeness factor keeps a half-filled roster from being read as a finished one.
A law of the sport gets enforced. Across all of Division I, every point scored is a point allowed. The league's average offense must equal its average defense. That identity is checked on every rebuild, and the projection is not published if it fails. It sounds like bookkeeping. It isn't — a system that skips this step can quietly drift the entire league up or down and never notice.
Defense. Defense is projected directly, from each player's defensive rates together with the shape of the roster around him — size, continuity, how much of the rotation is new. What matters most is that the team's defensive projection splits into real per-player slices. Each player's defensive number is his actual share of the team's projection, not a separate estimate placed beside it.
Each player is then graded against his own team's standard: how much cheaper or more expensive his slice is than his minutes alone would predict. Across a roster those differences sum to zero, so team-level defensive quality lives on a clearly labeled team line instead of being smeared across individual players who didn't earn it.
We only print what we can defend
Before publishing per-player defensive numbers, we tested whether they hold up — whether a player's defensive rate in one season tells you anything about the next.
Above a clear threshold of games and minutes, it does, reliably enough to print. In a middle band, it's closer to a coin flip. Below that, it's noise.
So the product prices defense only above that measured floor. Below it, a player carries an honest label — insufficient sample — instead of a fabricated decimal, and his minutes still count toward the team line where a roster-level unknown honestly belongs.
A number printed with false confidence would be the most misleading thing on the page. The label is the model telling you the truth about what it knows.
It adds up exactly
A team's RQI is its projected offense minus its projected defense.
Here is the part that separates this from a conventional rating: the individual player numbers, plus a handful of clearly named team-level lines, sum to the team's RQI exactly. Not approximately. The system refuses to serve a breakdown that doesn't reconcile.
That's not a technical flourish. It's what makes the number usable. Because every team number is an exact sum of player numbers, every roster decision is attributable — you can see precisely what each player is worth to this team, and precisely what changes if he isn't on it.
From a roster to a season
RQI feeds a season simulation.
Conference play is built as one shared schedule where every game exists exactly once, so league wins equal league losses by construction — another conservation law the system checks rather than assumes. Non-conference schedules are drawn to match how programs at that level actually schedule.
Every game becomes a win probability. Probabilities become a projected record. Record and schedule strength become a projected NET ranking. And NET becomes tournament probability, through a relationship calibrated on how the selection committee has actually behaved across recent seasons.
One chain, start to finish, with no hand-tuning between the stages.
Why this is different
Ratings built from games already played are very good at telling you what happened. They cannot tell you what a roster you just assembled is worth, for a simple reason: there are no players inside them, and at portal time the games don't exist yet.
RQI is built the other way around. Players first, team second, season third. So it exists in April. It can price a transfer target in your context before he commits. And when a roster changes, the whole season re-runs.
What RQI does not do
A projection is not a prediction of certainty, and we'd rather say so plainly.
RQI does not know about injuries that haven't happened, players who develop faster than their history suggested, or a locker room that does or doesn't come together. It prices the roster in front of it, using what the sport has historically rewarded. Teams will beat their projection and teams will miss it — that is what a projection means.
What it will always do is show its work, reconcile exactly, and refuse to print a number it can't stand behind.