Rules coverage & accuracy

Enforced every time you build a trade or a signing.

• ✓Every 2026-27 money line — the cap, luxury tax, both aprons, the mid-level and bi-annual exceptions, minimum salaries, and the max tiers — set from the CBA's own percentage-of-cap formulas, not typed in by hand.
• ✓Trade matching with the real expanded math — including the middle band that grows with the cap to $9,095,709 for 2026-27, not the flat $7.5M from 2023-24.
• ✓What each apron actually costs you: over the first apron you lose expanded matching and can't bring a player in by sign-and-trade; over the second you also can't combine salaries or send cash.
• ✓Every move that hard-caps a team, enforced for the rest of the season. Using the mid-level or bi-annual, taking a player by sign-and-trade, or taking back extra salary caps you at the first apron; the taxpayer mid-level, combining salaries in a trade, or sending cash caps you at the second.
• ✓Traded-player exceptions — the real league-wide ledger, plus any your own lopsided trades create — applied for you to absorb an incoming salary.
• ✓Sign-and-trades end to end: the acquiring team's hard cap, the contract length and structure limits, base-year compensation when you re-sign then trade a player, and poison-pill rookie extensions.
• ✓Re-signing your own free agents at their real Bird, Early-Bird, or Non-Bird ceilings — and the cap holds those rights keep on your books until you re-sign or renounce them.
• ✓Extensions, each with their own year limits, raise limits, and the extend-and-trade freeze window.
• ✓Draft picks that behave like the real thing: a pick you already owe is off your board, the Stepien rule stops you trading away first-rounders two years running, and every trade updates who owns what for the moves that follow.
• ✓Waive a player yourself and the guaranteed money sticks to your books as dead money — counting against the cap, tax, and apron — following the real guarantee, not the listed salary. Choose to stretch it and it spreads evenly over 2N+1 years under the 15%-of-cap guardrail, right alongside the real 2025-26 waives and stretches already on the sheet.
• ✓Your actual July 2026 starting point: which teams already spent an exception, who is hard-capped and why, and which cap-room teams gave up their mid-level and bi-annual for the year.

Directionally right, but limited by what public data makes available — and flagged where it matters.

• ~Two 2026 moves are on the board as done deals while the league is still reviewing them, and either could be unwound. Kawhi Leonard is a Raptor here — Toronto and the Clippers agreed that trade on June 30, but it was put on hold pending the NBA's investigation into whether the Clippers circumvented the cap, with no timetable. Gary Trent Jr.'s four-year deal with Milwaukee is under a circumvention review of its own. Both are booked the way every public roster page books them, because the reporting expects both to stand — but a verdict either way is a thing this board would have to be told about.
• ~Injury availability is projected from the injury's TYPE and DATE, not from games missed last season. A torn Achilles, ACL, patellar tendon, or a microfracture is assumed to cost twelve months from the day it happened, so one suffered late in a season eats into the next one; everything else is treated as healed by opening night. That is a rule of thumb, not a medical opinion, and it can't know about a setback or an early return.
• ~Apron Value — the 0–100 number on every player — is a rough read on impact from box scores and lineup data. It leans on BPM and carries an uncertainty band; it isn't trying to outdo the big public metrics like EPM or DARKO.
• ~A team's projected record is a read on the roster as it stands today — not a full-season forecast. There's no coaching, chemistry, or playoff translation baked in.
• ~How wrong that projection can be, measured rather than guessed: about 3.5 points of net rating, which is roughly SEVEN WINS either way. The model was tested against 240 team-seasons (2018-19 through 2025-26) by predicting each season using only what the players had done in the two seasons before it, and never training on the year it was grading. It explains a little over half the variation in team net rating. Two teams three or four wins apart in these standings are not meaningfully different, and shouldn't be read as if they are.
• ~The projection is built from two things: the talent in a team's projected rotation, and its perimeter defense, each measured against the league and then scaled so the spread of the standings matches what real net ratings actually do. A team-fit term used to sit alongside them and no longer does — across those 240 team-seasons it made predictions no better than leaving it out, and its own weight swung wildly depending on which season was held back. Fit is still computed and still shown on the board, because it's a real read on how a roster goes together; it just isn't allowed to move a projected record any more.
• ~Four known soft spots in that calibration, none of which change the conclusion: player positions in past seasons are taken from this season's; the historical talent measure is a simpler version of the Apron Value the site shows you; two of the eight seasons are the bubble and the 72-game year; and historical rosters mean everyone who played that season, including midseason arrivals.
• ~A rookie has no NBA minutes yet, so his value, fit, and playing time are projected from draft slot, college production, and scouting consensus — a read on his rookie year, not a ceiling. Most rookies grade out below average, top picks near it.
• ~A handful of recently retired or overseas veterans (Dwight Howard, Serge Ibaka, and the like) are on the free-agent board as minimum-deal flyers, valued for the realistic drop since their last NBA season — a 40-year-old Dwight reads as a deep-bench center, not the DPOY version.
• ~Positions come from where a player actually logged his minutes; a real second position is any spot he played at least ~12% of the time. Players without enough tracked minutes fall back to their listed position.
• ~Draft-pick trade value is projected from each team's strength, discounted for risk and how far out the pick is — the fairness meter just adds it up per side.
• ~Some traded-player-exception details are simplified: one exception per team per trade, whole players only.
• ~Any use of the non-taxpayer mid-level hard-caps the team at the first apron here. The real rule is softer in one narrow case: a use within the taxpayer MLE's size and 2-year length can later be re-characterized as the taxpayer MLE (Art. VII §6(f)(5)), trading the first-apron cap for a second-apron one.
• ~A matched offer sheet freezes the player from trades outright for a year here; the real rule is no trade without the player's consent — and never to the team that made the offer.
• ~A few fine adjustments to a team's apron salary need per-player bonus data that public feeds don't carry. When it isn't public, those small adjustments are left out rather than guessed.
• ~Which apron tier a team sits in is measured on salary actually signed. Cap holds for free agents you haven't re-signed or renounced still count against your room to sign and absorb, but they don't move you up an apron tier here. Denver is the live example: after matching Spencer Jones' offer sheet they read as a first-apron team on signed salary, while reporting places them in the second — the difference is Peyton Watson's unrenounced hold, which they're keeping on purpose while they try to bring him back.
• ~Reported deals are public as a term and a total, so the first-year salary is back-solved from the raise the deal is entitled to — 8% re-signing with your own team on Bird or Early Bird rights, 5% otherwise, including matched offer sheets and anything paid with an exception. That lands the big ones exactly (Austin Reaves' 4-year max back-solves to the 25% figure to the dollar), but a deal with unusual year-to-year structure — flat, declining, or front-loaded — is smoothed into standard raises.

Real rules or data we haven't wired into the offseason board yet.

• ✗Pick-swap outcomes and conditional incoming picks. You can trade swap rights on the board, but a swap never resolves to a final pick slot, and a conditional first you'd acquire isn't counted as Stepien coverage.
• ✗Incentive bonuses. Public contracts don't reliably say which bonuses count as likely or unlikely, so bonus money stays out of the live totals.
• ✗Signing second-round-pick exceptions and two-way contracts — you can't do those from the offseason drawer yet.
• ✗Designated “supermax” extensions — you can't build one from the board yet.
• ✗Freezing a future first-round pick for a second-apron team at all — the rule is modeled in the rules engine, but the board doesn't apply it to any pick yet, so a frozen first can still be traded here.