The Fantasy Basketball Encyclopedia

Projections

2026-27 fantasy basketball projection methodology

The season model projects availability, active-game minutes, and primitive box-score events first, then derives format-specific ranks without confusing analyst consensus with measured uncertainty.

Up: 2026-27 Fantasy Basketball Encyclopedia

Related: Using projection disagreement in fantasy basketball · Percentage categories and volume · Points-league volume and efficiency · Games played versus per-game value in fantasy basketball

Information cutoff

The first encyclopedia pass uses an information cutoff of 2026-07-18, America/New_York. It is an offseason baseline, not an opening-night promise. The 2026-27 schedule, training-camp roles, final standard contracts, and many medical ramps may still be unresolved. Each later update should retain its own cutoff rather than overwriting the epistemic state of this pass.

Projection chain

For player \(i\), the basic exposure decomposition is

\[ T_i = G_i \times M_i \times r_i, \]

where \(G_i\) is expected games played, \(M_i\) is active-game minutes, and \(r_i\) is production per minute. The displayed per-game line is \(M_i r_i\); the total line is \(T_i\). Changing minutes can also change usage, matchup quality, and efficiency, so \(r_i\) is scenario-dependent rather than mechanically fixed.

1. Roster and role states

Each player receives a roster state, projected depth-chart slot, and low/base/high role scenario. Pending free agents, two-way players, non-guaranteed contracts, unresolved trades, and injury replacements remain explicit branches. Official status is governed by 2026-27 fantasy basketball evidence and confidence policy.

2. Team minutes

The base active-game rotation sums to 240 regulation minutes per team. That invariant applies to a representative competitive game, not season totals. Overtime is excluded. Deep reserves can carry contingent minutes that are not simultaneously active; player pages distinguish rotation minutes from injury-path minutes so the roster does not pretend every player receives his ceiling at once.

3. Games and availability

Expected games are distributions, not moral judgments about durability. Inputs can include prior availability, current official medical evidence, age, known return windows, workload-management risk, role security, and team incentives. The model does not infer a diagnosis from an injury label. A mean such as 62 games can represent several different states and should be accompanied by the material branches.

4. Per-minute rates

Veteran rates begin with recency-weighted multi-season performance and are adjusted for age, role, teammates, coaching environment, and regression of noisy events. Attempts and makes are projected separately for percentage categories. Steals, blocks, shooting percentages, and rookie translations receive wider uncertainty than stable volume events.

Rookies and low-sample players use age, prior competition, role, physical translation, draft investment, and analogous players. Summer League supplies role and skill evidence but receives little weight as an NBA box-score sample because competition, rotations, and incentives differ sharply.

5. Category value

Nine-category rankings use 3PM, PTS, REB, AST, STL, BLK, FG impact, FT impact, and TO. Ratio value is based on makes and attempts relative to a declared baseline, not a z-score of raw percentages. Turnovers are directionally inverted. Rankings must state whether they are per-game or totals and identify the rosterable population used to estimate category distributions.

No single nine-category rank represents every build. Punt views remove or reweight categories only after preserving the primitive projection, and Robust punt builds in fantasy basketball governs the roster-level interpretation.

6. Points value

Points-league rankings apply the league's exact scoring equation to the primitive stat line. Yahoo High Score, ESPN, Sleeper, Fantrax, and custom leagues can disagree even with identical basketball projections. Generic points ranks therefore use a declared reference formula and are labelled as such.

7. Dynasty value

Dynasty rank is not a longer redraft list. It combines multi-year production distributions, age and development, role durability, injury risk, league depth, contention horizon, and replacement liquidity. Market value and projected fantasy production are retained as separate fields.

Consensus and disagreement

Public sources are not counted as independent votes merely because they occupy different domains. The baseline comparison uses a median after normalizing date, horizon, format, games, and minutes. Source-specific forecasts remain visible. Disagreement is decomposed into availability, minutes, rate, transformation, and unexplained residual; it is not called a confidence interval.

The encyclopedia's low/base/high cases are causal states, not the minimum, median, and maximum numbers found online. Where the model cannot establish defensible probabilities, it publishes thresholds and reversal conditions rather than false precision.

Required player fields

Each full player dossier should contain team and status, eligible positions when sourced, age/experience, depth-chart role, base GP and MPG, 9-cat per-game projection, points and dynasty context, category strengths and liabilities, low/base/high mechanisms, injury and transaction branches, source disagreement, confidence by component, dated sources, and open questions.

Audit gates

Open questions