The details matter.

Available modes

Basic roster audit ↗ · Basic simulation checks ↗

Every pair in the four playable leagues can field a unique, eligible roster in Basic or within the Pro budget. Basic uses starters without a budget; Pro adds a full roster and budget management. Each has separate leaderboards and challenges. These are game simulations, not forecasts or official league competitions.

Hockey is played on Puck Perfection’s NHL Double Team game ↗. Its rosters, results, profiles and leaderboards stay on Puck Perfection.

How the budgets work

NBA uses the $164.961M 2026–27 league cap; NFL uses the $301.2M 2026 league cap. Each selected player consumes their verified same-season cap charge. Percentages use the league cap, never a team’s payroll or adjusted cap room. The server adds exact dollar amounts before checking the limit; display rounding cannot create extra room. NBA cap charges come from SalarySwish; NFL charges come from Over the Cap. Each player links to their dated source.

These games apply a hard limit to selected-player charges. NBA exceptions, aprons, cap holds and two-way contracts are excluded. NFL rollover, unrelated dead money and transaction savings are excluded; practice-squad contracts cannot fill the 53-player roster. A published injured-player charge is retained while the simulation assumes that player is healthy. Missing, unconfirmed or conflicting current charges remain excluded. This is a roster-building game, not a simulation of every collective-bargaining rule.

MLB retains performance-derived credits. Premier League retains official FPL prices expressed as credits. Neither is labelled a real salary cap.

Basketball starts PG/SG/SF/PF/C with 10 bench players. Baseball has a nine-player batting lineup, four bench players, five starting pitchers and eight relievers. Soccer uses a fixed 4–5–1 starting XI and nine substitutes: one goalkeeper and eight outfield players. Signed organizational players are assumed healthy. MLB two-way players occupy one role.

Difficulty rewards the challenge

A weaker pair earns a larger multiplier on leaderboard score. It does not improve player ratings or win probability. NBA and NFL use the last completed season’s standings; baseball and soccer use their dated current standings. The multiplier is visible before you build. Scores remain separated by league and scoring version.

How your roster becomes a team

Every club is evaluated from its eligible player pool using the same model as your roster. Reliable depth matters over a full season. Weak slots receive a small penalty against a fixed quality threshold; improving a player never lowers that fixed lineup’s raw strength.

Basketball gives the starting five 70% of regular-season influence and 84% in the postseason, with the best reserves receiving more of the remaining minutes. Baseball’s postseason emphasizes its top starting pitchers and bullpen arms. NFL accounts for weak pass and run blocking while keeping its position-specific unit model. Soccer keeps its starting XI and substitutes throughout the league season.

These weights are game-design choices. Simulated scores determine wins, losses and scoring totals, including in friend challenges. A strong roster improves the odds; it cannot guarantee a title. Results show a short roster profile after the reveal.

What the simulation simplifies

NBA and MLB use balanced game schedules of the correct length, not the real fixture list. Premier League plays a full double round robin. The merged team replaces the weaker source club; the other source club fields the unused combined pool, so selected players never play against themselves. Tied standings use simplified deterministic tiebreakers.

Basketball includes the 7–10 play-in, baseball includes division winners and first-round byes, soccer has no playoffs, and NFL uses a seven-team bracket in each conference.

NBA evaluation uses recent box-score Game Score with small-sample shrinkage. MLB uses OPS, baserunning, ERA and strikeout rate. Premier League uses expected-goal involvement and defensive/goalkeeper outcomes with minutes-based shrinkage. These models cannot capture every defensive, tactical or park effect.

Friend challenges use neutral-site head-to-head: one match in soccer, one football game in NFL, and best-of-seven in basketball and baseball. This is a separate game format. Difficulty does not change the matchup, and challenge runs never enter the solo boards.

Rookie pedigree, with room to prove it

NBA rookies with verified standard contracts and a 2025 or 2026 draft record can play before their NBA debut. Draft position supplies a conservative starting projection, clearly labelled as an estimate. Established players retain their measured NBA performance ratings.

NFL uses 80% of the individual player assessment and 20% of a draft-position prior, with the final adjustment limited to ±3 points. MLB Pipeline’s current top-100 ranking supplies a prospect prior that loses weight as MLB plate appearances or innings accumulate. These are game formulas, not official scouting grades or a guarantee that a highly drafted player will succeed. Contract costs do not change with NBA or NFL pedigree.

Premier League has no draft. PL newcomers use verified playing performance and official FPL prices; we do not assign invented draft pedigree. Players needing outside-league evidence remain listed in the eligibility audit.

Roster audit

Players with missing performance evidence are individually listed as quarantined. Zero minutes or an absent contract never silently turns into an invented salary. These are eligible game pools, not a promise that every registered prospect has a playable evaluation.

Calibration results

These tests build an affordable combined roster from strong, middle and weak source pairs, then simulate 100 seeds per pair. They test game behavior; they are not historical single-team replays or a predictive backtest.

Roster seasons and team records

The roster season tells you which players you can choose. Team records come from the separate standings window below and set the difficulty bonus. A previous-season record is not a prediction for the roster season.

For example, Arizona’s 3–14–0 is its final 2025 NFL regular-season record. Coventry’s 0 Premier League points are after three matches in the September 9, 2026 snapshot, not a completed 38-match season. These records do not update live. Player performance windows appear separately in each player’s details.

Sources & snapshot dates

All 30 NBA, 32 NFL and 30 MLB team roster feeds, plus the complete official Premier League player pool, were rechecked September 10, 2026. Rosters are dated snapshots, not a live transaction feed. Performance windows remain labelled separately. Raw source URLs, fetch timestamps and player IDs are retained. NBA and NFL roster and draft records come from ESPN; current cap charges come from SalarySwish and Over the Cap respectively. MLB and Premier League roster feeds are official league sources.

NFL’s engine evaluates passing, rushing, pass defense, run defense and special teams separately. Opta’s blocking samples and the team-unit/fantasy references are diagnostic comparisons. Their raw scores are not blended into player assessments, and uncertain measurement windows are not labelled as current-season statistics.

Double Duty is an independent fan game, not affiliated with any league, team or data provider.

Your runs stay yours

One browser profile and display name works across NBA, NFL, MLB and Premier League here. Puck Perfection has its own profile. Server timestamps determine Daily and Weekly standings in America/Toronto; weeks start Monday. Saved results retain their original roster, prices, score, seed and data version. Refresh recovers unfinished drafts and reveals. Retry returns the same result instead of entering it twice.

Browser identity is not a cross-device account. Clearing cookies starts a new profile. Challenge links are persistent, but an archived data version becomes read-only to keep both sides on the same rules. A shared result or challenge link lets friends view the roster and build their answer.