# The dynasty context packet

Fill what you know. Use **unknown**, **not supplied**, **none**, or **not applicable** deliberately; they mean different things. You do not need a perfect dossier before asking a question. Exact league settings and the assets involved usually come first.

This is an intake template. Blank fields are intentional. Keep player IDs and league IDs as text, preserve source timestamps, and do not overwrite old observations with new ones.

You can start with your league link instead of typing the whole roster. The [supplemental league-access guide](08-league-access-guide.md) explains what to copy for each platform and what to send if your AI cannot open it. Only fill the gaps it could not read.

## Minimum useful packet

```text
AS OF / DEADLINE:
Date, time zone, season/week:
Decision deadline:
Research mode: live / supplied evidence only / historical as-of

LEAGUE:
Platform, league URL/ID, season, number of teams:
Exact starting slots and scoring:
Bench/IR/taxi size and relevant restrictions:
Playoff format and rookie draft-order rule:
Special rules that affect this decision:

MY TEAM:
Roster and player IDs if available:
Picks: year, round, original franchise, current owner, known slot/range:
Objective and horizon:
Constraints or preferences:

DECISION:
Exact question or offer, with send/receive directions:
Offer status: hypothetical / available / received / accepted / completed
Relevant counterparty roster and picks:
Available waiver or trade alternatives:
Sources/snapshots and dates:
```

If time is short, begin here. The assistant should identify the most consequential missing input and continue with the useful work available.

## Full packet, loaded only when relevant

### A. League constitution — stable until changed

```text
Context version:
Platform / league ID / season / previous league ID:
League rules last verified:
Number of franchises and player copies:
Lineup type: managed / best ball / other

STARTING SLOTS:
Slot | number required | eligible positions | minimum/maximum restrictions

SCORING:
Passing yards / touchdowns / interceptions / bonuses:
Rushing yards / touchdowns / attempts / first downs / bonuses:
Receiving yards / touchdowns / base reception points:
Position-specific reception bonuses and whether they stack:
Fumbles / return scoring / other:
IDP categories, event stacking, position eligibility and bonuses:
Link or export of complete settings:

ROSTER:
Active roster / bench / offseason capacity:
IR capacity, eligibility and required activation:
Taxi capacity, experience limits, deadline, promotion and re-entry:
Roster minimums, position limits, duplicate-player rules:

COMPETITION:
Schedule, divisions, median game or all-play rules:
Playoff teams, weeks, byes, reseeding, round length:
Payout structure and any regular-season prizes:
Tie-breakers:
Expected league continuity / empire reset or end conditions:

DRAFT / TRANSACTIONS:
Startup or rookie draft; snake/linear/auction; third-round reversal:
Rounds, draft pool, devy depletion, clock and trading permissions:
Draft order for playoff and non-playoff teams:
Max PF definition, taxi/IR treatment, lottery and tie-breakers:
Future-pick trading limits and payment obligations:
Trade deadline, review process and other restrictions:
Waiver schedule, FAAB, zero-dollar bids and tie-breaks:

CONDITIONAL FORMATS:
Contracts: salaries by year, escalators, expiry, dead money,
  guarantees, extensions, franchise rights, cap and available space:
Devy/C2C: college scoring, eligibility, NFL rights, promotion,
  depleted rookie pool and graduation limits:
Keeper/empire/multi-copy: precise rules:
```

### B. Manager objective — explicit preferences

```text
Primary objective:
Secondary objectives:
Preferred competitive horizon:
Willingness to sacrifice this year / next year:
Risk preference and tolerance for large downside:
Practical time available for waivers/trades/lineup management:
Favorite players or no-trade preferences:
Budget or future dues constraints:
My own assessment of team direction:
What evidence would make me reconsider that assessment:
```

Enjoyment is a legitimate preference. If keeping a favorite costs projected competitive value, show the cost rather than pretending it does not exist. A manager's stated desire to win now is an objective; the claim that their roster is a contender is a hypothesis.

### C. Current league state — dated evidence

```text
Snapshot observed at:
Underlying data through:
Source and access method:
My roster ID:

PLAYERS:
Provider ID | name | NFL team | fantasy eligibility | active/IR/taxi
  | contract if relevant | last verified

PICKS:
Draft year | round | original franchise ID | current owner ID
  | exact slot if fixed | projected range if not | protections
  | source and last verified

TEAM POSITION:
Record, points scored, Max PF, all-play record where available:
Projected legal lineup and source/method:
Important absences and bye coverage:
Future starter coverage, major role/age concentrations:
FAAB, waiver priority, cap space, obligations:

LEAGUE LANDSCAPE:
All rosters/picks or relevant counterparties:
Strongest competing teams and evidence:
Waiver pool at relevant positions:
Recent completed league trades, exact packages and dates:
Current trade block, asks and offers with timestamps:
```

Do not infer a roster's strength only from its record or total market value. Keep forward projections, past scoring, and roster market values in separate columns.

### D. Player thesis card — for assets carrying the decision

```text
Player ID / name / age as of decision / position:
Current NFL team and role verified at:
Evidence of ability and limitations:
Role and opportunity: routes/snaps/touches/targets, with denominators:
Opportunity quality: scoring-area, passing-down, depth and alignment:
Efficiency measures: definitions, comparison group, sample:
Injury/availability: report, date, status, unknowns:
Organizational commitment and competing players:
Contract facts and team options, without inferring guaranteed usage:
Relevant base rate or comparison group:

Bear case / mechanism / evidence needed:
Base case / mechanism / evidence needed:
Bull case / mechanism / evidence needed:
Expected time to better information:
What would falsify my thesis:
What remains useful if the thesis is wrong:

Football forecast:
Usable lineup contribution:
Public market observations:
Local asking price / actual offer:
My buy ceiling / sell floor / hold rationale:
```

### E. Evidence and market ledger

One row per claim or observation; link related rows to their common origin.

| Field | What to record |
|---|---|
| Claim | The exact fact, metric, report, package, or price observation |
| Category | Fact / report / projection / preference rank / ADP / completed trade / ask / offer / interpretation |
| Identity | Player IDs; pick year, round, origin and owner |
| Source | Direct URL or authorized local record; original reporter or dataset |
| Time | Event date; publication/update date; data-through date; retrieval time |
| Comparability | League settings, season, player pool, sample window, population |
| Measurement | Native units, denominator, sample size and dispersion if available |
| Dependency | Shared original report or underlying data; possible duplicate |
| Support | What the observation establishes and what it does not |
| Status | Current / superseded / disputed / incomplete |
| Decision impact | Which assumption or action could change |

Never fill an unavailable sample size with a guess. Use a short ledger for small decisions; archive the detail outside the active prompt.

### F. Decision brief — new for each question

```text
Decision ID / as-of timestamp / deadline:
League context version:
Question and objective:
Action being considered:
I send:
I receive:
Counterparty sends/receives:
Offer status and evidence it is still available:
Required cuts, promotions, salary changes or follow-on moves:
Alternatives including hold:
Counterparty facts:
Counterparty assumptions:
My current leaning and why:
What I most want challenged:
```

### G. Decision journal — preserve before knowing the result

```text
Timestamp / decision ID / context version:
Recommended action and best alternative:
Action actually taken and execution price, if any:
Objective served:
Verified facts carrying the decision:
Forecasts and assumptions carrying the decision:
Why this beats the best feasible alternative:
Price boundary and conditions that reverse the answer:
Forecast event definitions, horizons, estimates and methods:
Confidence in facts / forecast / robustness of action:
Review date or event trigger:

APPEND LATER:
New evidence and when it became knowable:
Observed football outcome:
Observed market outcome and whether an exit was available:
Observed execution outcome:
Process error / forecast error / execution failure / ordinary uncertainty:
Change to future decision practice:
```

## Refresh rules

Refresh ownership, offer status, eligibility, and decision-sensitive news before a transaction recommendation. Refresh the draft board when the clock is live. Recheck rules after season rollover or amendments. Refresh market comparisons after a material event or when the source's sample window no longer fits the question. There is no universal seven-day freshness rule.

For historical exercises, allow only evidence available at the cutoff. A present-day page is not an archive of its historical state. Models may already know historical outcomes; use fictional fixtures and prospective decision records when testing reasoning or forecasting skill.

## A platform detail that can break otherwise good reasoning

Sleeper distinguishes a pick's original franchise from its current owner. Its read-only API documents league settings, rosters, transactions, draft information, and traded picks. Reconstruct the full pick inventory from the draft rules and transfers; the traded-picks list alone is not every pick. Verify the relevant league season and draft. [Sleeper API documentation](https://docs.sleeper.com/)

The same discipline applies to other platforms: use their actual definitions rather than copying field meanings between systems.