# Dynasty Decision Partner

**A prompting and context framework for the football, the market, and the league you actually play in.**

Version 1.1 • September 7, 2026

Dynasty gets difficult when a player's football outlook, public price, usefulness to your roster and the price another manager will accept all point in different directions. This kit is built to make those disagreements explicit and turn them into actionable decisions.

It combines reusable operating instructions, a context packet, a deeper reasoning guide, task prompts and an evaluation kit. It is a proposed framework ready for use and testing; no claim of proven forecasting superiority is made.

## Start with one real decision

1. Paste the master prompt into a new conversation or your assistant's persistent instruction area, if available. If there is a length limit, use the short version at the end of the task-prompts file and attach the full master as an explicitly adopted instruction document.
2. Fill the minimum context packet or start with your league link using the [supplemental league-access guide](08-league-access-guide.md). Start with exact scoring, starting slots, your roster/picks, the offer or question, and your objective.
3. Add the single task prompt matching the decision. Let the assistant research and identify consequential gaps.
4. Check the answer for the action, price boundary, before/after roster, decisive evidence and what would change the call.
5. Preserve the decision record before you know the result.

Do not paste the entire library into every question. Keep stable instructions available, load relevant league/current-state context, and bring in deeper modules only when needed. This follows the general context-engineering principle of curating high-signal information rather than accumulating everything. [Anthropic's context-engineering guidance](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents)

## What is in the kit

| File | Purpose |
|---|---|
| [01-master-prompt.md](01-master-prompt.md) | Complete reusable instructions for a dynasty decision partner |
| [02-context-packet.md](02-context-packet.md) | Minimum intake, full league/team/evidence templates and decision journal |
| [03-dynasty-reasoning-guide.md](03-dynasty-reasoning-guide.md) | The underlying strategic model, subtle traps and format switches |
| [04-task-prompts.md](04-task-prompts.md) | Fourteen task prompts plus a short operating prompt |
| [05-test-kit.md](05-test-kit.md) | Fair comparison protocol, rubric and twelve synthetic test cases |
| [06-test-answer-key.md](06-test-answer-key.md) | Expected test behavior; keep separate from the model being tested |
| [07-source-notes.md](07-source-notes.md) | Verified methodology/platform sources and the limits of this research |

These are the original eight documents, including this start-here file. The [supplemental league-access guide](08-league-access-guide.md) was added afterward to help you give an AI your league context without retyping everything; it was not part of the original one-shot experiment.

The complete ZIP now contains all nine documents. Extract it before following the individual file links above. The files use plain Markdown so you can edit or share them without a particular app.

## The four questions that organize everything

| Question | What a strong answer must distinguish |
|---|---|
| How good might the player become? | Ability, role, opportunity, health, competition and plausible outcomes |
| How much can I use? | Legal starting lineups, available replacements, coverage and cuts |
| What does the market actually show? | Preference rankings, ADP, completed trades, asks and current offers |
| What should I do? | My objective, price, alternatives, timing and the other manager's incentives |

Researching more articles does not solve a missing league rule. Better football analysis does not create a willing buyer. A balanced calculator does not show whether a package fills your lineup or clogs your bench.

## The shortest useful decision request

> Use the dynasty framework and this league context. As of [date/time], I am deciding whether to [exact action] at [exact price]. My objective is [objective and horizon]. Compare it with [alternatives] and holding. Verify the facts the research mode permits, rebuild the resulting lineup and roster, separate football from market value, and give me a verdict, a walk-away threshold, the strongest objection and what would change your mind. Ask only for missing information that could reverse the decision.

## The deeper standard

A player can be improving while becoming overpriced. A veteran can be declining while remaining worth more to your team than their available sale return. A pick can be attractive before a trade and less attractive after you strengthen its originating roster. A fair deal can help a direct playoff rival more than it helps you. A useful stash can stop being useful when the information you were waiting for arrives.

The assistant should recognize these possibilities, identify which matter to this decision, and still recommend an action. Complexity should make the answer more precise about its conditions, not less willing to help.

## What this package verifies

The cited source distinctions were checked during preparation. The synthetic arithmetic and package structure were checked. The test cases and rubric are prepared for evaluation; they are not evidence that any model passed, improved, or can predict dynasty outcomes reliably.