# Source notes and limits

Researched September 7, 2026. These sources support specific data, methodology and platform distinctions. They do **not** validate this entire framework or establish a dynasty forecasting edge. Most of the strategic organization, task prompts and test fixtures are proposed design choices synthesized for this kit.

## Market evidence

**KeepTradeCut:** its FAQ describes preference voting in a roster-neutral setting and asks voters to exclude perceived trade value. It separately documents a trade-frequency liquidity measure and calculator adjustments. Treat these as different signals. The FAQ also contains conflicting older and newer statements about TE-premium support, so inspect active settings when using the tool. [KTC FAQ](https://keeptradecut.com/frequently-asked-questions)

**FantasyCalc:** its FAQ describes inferring values from completed trades, outlier filtering, format adjustments and recency weighting. It also explains roster/waiver adjustments and greater sensitivity for infrequently traded elite assets. Its estimate is not a standing offer from a manager in your league. The rendered FAQ and indexed text were checked; plain text opening returned no content. [FantasyCalc FAQ](https://fantasycalc.com/frequently-asked-questions)

**Dynasty League Football:** the accessible methodology introduction distinguishes its mock ADP from actual MFL rookie drafts and notes differences in player pools. This supports labeling draft population and selection rules. The premium player-by-player analysis was not used. [DLF mock versus MFL ADP](https://dynastyleaguefootball.com/2026/04/18/mock-draft-scenarios-dlfs-mock-adp-vs-early-mfl-adp/)

**DynastyProcess:** the values documentation describes transformations of consensus rankings and assumptions for pick treatment; portions are explicitly labeled legacy methodology. Use it as an example of why a calculator's definitions and version matter, not as a universal valuation formula. [DynastyProcess market values](https://dynastyprocess.com/values/)

## League and football data

**Sleeper API:** documents read-only access to league settings, rosters, transactions, drafts and traded picks, including distinct original-franchise and current-owner fields. Platform documentation does not replace the league's complete constitution or an actual current-state check. [Sleeper API](https://docs.sleeper.com/)

**Sleeper reception scoring:** explains the interaction of base reception scoring, positional bonuses and primary position. Verify a league's settings before applying it. [Reception bonuses](https://support.sleeper.com/en/articles/3652730-how-are-reception-bonuses-calculated)

**Sleeper Max PF:** documents taxi-player inclusion in fantasy-football potential points. Do not transfer that rule blindly to another platform or to custom league bookkeeping. [Taxi and Max PF](https://support.sleeper.com/en/articles/4028606-do-taxi-players-count-toward-max-pf)

**nflverse:** the update schedule distinguishes dataset cadences and availability. At research time, it states that 2023-onward participation data arrives after the postseason, while also flagging an injury-source interruption after 2024. Thus a useful historical dataset may not supply current participation or injury evidence. Check the specific file's coverage and timestamps instead of assuming a library exposes live facts. This was a documentation check, not a live audit of every pipeline. [Data update and availability schedule](https://nflreadr.nflverse.com/articles/nflverse_data_schedule.html)

**NFL:** a source for official practice and game statuses. [NFL injury reports](https://www.nfl.com/injuries/)

**Footballguys / Joe Bryant:** the original value-based drafting explanation motivates comparison with positional peers under scoring and lineup rules. The specific dynasty extensions about feasible replacements, cuts and paths in this kit are our proposed application. [Principles of Value Based Drafting](https://www.footballguys.com/article/bryant_vbd?article=bryant_vbd)

## Context and evaluation

**Anthropic:** its context-engineering article supports curating relevant context, separating persistent guidance from retrieved information and managing growing histories. This kit applies that principle to dynasty; it does not assume a fixed universal context-window capacity or a magic prompt length. [Effective context engineering](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents)

**Zheng and collaborators:** their evaluation research describes limitations of model judges, including answer-order and verbosity effects. This motivates blinded, randomized comparisons and deterministic checks alongside expert review. [Judging LLM-as-a-Judge](https://arxiv.org/abs/2306.05685)

**Gneiting and Raftery:** their paper provides the foundation for proper scoring rules for probabilistic forecasts. This supports forecast evaluation; it does not validate subjective probabilities generated by an assistant. [Strictly Proper Scoring Rules, Prediction, and Estimation](https://sites.stat.washington.edu/people/raftery/Research/PDF/Gneiting2007jasa.pdf)

## Research limits

No authenticated dynasty league was analyzed for this package. No current player rankings or trade recommendations are asserted. No model-versus-model or baseline-versus-framework benchmark has been run. The fictional cases are designed to expose errors; their answer keys do not demonstrate a model passed them.

The screenshot supplied the motivating problem and the idea of combining current market research with dynasty context. It was not treated as a benchmark result or an instruction to post publicly. The framework is portable across assistants that can accept instructions and context; actual browsing, file retrieval, persistence and other capabilities depend on the environment.