yahoo-fantasy

name	yahoo-fantasy
description	Analyze your Yahoo Fantasy sports team, evaluate roster health, compare free agents, and recommend add/drop moves. Use when user asks about fantasy team, waiver wire, roster moves, or player analysis.

Yahoo Fantasy Team Analyzer

Comprehensive analysis of the user's Yahoo Fantasy team with actionable roster moves.

Opt-in deep-dive phases

Load these detail files ONLY when the user's request triggers them. Skip otherwise — web searches are expensive.

Trigger keywords	Load file
"深度分析" / "advanced stats" / "regression" / "xStats" / "dig into [player]"	phases/advanced-metrics.md
"找潛力新秀" / "breakout" / "prospect" / "rookie" / "young player"	phases/breakout-scanner.md

Phase 1: Auth & league discovery

Call get_leagues. On auth error: authenticate (no args) → user pastes code= from redirect → authenticate(code=...) → retry.

If multiple leagues, ask which. Record league_key and my_team_key.

League settings (canonical cat list): get_league_settings(league_key) returns stat_categories with stat_id → display_name and sort_order (1 = higher better, 0 = lower better). Use this — don't hardcode or guess cats. Feeds Phase 4 matchup diff and Phase 7 category-fit rubric.

Player key lookup: search_players(league_key, name="...") returns player_key for stats calls. Never guess keys.

Phase 2: Roster analysis

2a. Week range

get_matchup(team_key) → week_start, week_end.

2b. Daily lineup scan

For each day in range, parallel-call get_roster(team_key, date="YYYY-MM-DD"). Per player per day record: selected_position, is_bench, is_injured_reserve.

2c. Schedule verification (avoid flagging off-days)

Cross-reference GP from Phase 3's lastweek result: if GP < active days, the gap is team off-days — not a lineup mistake. No extra API needed.

Flag decision table:

Situation	Flag?
BN + team had game + player was scheduled to play/start	Yes
BN + team had game + player not scheduled (e.g. SP on non-start day)	No — correct
BN + team had no game	No — off-day
BN + is_injured_reserve: true	No — IL slot is separate
DTD player active at risk	Yes
NFL bye-week player in active slot	Yes
NA-status player in active slot	Yes — blocks the slot
Pitcher in P slot vs SP/RP slot	No — stats score identically

2d. Today's lineup check

get_roster(team_key) (no date = today).

For every is_bench: true, is_injured_reserve: false:

• Hitter: flag if clearly stronger than active player at an eligible position
• Pitcher: flag only if confirmed start (SP) or save/hold opp (RP) today
• NA in active slot: always flag

Output per flag:

• Hitter: SWAP: [BN player] → [slot], bench [active player]
• NA: MOVE: [NA player] out of [slot] → BN, activate [BN player]

Phase 3: Player stats

Batch all rostered players in 2 parallel calls:

get_roster_stats(team_key, stat_period="lastweek")
get_roster_stats(team_key, stat_period="season")

get_roster_stats accepts season | lastweek | last30. last14 is NOT supported. last30 returns HTTP 400 in the first ~2 weeks of the season — skip if it fails.

Single-player fallback: get_player_stats(player_key, stat_period=..., date=...). Periods: season | week | lastweek | last14 | last30 | date (date requires date="YYYY-MM-DD").

Identify:

• Underperformers: lastweek >20% below season in key cats
• Hot streaks: lastweek well above season average

Phase 4: Current matchup

Reuse get_matchup from 2a. Note: score vs opp, cats ahead/behind, cats still live, streaming opportunities.

Phase 5: Free agents

get_free_agents(league_key, position="...", include_stats=true)

include_stats=true avoids separate stats calls. count must be an int — omit to use default (25).

Priority positions: injured spots, underperforming starters, streaming slots (RP in MLB, flex).

Ownership trend (percent_owned + percent_owned_delta):

• Rising >10% WoW → market sees breakout, act fast
• Falling >10% WoW but production still solid → possible dip-buy

Phase 6: News verification

Web search for: injured rostered players (return timeline), top FA candidates, unusual stat drops (role change / trade rumor).

Pattern: "[Player Name]" fantasy [sport] news injury

Yahoo injury status can lag — news is authoritative.

Phase 7: Scoring & report

Every add/drop candidate gets a composite 0–100 score so trade-offs are explicit. Trust the tier, not 2-point gaps — inputs include estimation.

Composite score

Dimension	Weight	Notes
Category fit (live matchup / team cat needs)	25	See rubric below — dominant factor
Production vs replacement	20	Season stats vs median FA at same position
Role / opportunity	15	Starter lock, closer role, bell-cow, target share secured
Trend	15	lastweek vs season; direction over last 14–30 days
Underlying agreement	10	xStats vs surface — ONLY if advanced-metrics phase ran
Positional scarcity	10	C / SS / TE → full 10; OF / WR / RP → 3–5
Schedule	5	Games left this week; ROS opponent strength

If advanced-metrics did NOT run, rescale remaining weights ×100/90 so max stays 100.

Category fit rubric (0–25)

This is what turns "good player" into "good FOR US, THIS WEEK". For each scoring cat in your league:

1. Get current matchup gap (you − opponent) from Phase 4.
2. Project the player's contribution this week (season rate × games remaining).
3. Apply matchup-state multiplier:

Cat state	Multiplier
Tight, gap ≤ 2 — swing cat	×3
Behind 3–5, games remain	×1.5
Leading 3–5	×1.0
Locked (ahead ≥ 6, OR behind ≥ 6 with no games left)	×0.2

Negative cats (ERA, WHIP, NBA TO): invert direction — a high-ERA pitcher actively hurts when you lead ERA.

Sum weighted contributions across all cats → normalize to 0–25.

Special cases:

• ROS / season-long questions (no live matchup): substitute matchup-state with your team's season cat ranks — bottom-3 ranks ×2, top-3 ×0.5.
• H2H-points / rotisserie leagues: collapse cat fit to "projected points / z-score". No matchup multipliers.

Tiers & swap threshold

Score	Tier	Action
85+	Must add	drop almost anyone
70–84	High priority	drop weakest bench / cold player
55–69	Solid	add if clear drop candidate exists
40–54	Streamer / watch	weekly plug-ins or injury cover only
<40	Pass

Recommend a swap only if ADD − DROP ≥ 15. Smaller gaps = churn, don't move.

Report layout

Roster health block:

ROSTER
Active healthy: ...
Day-to-day: ... (injury note)
IL/Injured: ... (return ETA, stash-worthy?)
Cold: ...
Hot: ...
Regression watch: ...  (only if advanced-metrics ran)

Scored move block (full breakdown for top 3, one-liner for rest):

ADD: [Player] ([Pos]) — [team]    Score: XX (Tier)
DROP: [Player] ([Pos]) — [team]   Score: YY (Tier)
Gap: +Z

Breakdown — ADD:
  Cat fit       22/25  (swings SB +3, HR +1)
  Production    15/20
  Role          14/15
  Trend         11/15
  Underlying     7/10
  Scarcity       3/10
  Schedule       5/5

Why ADD: [one line]
Why DROP: [one line]
Risk: Low | Med | High — [reason]

One-liner for lower-priority candidates: ADD X (72) / DROP Y (58) — gap 14, skip.

Streaming (if applicable)

MLB 2-start SPs, NBA back-to-back plays, NFL favorable matchups. Score using only Category fit + Schedule — short-term plays; other dimensions don't apply.

Sport-specific cheat sheet

MLB

• SP core: ERA, WHIP, K/9, W, QS
• RP core: SV, HLD, ERA, WHIP
• Bats core: AVG, HR, RBI, R, SB
• Drop: ERA > 5.00 (SP), AVG < .220 no power, lost closer role
• Add: new closer, 10+ day hot streak, call-up getting starts
• IL stash: IL slots are separate — hold high-upside injured players

NBA

• Core: PTS, REB, AST, STL, BLK, 3PM, FG%, FT%, TO (neg)
• Drop: bench role, minutes restriction, cold 2+ weeks no role security
• Add: injury-driven promotion, usage/minutes uptick, favorable b2b

NFL

• QB: pass yds/TD, rush, INT (neg)
• RB: rush yds/TD, rec, snap %, target share
• WR/TE: targets, receptions, yds, TD, air yards
• Drop: lost WR1 role, RB committee reduction, offense collapsing
• Add: handcuff after starter injury, emerging target share, favorable ROS

Final output structure

1. Quick summary (3–5 bullets) — include 1–2 line callout of which cats are swing cats this week
2. Roster status table
3. Top 3 scored moves (full breakdown) + one-liners for the next 3–5
4. Watch list
5. This week's focus

Keep it actionable. Injuries and significant underperformance > marginal upgrades.