Polymarket Player Goal Value

This skill implements a value framework for player-goal YES markets (World Cup + league + match-level props) using real player-level scoring data.

It models fair value using:

• real historical goals/90
• expected minutes per appearance (from real minutes/games)
• role from position (F/M/D/G)
• small penalty-goal uplift (from goals vs non-penalty goals)

Then it places patient limit buy ladders only when fair value edge exceeds a threshold.

Reference inspiration: https://x.com/Predicti0r/status/2061791808158400570

What it does

• Scans active player-goal markets from Simmer imports plus Polymarket sports-page premade combo legs
• Loads single-source player data from data/wc_players_filtered.csv generated by scripts/fetch_wc_stats.py
• Merges data/custom_players.json into the refresh pipeline for non-Understat players such as MLS/Eredivisie players
• Converts player stats into fair P(score >= 1) with a Poisson model
• Skips unknown players (no fallback guessing)
• Uses ask price for edge checks (fair - ask_yes >= min_edge)
• Applies spread/slippage quality gates + optional context safeguards
• Places laddered limit orders (GTC) at discounted prices from fair
• Enforces cooldown + daily budget controls
• Supports --venue (sim, polymarket, kalshi) for testing and deployment

Defaults

• Dry-run by default (no real orders)
• min_edge: 0.06
• max_position_usd: 12
• daily_budget_usd: 40
• max_trades_per_run: 3
• limit ladder: 8c / 5c / 3c below fair with 25% / 35% / 40% allocation

Run

cd skills/polymarket-world-cup-player-goal-value

# refresh player stats from the free Understat connector
python scripts/fetch_understat_players.py --seasons 2026,2025,2024 --min-minutes 300

# refresh World Cup stats and rebuild filtered player data (incremental cache)
python scripts/fetch_wc_stats.py

python player_goal_value.py --config
python player_goal_value.py --venue sim                  # dry run in $SIM venue
python player_goal_value.py --venue sim --positions      # inspect open sim positions
python player_goal_value.py --venue polymarket --live    # live orders on Polymarket

Tune

python player_goal_value.py --set min_edge=0.08
python player_goal_value.py --set max_position_usd=20
python player_goal_value.py --set daily_budget_usd=75
python player_goal_value.py --set limit_offsets_cents=10,6,4
python player_goal_value.py --set limit_splits=0.2,0.3,0.5

Notes

• Uses free Understat player data (top 5 leagues) for goals, non-penalty goals, minutes, and role.
• Non-Understat players are supported through data/custom_players.json; scripts/fetch_wc_stats.py merges those base stats with ESPN World Cup goals/appearances before writing data/wc_players_filtered.csv.
• Latest season in --seasons is included only if upstream sources already expose player rows (e.g. 2026 may be empty pre-kickoff).
• Unknown player markets are skipped instead of falling back to synthetic priors.
• Designed for low-liquidity conditions where market chasing is penalized.
• Start in dry-run/sim and tune expected_tournament_matches, expected_single_market_matches, expected_season_market_matches, min_edge, and minute filters after collecting outcomes.

Bugfixes applied in v0.1.8

• Added Polymarket Gamma /events/slug/{event}/premade-combos discovery so sports-page player-goal legs like fifwc-bra-jpn-2026-06-29 are found even when Simmer indexed search returns zero.
• Added SIMMER_WCPGV_COMBO_EVENT_SLUGS to focus runs on specific World Cup event slugs and auto-wire active ESPN game windows to Polymarket event slugs in adaptive_game_runner.py.
• Added regression coverage for Brazil/Japan premade combo player-goal import and event-slug focused discovery.

Bugfixes applied in v0.1.7

• Added game-window adaptive runner support so frequent cron ticks stay silent off-window and run only around relevant World Cup fixtures.
• Added player-data freshness gating and scan-limit controls so live/prep window runs do not repeatedly refresh slow stats or exceed scheduler timeouts.
• Added local client factory wiring for safer venue/live client construction in local dogfood runs.
• Refreshed bundled World Cup player data from the merged Understat + custom player + ESPN World Cup pipeline.

Bugfixes applied in v0.1.6

• Added fallback discovery search via /api/sdk/markets?q=... and merged + deduped results with snapshot feed to recover missing player-goal candidates.
• Added sim-only proxy quote option (allow_proxy_price_in_sim_only) so missing asks in $SIM no longer block all execution tests.
• Fixed venue-specific trade payload: only send price + GTC on polymarket; sim/kalshi now submit supported market orders.
• --positions now queries live account context so it reflects executed sim orders without requiring a separate --live flag.
• Existing v0.1.1 fixes retained: venue plumbing, dry-run state isolation, ask-based edge checks, SDK-managed price normalization.

Deterministic spec (Skill Builder style)

Signal

• Player fair-value estimate from weighted priors:
  • penalties, expected matches, minutes certainty, role centrality, mismatch upside

Entry logic

• Require fair_yes - ask_yes >= min_edge
• Build discounted limit ladder below fair value
• Enter only when spread/slippage/cooldown/budget gates pass

Exit logic

• v0.1 focuses on disciplined entry only
• Exit handling can be layered as explicit sell rules in future version

Market selection

• Active Polymarket-imported World Cup “player to score at least one goal” markets

Position sizing

• Fixed per-market cap max_position_usd
• Ladder split from limit_splits

Risk controls

• max_spread, max_slippage_pct
• cooldown_hours
• max_trades_per_run
• daily_budget_usd
• optional context safeguards (disable with --no-safeguards)