machina-sports/skills/world-cup-intelligence

World Cup 2026 Intelligence Skill

This skill organizes and packages the real-time predictive analytics, market data aggregation, and social sentiment indexing capabilities of the FIFA World Cup 2026 pod into a cohesive, SDK-discoverable, and Studio-renderable capability.

Overview

The world-cup-intelligence skill wraps low-level background workflows (ingestion, identity crosswalking, Dixon-Coles model solving) into high-level, client-facing methods exposed via the @machina-sports/sdk or the Machina Studio UI.

By bundling these workflows under a unified manifest, we provide Studio operators with real-time "hot cards" (such as Market Watch and Fan Pulse) and conversational agents with verified data retrieval primitives.

---

Bundled Workflows

This skill exposes three primary executable workflows.

1. worldcup-market-watch

Generates a tournament-wide, composite market intelligence card showing odds movers, price spreads, and candidate arbitrage edges.

• Runtime Semantics: Serves a cached card unless force_regen is true or the cache is older than 20 minutes (TTL).
• Inputs:
  • force_regen (boolean, optional): Forces re-execution of the aggregation connectors and Gemini-based summary authoring.
• Outputs:
  • skill_card (object): A structured JSON object containing compiled tournament-wide statistics, top movers, and anomalies.
  • served_from (string): Either 'cache' or 'generated'.

2. worldcup-fan-pulse

Extracts real-time public sentiment, trending news storylines, and fan pulse for either a specific fixture or the tournament globally.

• Runtime Semantics: Relies on xAI Grok search to query the live Web/X index, summarizing results into a structured card. Cache TTL is 1 hour.
• Inputs:
  • event_urn (string, optional): The canonical Machina URN of the event (e.g., urn:machina:sport:soccer:event:scotland-vs-morocco:20260619:wor). If omitted, generates a global tournament pulse.
  • query (string, optional): Custom search parameters. Defaults to "FIFA World Cup 2026".
• Outputs:
  • skill_card (object): The structured pulse card containing sentiment dials, key themes, and source attribution links.
  • served_from (string): Either 'cache' or 'generated'.

3. worldcup-get-signal

Fuses our proprietary Dixon-Coles mathematical model's probabilities with the best real-time market prices across Kalshi and Polymarket.

• Runtime Semantics: Strictly read-only, informational decision support. Evaluates the model-implied probabilities against the line-shopped price per 1X2 outcome, outputting edge, EV, and Kelly recommendations.
• Inputs:
  • event_urn (string, required): The target fixture's URN (e.g., urn:machina:sport:soccer:event:scotland-vs-morocco:20260619:wor).
  • bankroll (string, optional): Simulated bankroll size; when supplied, each leg also returns a stake_amount.
  • kelly_fraction (number, optional): Kelly criterion scaling factor. Defaults to 0.25 (quarter-Kelly).
  • min_edge_bps (number, optional): Minimum net edge (basis points) for a leg to be flagged as value. Defaults to 200 (2%).
  • fee_bps (number, optional): Venue fee/slippage in basis points; edge, EV, and Kelly are computed net of it. Defaults to 0.
• Outputs:
  • signal (object): Per outcome — model_prob, fair odds (fair_decimal/fair_american), best_price+best_venue, edge/edge_pct, ev_per_dollar, kelly_full+kelly_stake, confidence_tier, and risk_flags.
  • recommendation (string): Standardized plain-text directive — e.g. "No actionable edge -- model and market broadly agree; pass." or "Value: back Draw at kalshi @0.28 -- model 35% vs market 28%, edge 7.1%, suggested stake 2.46% of bankroll (quarter-Kelly). Model confidence: low."
  • top_pick (object): The single highest-EV value leg (suppressed when flagged edge_likely_model_noise).

---

Setup & Execution

SDK Integration

To call these capabilities programmatically inside a fan app or service:

import { MachinaClient } from "@machina-sports/sdk";

const sdk = new MachinaClient({ token: process.env.MACHINA_API_TOKEN });

// Retrieve the hot tournament-wide market watch card
const watchCard = await sdk.skills.run("world-cup-intelligence", "worldcup-market-watch");
console.log(watchCard.skill_card.title); // "World Cup Market Watch"

CLI Execution

You can run these workflows directly via the Machina CLI:

# Force-regenerate the tournament market watch card
machina workflow run worldcup-market-watch force_regen=true

# Get betting signal for a specific match
machina workflow run worldcup-get-signal event_urn="urn:machina:sport:soccer:event:scotland-vs-morocco:20260619:wor"

---

Architectural Separation of Duties

To maintain long-term reliability and isolation, the World Cup Pod separates duties into three strict layers:

┌────────────────────────────────────────────────────────┐
│                      AGENTS LAYER                      │
│     (Reasoning loops: Copilots, Cron Publishers)        │
└───────────────────────────┬────────────────────────────┘
                            │ (chooses / orchestrates)
                            ▼
┌────────────────────────────────────────────────────────┐
│                      SKILLS LAYER                      │
│        (Discovered capability package: skill.yml)       │
└───────────────────────────┬────────────────────────────┘
                            │ (invokes)
                            ▼
┌────────────────────────────────────────────────────────┐
│                    WORKFLOWS LAYER                     │
│    (programmatic pipelines, model-solvers, ingestion)  │
└────────────────────────────────────────────────────────┘

1. Workflows (programmatic DAGs): Own raw data movement, calculations, and database state modifications (e.g., worldcup-ingest-fixtures, worldcup-sync-model-forecasts).
2. Skills (reusable packages): Bind specific workflows with human-readable guidelines, strict schemas, and UI elements. They are product-facing.
3. Agents (guided personas): Are non-deterministic reasoning loops that use Skills as tools to fulfill conversational goals.