Skip to main content
Back to challenges
Challenge

Multimodal Prediction Market Agent

This challenge focuses on building a highly sophisticated agent for analyzing real-time prediction markets. Leveraging Google's Agent Development Kit (ADK) with its strong multimodal capabilities, developers will create an agent that aggregates diverse data sources—market data, news feeds, social media sentiment, and expert opinions—to forecast event outcomes and identify potential arbitrage opportunities. This goes beyond simple data analysis to synthesize complex information for strategic decision-making. The system will use Claude Opus 4.5 for nuanced textual analysis and advanced strategy generation, interpreting qualitative market signals. Grok 4 Heavy will be integrated for high-performance, real-time complex pattern recognition and scenario simulation, enabling rapid assessment of market dynamics. Firebender AI will serve as a specialized service, offering advanced market data aggregation and pattern recognition as a tool that the Google ADK agent orchestrates. The agent will interact with a simulated Prediction Market API to fetch real-time contract data and execute simulated trades. This challenge emphasizes multimodal data fusion, multi-LLM orchestration, and the integration of specialized agentic services to create a powerful, real-time market intelligence system capable of providing actionable insights to 'pro gamblers' and financial strategists.

Special Purpose AgentsHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Design a multimodal prediction market analysis agent using Google ADK, Claude Opus 4.5, Grok 4 Heavy, and Firebender AI for real-time forecasting and arbitrage detection.

Delivery guide

How work is evaluated

Evaluation

The evaluation will focus on the agent's ability to accurately predict market outcomes, identify profitable arbitrage opportunities, and effectively integrate multimodal data sources.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
4 scoring checks
Binary
4 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

PredictionAccuracy

Agent must correctly predict the market outcome for at least 80% of test cases.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

ArbitrageIdentificationRate

Agent must correctly identify 90% of intentionally planted arbitrage opportunities.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 3

Average Prediction Confidence

Average confidence score for all market predictions made by the agent. • target: 0.8 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 4

Arbitrage Profitability (Average Margin)

Average profit margin of detected arbitrage opportunities. • target: 0.05 • range: 0-0.1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Master Google ADK for building multimodal AI agents, focusing on ingesting and processing diverse data formats (text, images, video snippets from news).

  • Implement multi-LLM orchestration, using Claude Opus 4.5 for sophisticated sentiment analysis from news and social media, and Grok 4 Heavy for high-throughput pattern recognition in numerical market data.

  • Integrate Firebender AI as a specialized agentic service tool, enabling the Google ADK agent to delegate complex market data aggregation and statistical pattern recognition tasks.

  • Design and build real-time data pipelines to ingest live prediction market contract data and news feeds from a simulated Prediction Market API.

  • Develop advanced analytical algorithms within the agent to identify subtle market inefficiencies and potential arbitrage opportunities across different prediction markets.

  • Implement robust evaluation metrics to assess the accuracy of market predictions and the profitability of identified arbitrage strategies.

  • Explore ethical considerations and potential biases in using AI for financial prediction markets, ensuring transparency in decision-making.

Resources and assets

Reference links and supporting material

Dataset notes

Sample data for 2 tasks

Challenge setup

The evaluation will focus on the agent's ability to accurately predict market outcomes, identify profitable arbitrage opportunities, and effectively integrate multimodal data sources.

Challenge input

{ "market_id": "M001", "historical_data": {...}, "news_articles": ["article_text"], "social_sentiment": "positive" }

Agent execution

No agent tool is listed for this challenge.

Evaluated output

{ "predicted_outcome": "[Outcome A]", "confidence": 0.85, "justification": "[reasoning]" }

Evaluation checks
  • Agent must correctly predict the market outcome for at least 80% of test cases.
  • Agent must correctly identify 90% of intentionally planted arbitrage opportunities.
  • Average confidence score for all market predictions made by the agent. • target: 0.8 • range: 0-1
Configured evaluation evidence
  • Average Prediction Confidence target: 0.8
  • 2 public reference cases
Configured runtime
  • Python execution harness

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

No tool records are attached.

Evaluation contract

  • PredictionAccuracy · Weight 1
  • ArbitrageIdentificationRate · Weight 1
  • Average Prediction Confidence · Weight 1
  • Arbitrage Profitability (Average Margin) · Weight 1

Metrics

  • Average Prediction Confidence · Target 0.8
  • Arbitrage Profitability (Average Margin) · Target 0.05

Reference cases

  • 2 public reference cases

Workflow tasks

  1. Task 1

    Input: { "market_id": "M001", "historical_data": {...}, "news_articles": ["article_text"], "social_sentiment": "positive" }

    Output: { "predicted_outcome": "[Outcome A]", "confidence": 0.85, "justification": "[reasoning]" }

    Execution harness: Python

  2. Task 2

    Input: { "market_data": { "platform_A": { "outcome_X_buy": 0.45, "outcome_X_sell": 0.52 }, "platform_B": { "outcome_X_buy": 0.48, "outcome_X_sell": 0.55 } } }

    Output: { "arbitrage_found": true, "strategy": { "buy": { "platform": "A", "outcome": "X" }, "sell": { "platform": "B", "outcome": "X" } }, "profit_margin": 0.03 }

    Execution harness: Python

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Multimodal Prediction Market Agent