data feature

LLM-Ready Exports

Pre-structured context formats for Claude, GPT-4, Gemini, and local models. Feed odds data, line movement, and resolved outcomes directly into LLM reasoning pipelines without format wrangling.

context-export.sh

Claude context format · BOS @ MIA example

<game_context>
  GAME: BOS Celtics @ MIA Heat
  DATE: 2025-04-14T19:30:00Z
  LEAGUE: NBA

  SPREAD (FULL_GAME):
    BOS -3.5 | FD: -110, DK: -112, BM: -115
    LINE_OPEN: -4.5 | CURRENT: -3.5 | MOVED: +1.0 (toward MIA)

  TOTAL (FULL_GAME):
    o/u 224.5 | FD: -110/-110, DK: -110/-110
    LINE_OPEN: 222.5 | CURRENT: 224.5 | MOVED: +2.0

  MONEYLINE:
    BOS: FD -165, DK -170 | MIA: FD +140, DK +145
</game_context>

What to build with LLM-ready data

Game analysis pipelines

Feed structured game context into an LLM to generate pre-game analysis, flag line movement anomalies, or summarize betting trends across a slate of games.

eg: Ask Claude to identify steam moves across tonight's NBA slate

Model explanation

Use LLMs to explain your quantitative model's outputs in plain language. Feed in the odds data and model predictions, get back a human-readable rationale.

eg: Explain why the model likes BOS -3.5 given the line movement

Data Q&A

Build a natural language interface to your betting data. LLMs with structured context can answer questions like 'which spread favorites beat the close last week?'

eg: Query historical resolved data in plain English

Alert generation

LLMs can interpret line movement data and generate plain-language alerts: 'Sharp money hit the BOS total, moved 2pts in 30 minutes across all books.'

eg: Auto-generate steam alerts from tick data

What's included in each export

Game metadataTeams, date, league, venue

Current oddsAll books, all market types, best available

Line movement summaryOpen, current, and direction for each market

Resolved outcomeWin/loss/push result (for historical exports)

CLV snapshotClosing line vs your odds if applicable

early access

Plug odds data straight into your LLM

Claude, GPT-4, Gemini, and local model formats. No preprocessing required.

Join Waitlist →