Building the dataset.
simulation-powered · research-backed · paper-onlyGameTimePicks is a simulation-powered sports analytics platform. Behind the public 10,000-run game simulations, an automated research engine is quietly building a long-term, leakage-safe historical dataset — so that any future model can be evaluated against the market, honestly and out-of-sample.
Research milestones
- Automated pregame data capture
Every day, leakage-safe pregame features are captured for the MLB slate — starters, lineups, bullpen, matchup, park, team form, and more — each timestamped strictly before first pitch.
- Settlement pipeline
Official box scores from the MLB Stats API are joined to the pregame snapshots after games finalize, producing a clean, labeled research record.
- Observation quality validation
An automated quality gate checks every observation for stable IDs, correct outcomes, timestamp integrity, and no leakage before anything enters the dataset.
- Benchmark framework
A market-baseline benchmark is in place to evaluate any future model out-of-sample — always compared against the market first, never assumed better.
- Next milestone: 30 qualifying MLB observation dates
The dataset grows one finalized, market-covered slate at a time. We are building a long-term historical foundation so that any future model can be evaluated honestly.
Everything here is paper-only and educational. The public simulator is deterministic — the same 10,000-run result for every user — and is offered as an analytics tool for exploring probabilities and comparing them to the market, not as betting advice or a claim of superiority. The science continues in the background while the product is in public beta.