Before any model goes live on Sharp Picks, it runs in shadow mode. Shadow mode means the model generates picks in real time against real lines, but those picks are never shown to users. They're graded silently in the background.
This is how we're preparing for MLB and WNBA.
Why Shadow Mode Exists
Backtesting is necessary but not sufficient. A model that looks great on historical data can fall apart in production for dozens of reasons: data pipeline delays, line movement it didn't train on, feature drift, or simply overfitting to the past. Shadow mode catches these problems before they affect your bankroll.
The Evaluation Gate
For WNBA, we're running shadow mode for the entire 2025 season. The model has to demonstrate profitability across that sample before going live. The gate is strict: we need to see positive CLV, a reasonable win rate, and no systematic blind spots in specific game types.
For MLB, the shadow period will cover the first portion of the 2026 season. Same evaluation criteria, applied to a larger sample given MLB's 162-game schedule.
What You'll See
When a new sport goes live, it'll show up in the app alongside NBA with its own signal feed, its own discipline filter stats, and its own track record starting from day one of live picks. The shadow period record won't be published as the official track record - it's testing data, not live performance.
This process isn't fast. That's intentional. We'd rather launch late with a validated model than launch early with a guess.