What Calibration Phase Means (And Why You Should Care)
Every MLB signal you see in the app right now has a label on it: Calibration. This is not a disclaimer. It is a statement about where the model is in its lifecycle.
Calibration phase means the model is live, generating real picks against real lines, and grading every result against the closing line. The picks are not hypothetical. They are tracked, timestamped, and published before the game starts. The only difference between calibration and deployment is how much live data the model has accumulated.
Why We Do This in Public
Most services build a model, backtest it on historical data, and launch it as "proven." Backtesting is necessary, but it is not sufficient. A model that looks profitable on three years of historical data can fall apart in its first month of live play for dozens of reasons: data pipeline delays, feature drift, market regime changes, or simple overfitting.
> Calibration phase catches these problems in real time.
You can see every pick the model makes and grade it yourself. You can watch the CLV numbers accumulate. You can see whether the model's edges are real or an artifact of sample noise. We are building the track record in front of you, not behind a backtest.
When Does Calibration End
We have internal criteria for promotion to deployment phase. Those criteria are based on sample size, CLV consistency, and edge-to-outcome correlation. We will not announce a specific date because the data decides, not the calendar. When the model earns deployment status, you will know.