Sports prediction model shows edge vs closing lines—but edge may vanish in real betting

The developer is building a sports prediction model and discovered a puzzle in the backtesting results. When they tested the model against closing lines—the final odds set immediately before an event starts—they found consistent edge, meaning the model's predictions outperformed those final lines.

However, in real deployment (inference time), there is a fundamental timing problem. The model makes predictions 12–24 hours before the event, at which point closing lines do not exist yet. Instead, the developer uses the current line available at prediction time. This creates a practical challenge.

The model's single strongest predictive feature is line movement—specifically, the implied probability shift from the opening line to the closing line. At backtest time, this feature is complete data. But at inference time, when predicting before closing lines form, the line movement feature is incomplete; the market has not finished moving yet.

This produces a paradox. Closing lines are considered nearly impossible to beat in the betting world because they contain all available information: sharp money (professional bettors), injury news, public sentiment, and everything else. Yet the backtest shows the model beating them consistently. If closing lines are truly efficient, then beating them should imply the model has found genuine signal. But at inference time, the model bets against earlier, less-efficient lines while using only a partial version of its strongest feature.

The unresolved question is whether the edge against closing lines will transfer to earlier bets where the lines are less efficient, or whether the incomplete line movement signal degrades prediction accuracy enough to wipe out any real edge.

The developer faces a classic challenge in quantitative sports betting: the gap between backtest conditions and live deployment. Closing lines are widely regarded as nearly impossible to beat because they represent the aggregate judgment of professional sharp money, injury reports, and all other public information—a standard benchmark for market efficiency. Yet the backtest shows consistent edge, which, if genuine, implies the model has extracted real signal.

The tension lies in the timing mismatch and feature incompleteness. The model's strongest predictor is line movement—the probability repricing that happens between opening and closing. In backtest, this feature is complete; in live prediction 12–24 hours earlier, it is incomplete. This creates two competing forces: betting against earlier, less-efficient lines (a potential advantage) while wielding a crippled version of the feature that drove the backtest edge (a potential disadvantage). Whether the edge persists depends on whether the model's signal is robust enough to overcome the feature degradation, or whether line movement was doing most of the heavy lifting in the backtest.