Published · Updated
Closing line value (CLV) is widely treated as the best single predictor of long-term sports betting ROI. Beating the line just before first pitch correlates with future profitability more reliably than win rate alone. Here's why.
CLV compares your bet's odds to the closing line, the final price posted just before the game starts. If you bet a team at -110 and the line moved to -130 by close, you got +20 cents of CLV. If you bet at -130 and the line closed at -110, you got -20 cents (negative CLV).
The closing line is the market's most accurate read on a game, all sharp money has hit, all information is incorporated. If you're consistently beating the close, you're seeing the game more accurately than the entire betting market collectively.
Win/loss in any single sample is noisy. CLV is signal. Over a large sample, consistently positive CLV is a strong sign of long-term profitability, even when short-term W-L results lag.
CLV % = (closing_implied_prob - your_implied_prob) × 100
Or in dollar terms: track the difference between what you would have paid and what the line settled at. Positive CLV = you got a better price than the close.
If you consistently beat the close, you're effectively predicting where the line will move. Books interpret this as sharp money, even if the bet itself loses. Many US books limit consistently CLV-positive bettors, often by lowering their max-bet caps to recreational levels.
Moneyline picks log their price at publication time, and we compare it with the latest price our line-history feed recorded for that game. The result is written to our internal ledger. Coverage is partial: first-inning reads carry no price and so no CLV, most recent picks have no CLV value, and the track record page does not show CLV.
Per-bucket results (sport × pick type × confidence tier) tell us where the model is working. analytics/bucket_roi.py disables a bucket automatically when its recent ROI (or, for markets with no logged price, its hit rate against breakeven) falls below a set threshold, or when its recent Brier score worsens sharply. That check uses results, not CLV.
Theory is one thing, using the concept on tonight's slate is the value. These hubs are the practical application: