Reading Line Confirmation Signals

BaseCase detects an edge by comparing a book's price against an internally computed fair value. That comparison can be wrong in three ways: the fair value is biased, the book's price was correct and BaseCase's was off, or the book is about to re-price and the edge is genuine but fleeting. Line confirmation is the validation step that distinguishes the third case from the first two by tracking what the rest of the market does after detection.

The validation problem

Edge detection is a one-shot computation. It tells you that, at time T, this book's price implied a probability that BaseCase's fair value disagrees with by some amount. It does not tell you whether BaseCase or the book is right. In an efficient market, an edge detected at time T should be partially confirmed by other books re-pricing toward the fair value at time T+1, T+5, T+15. If no other books move, the original detection was likely a false signal — BaseCase's fair value may have been off, or the book may have had information BaseCase didn't.

The Hasbrouck (1995) information-shares framework formalizes this: in a market with multiple competing price sources, each source's contribution to price discovery can be measured by how much subsequent movement loads onto its prior signal. BaseCase adapts this to per-edge confirmation tracking.

How BaseCase scores it

When the pipeline detects an edge on a book, it begins watching the same market on other books. Over the subsequent minutes, it tracks whether other books move toward BaseCase's fair value. The resulting confirmation_strength field is a number between 0 and 1, weighted by the historically measured reaction speed of each confirming book. A move from a slow-reactor book carries more weight than a move from a fast-reactor: slow books re-price last, so when they finally move, the signal is more reliable than when a fast book that always re-prices does the same.

The Edge Finder displays one of three badges based on the resulting score:

• Market validated (green) — line_movement_confirmed = true and confirmation_strength >= 0.7. Strong cross-book evidence the edge is real.
• Line moving (light green) — line_movement_confirmed = true and confirmation_strength < 0.7. Some confirmation, but partial.
• Unvalidated (amber) — line_movement_confirmed = false and confirmation_strength >= 0.5. The fast-reactor book did not confirm; treat with skepticism.

Edges with no confirmation signal at all render with no badge — the absence is informational on its own. Validation badges are gated to Sharp tier and above; they appear inline on every edge row for paid users.

Why slow-reactor confirmation is more informative

Fast-reactor books — those that adjust to news within seconds — confirm everything that's about to move. Their movement is necessary but not sufficient evidence of a true edge. Slow-reactor books that adjust only when several other sources have already moved confirm only the moves that the broader market converges on. A slow-book re-price is rarer and more informative per occurrence. The confirmation_strength weighting reflects this asymmetry directly: per-book reaction-speed coefficients are estimated from historical data and applied as multipliers on the per-book confirmation signal.

Worked example

Suppose BaseCase detects a +2.4% EV edge on DraftKings at 1:30 PM. By 1:33, FanDuel has moved 4 cents toward BaseCase's fair value, BetMGM has moved 2 cents, and Pinnacle has moved 6 cents. Pinnacle is a notoriously fast reactor; FanDuel and BetMGM are mid-speed. The composite confirmation score weights Pinnacle's movement lower (because Pinnacle moves on most signals, true or not) and FanDuel's higher (because FanDuel doesn't move unless it's the consensus direction). If the weighted total clears 0.7, the row is tagged Market validated.

Caveats

Line confirmation does not eliminate edge bias. If BaseCase's fair value is systematically biased on a particular sport or market type, the confirmation framework will systematically validate the bias along with the genuine edges. The badges should be read as one signal among many — alongside the fair-source badge, the cross-book dispersion estimate, and your own knowledge of the market. Confirmation is also forward-looking: an unvalidated edge may become validated in the next minute, or never. The badge reflects the state at the moment the row was rendered.

A confirmed edge is one the market has agreed with. An unconfirmed edge is one BaseCase is asking you to take on faith.

Further reading

• Why fair value beats consensus — what the confirmation framework is validating
• Expected value — the underlying signal being scored