Skip to content

Methodology

How Otterline Consensus Works

Otterline uses league-specific pregame pipelines. It records the signals available before a game, preserves the market context captured at that time, and grades published decisions from persisted outcomes. The number and type of sources vary by sport.

Current league structure

Sources are collected separately, but Otterline does not claim they are statistically independent. The current source abbreviations and available predictions are shown on each league's consensus board.

LeagueExpected signalsPublic scope
NHL4Four collected pregame model signals
NBA3Three collected pregame model signals
MLB3Three collected pregame model signals
NFL3Three collected signals with a sealed consensus record

What the numbers mean

Model agreement

The number of available model signals selecting the same side. Agreement is a count, not a probability or guarantee.

Model confidence

A probability produced by a specific model when that source supplies one. It is kept separate from market pricing.

Market-implied probability

The price-derived probability captured from an external market. It provides context and is not presented as an Otterline model score.

Historical combo rate

The win rate of a specific group of agreeing signals over previously graded games. It is shown with its sample size when available.

From evidence to a tracked result

  1. 1Collect the scheduled game and available source predictions before the league cutoff.
  2. 2Reject or clearly mark missing, stale, ambiguous, or post-start evidence.
  3. 3Apply that league's current qualification rules and seal any published consensus before the game.
  4. 4Keep the market snapshot available at the decision time as context and, where required, for unit calculations.
  5. 5Grade the sealed decision from a persisted final outcome without changing the original prediction.

Important limits

  • Agreement does not prove an edge and does not guarantee an outcome.
  • Market percentages are not reused as Otterline model confidence.
  • Historical rates describe the recorded sample shown; they are not a promise of future performance.
  • League records and unit figures can use different evidence scopes, which are identified on Results.