Glossary

Reference

Definitions for terms used across consensus, picks, and performance.

Consensus & tiers

Consensus: Multiple models picking the same side in the same matchup. Higher agreement (e.g. 3 of 4 or 4 of 4) is treated as a stronger signal. We also use historical win rates for each combination of agreeing models.
Elite / 4-of-4: All tracked models agree on the same pick. Shown on consensus pages with the strongest agreement tier.
Verified / 3-of-4: Three of four models agree. Often used as a minimum bar for curated buckets; combo win rate is still considered.
Lean: Partial agreement (e.g. 2 of 4). Weaker signal; used for context rather than as a primary pick bucket.

Implied probability: The probability implied by the odds (e.g. -200 implies ~67%). Used to compare model confidence to market price.
Edge: When model-derived probability suggests a better outcome than the market price. Value buckets (e.g. Value Edge, Verified Value) filter for these spots.
Chalk: Heavy favorite. We avoid overexposure to chalk in some buckets to keep risk and payout in balance.

Unit (u): A fixed fraction of bankroll per bet—typically 0.5–1u per game. Units keep risk consistent and remove emotional sizing. See the Bankroll Calculator to set your unit.
Win rate: Wins divided by total decided games (W / (W + L)). Reported per bucket and for consensus overall.
Combo win rate: Historical win rate for a specific combination of agreeing models (e.g. when these 3 models agree). Used to weight which consensus levels are worth following.

For more detail, see Methodology and FAQ.