GLOSSARY

GLINTARY Glossary

Plain-English definitions for the markets, odds and research terms used across GLINTARY.

Markets

The priced questions GLINTARY evaluates.

Moneyline: A selection on who wins the game or match, without a point or goal handicap. The line plays no part: only the winner decides it.
Match Winner: A provider label for selecting the winner of a match. In many sports this is functionally the same idea as a moneyline, although settlement rules can differ by sport and provider, so GLINTARY records the two separately rather than merging them.
Spreadalso: Handicap: A market that applies a virtual points, goals or games advantage or disadvantage to a side before determining the result. Also called a handicap. The figure applied is the line.
Totalalso: Over/Under, Over, Under: A selection on whether the combined score or statistic finishes above (Over) or below (Under) a specified line. Also called over/under.
Line: The handicap or total the selection is measured against. In “Under 49.5 at 1.94”, 49.5 is the line and 1.94 is the odds.

Odds and pricing

What each odds figure on the record means, and what it does not claim.

Recorded Odds: The decimal odds attached to the locked position, observed from the recorded source at the recorded time.
Best Observed Odds: The highest relevant odds GLINTARY observed in the recorded market evidence at that time. It describes what was observed, not what is available to any particular reader. GLINTARY defines this term but does not display a value for it. It cannot be reconstructed reliably from the immutable record, and a figure that cannot be verified does not belong on the public record.
Minimum Qualifying Odds: The lowest odds at which the position still satisfies the published GLINTARY research threshold. It is derived per candidate from that candidate's own de-vigged benchmark plus the policy's edge buffer and uncertainty margin; there is no global minimum.
Fair Odds: The de-vigged market-implied benchmark derived from the reference market. The de-vig removes the operator margin from the reference book; it is a benchmark for comparison, not a probability GLINTARY claims to know.
Market Reference: A neutral comparison line recorded alongside a position. It is not the Fair Odds benchmark and not the Recorded Odds, and it is never a recommendation.
Closing Odds: Comparable recorded market odds near closing, shown where reliable closing data exists for the same line.
Operator: A sportsbook or market source whose odds observation contributes to the market data. Where a count of operators is shown, it is the number of distinct operators whose quotes were represented for that selection when it was evaluated. It does not mean every one of them quoted the same figure, and naming an operator is evidence attribution, never a recommendation or a claim that the odds are available to you.

The engine

How a fixture becomes a candidate, and a candidate an official position.

Event: One fixture, game or match.
Market: One priced question about an event, such as the winner, a spread or a total.
Outcome: One selectable side or result within a market.
Coverage scan: One complete pass of the engine over the eligible pre-match schedule. It ingests the fixtures, pulls the odds for every supported market, archives the raw evidence and screens every selection deterministically. A scan decides nothing on its own: it produces the candidates that research and the policy gates then judge.
Screening: The deterministic first pass over every selection a scan collected. It applies fixed arithmetic rules about market depth, price against the de-vigged benchmark, freshness and event integrity. No model takes part and nothing is judged on merit here; screening only decides what is worth researching.
Initial candidate: A selection that passed deterministic screening and so became eligible for deeper research. GLINTARY takes at most one per event, the highest-ranked by its screening score. An initial candidate is not a position and has had no research applied to it yet.
Deep research: The complete research packet for one candidate. Seven agent roles contribute: sports data, intelligence, local sources, market context, source verification, an adversarial red team and a synthesiser. The result is one structured record of claims, their support, the strongest objection and the open questions. Research does not decide anything; the deterministic policy gates read the packet and produce the decision.
Official position: A candidate that passed every policy gate, including a final fresh-odds check, and was then locked, cryptographically committed and published before the event started. Only an official position appears on the Board, in the Ledger and in the performance record.
Declined after researchalso: PASS: The candidate received a complete research packet, the policy gates assessed it and it did not qualify for publication. Most candidates end here, which is the standard working rather than the engine failing. Internally this decision is recorded as PASS, which means the candidate passed *through* without a position, not that it passed the standard; the public wording avoids that trap.
Watchlist: A researched candidate that did not qualify for publication because only the price or recoverable data fell short, and whose case otherwise survived. The condition that would have to change is recorded with it. A watch condition is a statement of what is missing, never a prediction that it will be met.
Review required: A contradiction or presentation defect that code should not resolve on its own.
System blocked: Something the decision depends on, such as a data source or the archived evidence, was missing or unhealthy, so the candidate could not be assessed. It is never reported as a merit decision: “we could not assess this” and “we assessed it and declined” are different statements.
Lock: Freezing the complete position record before the event starts and publishing a cryptographic commitment to it. Nothing in a locked record is edited afterwards; updates, corrections and the result are appended next to it.
Evidence grade: A letter the research step records for the strength of the evidence behind a position, from A (strongest) through B to C, or UNGRADED when no grade was recorded. It describes the evidence that was available and confirmed, not the probability that the position wins.

Performance

How the public record measures itself.

Unit: GLINTARY uses a flat 1.00u risk for every official position. A unit is a neutral performance measure, not a currency amount. At decimal odds 2.18, a winning 1.00u position produces +1.18u profit; a loss produces -1.00u. Pushes and voids produce 0u. A flat 1u risk prevents discretionary stake sizing from distorting the public record: every position counts the same, so the record measures selection, not confidence.
Win: A settled position with a positive result. At a flat 1.00u risk a win returns the odds less the stake, so a win at decimal odds 2.18 is +1.18u.
Loss: A settled position with a negative result. At a flat 1.00u risk a full loss is -1.00u.
Push: A settled position with neither a win nor a loss, typically a handicap or total that lands exactly on the line. The stake is returned and the position contributes 0u.
Void: A position voided under the settlement rules recorded with it, for example a cancelled event or a market that never resolved. The position contributes 0u, stays on the public record and is counted separately from wins, losses and pushes rather than being removed.
Net units: Net Units = cumulative unit profit/loss across settled official positions. Only settled positions contribute; open positions count towards neither profit nor loss until they settle.
ROI: ROI = Net Units / Units Risked. Units Risked is the sum of the risk of every settled position, so with a flat 1.00u it equals the number of settled positions. Open positions are excluded from both sides.
CLValso: Closing-line value: Closing-line value compares the Recorded Odds with comparable closing odds for the same line and, where possible, the same operator. Comparable recorded market odds near closing, shown where reliable closing data exists for the same line. It is published only when a validated closing source exists and the sample reaches the published minimum, and it measures pricing, not whether a position won.

Where these terms appear

Every figure defined here is rendered from the public record itself. The GLINTARY Standard explains what a candidate has to survive before it is published, How It Works describes the pipeline, and the Live Engine shows what the engine is doing now, with every researched candidate and every completed scan.

34 terms. Definitions are versioned with the site and never rewritten to flatter a record.