01
Data Collection
Connected sources provide schedules, prices, results, and the sport context available for each event. Fields that do not arrive stay unavailable.
EdgeDesk Methodology
How We Grade Projections
Inside The EdgeDesk research engine: source-attributed data, reproducible rules, and transparent states.
EdgeDesk Engine Flow
01
Data
02
Normalize
03
Features
04
Score
05
Gate
06
Record
The 6-Step Process
Accepted candidates pass through six stages before publication. Missing inputs and unresolved outcomes retain their real state.
02
Normalization
The system aligns identifiers, times, markets, and available statistics into canonical formats for deterministic processing.
03
Feature Building
The engine transforms available inputs into sport-specific signals. An absent signal is never replaced with an invented value.
04
Scoring Algorithm
The model scores each eligible candidate and, when a valid price exists, compares its estimate with the no-vig implied market probability.
05
Quality Gate
Not every model output is published. The gate applies current quality, product, season, and data-availability policy; it does not turn a model discrepancy into a proven edge.
06
Reporting & Grading Loop
Each accepted generation is stored with a content-addressed receipt. Outcomes are reconciled into the Power and Premium record since June 6 and, when enough data exists, closing-line calibration.
Philosophy
Core Principles
These four principles govern every decision the engine makes. They are not aspirational ideals — they are rules coded into the system.
01
No emotion
Every variable is processed mathematically. Feelings do not enter the model.
02
No forced projections
If the engine finds no value, we do not publish. We prefer no projection over a bad projection.
03
Receipt-backed record
Accepted generations and record summaries are hash-bound, and each product bucket stays separate.
04
Continuous calibration
We compare our confidence bands against real outcomes to constantly improve.
Transparency Without Excuses
We publish the available graded archive and a receipt-backed record summary, including losses and pushes. Scope and product buckets are explicit so the evidence can be reviewed.
Verifiable public record
Separate product record buckets
Ready to research with real data?
Create your account and review our methodology and the public archive of graded projections before choosing a plan.