/* 2026-11-02 usability pass: prioritize fixtures, predictions and one ticket builder. */

INDEPENDENT RESEARCH ENVIRONMENT

Virtual Lab

eFootball-first research for GT and eAdriatic, with VFootball retained as a secondary observation lane.

Standalone path: /virtual-lab/ · No changes to the core Match Signal football/tennis pages. Paper research only.

MATCH SIGNAL Public owner: Diddy Build: MS-VIRTUAL-LAB-V1-2026.10.14-V3 · LIVE FEED v1 · Product-scoped participant research engine

PRIMARY RESEARCH LANE

eFootball first. Evidence before promotion.

GT and eAdriatic receive the main research focus. VFootball remains available as a secondary comparison lane, with its history preserved. No market is promoted by the UI.

Open prediction desk → System reports → Private AI Analyst →

Primary laneeFootball GT

Current fixtures, participant evidence, O/U line validation.

Second laneeAdriatic

Collected separately so GT volume cannot crowd it out.

Experimental marketsWinner · GG/NG

Observed and settled separately; blocked from automatic promotion.

Execution boundaryPaper only

Read-only market data and research. No live betting automation.

LAB PULSE

SportyBet feed → model → evidence

A live product view of what is being observed, calculated and currently eligible for research.

Live / fresh—

Research / watch—

Desk cards—

Settled accuracy—

Model regime—

Upcoming SportyBet fixtures

SPORTYBET NG · READ-ONLY

Connecting to live feed…

Waiting for SportyBet market data

eFootball is the default research view. VFootball and Zoom remain available as secondary observation lanes. Winner and GG/NG monitoring remains separate from validated Lab O/U eligibility.

Upcoming prediction desk

0 upcoming priced fixtures

Only research-eligible upcoming fixtures are shown. Refresh rechecks the live SportyBet feed, eligibility artifact, historical model state, and prediction calculations.

Not checked yet.

Waiting for live upcoming fixtures…

Near-Qualified Radar

0 monitored candidates · not promoted

The strongest upcoming raw-evidence candidates that are close to the research desk, but still blocked by a specific evidence/model condition. These cards never enter the Odds Builder until promotion is earned.

Waiting for current evidence.Upcoming raw-evidence candidates will appear here after the live feed and model state load.

Odds Builder

2–4 legs · current prices

Builder-ready: 0

Build a paper slip from qualified upcoming predictions.

0 legs

The builder uses the live SportyBet price attached to each displayed prediction. It does not place a bet or guarantee an outcome.

CONTROL ROOM

Research controls

Change scope before comparing evidence. Product families are kept separate.

Product

Market

Validation split

Load CSV / JSON

Observed rounds0

Win rate—

OOS win rate—

Paper ROI—

Sequence delta—

Automatic research history

CONNECTING…

The Lab is loading its automatically collected settled observations.

Research boundary This lab analyzes legitimately observed/publicly available results and published market information. It does not bypass RNGs, bookmaker security, private endpoints, or account controls.

EVIDENCE SUMMARY

Product fingerprints & signal gate

These panels summarise the selected product scope and its current research boundary.

Product fingerprint

No data

Load an observed dataset to compare product families.

Signal gate

PAPER ONLY

NO DATA

At most one research signal per calendar day. No automatic wagering.

VALIDATION

Sequence checks & unseen results

Exploration and out-of-sample evidence must remain separate.

Sequence / dependence tests

Lag-1

Waiting for data.

Out-of-sample test

—

Waiting for data.

Participant / player-team recurrence desk

Exact-line evidence · time-safe

Loading recurring participant evidence…

Product-specific O/U model

Walk-forward scoring

Loading line-ladder model evaluation…

Lab trace / health

V1 pipeline

Waiting for the Virtual Lab v1 renderer…

Candidate strategy laboratory

Discovery → freeze → unseen test

Low-threshold target

Tests a configurable low payout threshold from observed odds. This evaluates hit frequency and expected return; it does not assume the threshold is predictable.

Minimum odds

No strategy test has been run.

Research conclusion

Do not force a pick

Load data to generate a selection instruction.

How to use the Lab / what to select

Prediction → builder

1 — Pick the exact product

Match the bookmaker screen to GT eFootball, eAdriatic, Virtual Football or Zoom/Turbo. Never mix their statistics.

2 — Pick one market

Use the exact market and line shown by the bookmaker. A signal for 1X2 is not a signal for a different market or total line.

3 — Wait for a research-qualified candidate

The engine must have enough settled history, an untouched test period, stable results and positive paper economics. Raw win rate alone never qualifies.

4 — Match today's price

A historical signal can disappear at a worse price. Current odds must satisfy the candidate's validated price condition.

5 — Select only the displayed candidate

When the Lab shows RESEARCH-QUALIFIED, the card identifies the exact product, market and selection. Otherwise the instruction is NO SIGNAL.

6 — One daily paper signal

The workflow is capped at one paper signal per day. There is no automatic wagering or forced selection.

Dataset contract

Real-data only

Required fields

product, provider, event_id, timestamp, competition, participant_1, participant_2, market, selection, odds, result, score

For reliable settlement, each row should represent a completed event/market observation and carry the actual outcome. Do not mix discovery and test data manually.