Editors and research teams
People who repeatedly check fixture context, player availability and changing reports before producing a briefing. Their frequency of use and willingness to pay still need validation.
AI SPORTS INTELLIGENCE
/ 03
See the signal behind every market.
The first proposed EX Insight workflow helps sports editors and researchers prepare source-linked pre-match briefs. Model comparisons and enterprise delivery are later validation areas.
ILLUSTRATIVE INTELLIGENCEManchester City vs Arsenal
DEMOAI PROBABILITY52%
MARKET47%
MODEL–MARKET GAP+5 pp
Illustrative example — not live data or a recommendation
52%AI probability
47%Market probability
+5 ppModel–market gap
Illustrative example values only — not live data, advice or an indication of future results.
One fixture. A briefing you can check.
The proposed pilot focuses on sports editors and researchers preparing pre-match coverage. The task is to reconcile scattered updates into a source-linked briefing—not to provide a betting recommendation. The first competition and data sources would be agreed before a pilot.
Reconcile the information
Fix the event and cut-off time, preserve the source for each claim, flag conflicting reports and separate confirmed facts from interpretation.
A reviewable research brief
A concise summary, source list, unresolved questions and revision notes. The pilot would test whether this reduces checking effort without sacrificing accuracy.
This is a proposed pilot scope. The current website shows illustrative interfaces, not a connected research service or verified customer results.
From a conflicting update to a clear review task.
Imagine an editor preparing a Northbridge–Riverside preview. One report says a player will start; another says the decision is pending. This example explains the intended workflow, not an analysis generated by a live EX model.
Keep both reports
Attach source, publication time and the exact availability claim to the same fixture. Do not silently discard the conflicting report.
Mark the uncertainty
The brief should say that availability is unconfirmed and identify which official update would resolve the conflict. It should not invent a confirmed lineup.
Give the editor a next step
Present the two sources, a cautious summary and a recheck task. If the information changes, preserve a revision note for the reader to inspect.
Pilot measures: time to a reviewable brief, citation accuracy, unresolved-claim handling and repeat use. Record a manual-workflow baseline and agree thresholds before testing; no improvement rate is claimed yet.
PRODUCT IN PRACTICE
Put the capabilities in context.
Explore the intended information structure and workflow below. These examples are not live data or an available service.
DEMO-001 / FICTIONAL FIXTURE
Northbridge vs Riverside
A useful briefing shows what is known, what changed and what still needs confirmation.
- Research window
- Illustration: 30 minutes before kickoff
- Outcome scope
- Home win / draw / away win
- Data context
- Synthetic inputs, no live source
Questions to investigate
- Are lineup assumptions confirmed?
- Do all inputs refer to the same update time?
- Which change has affected the estimate?
The fictional teams and inputs explain the intended interface. They do not represent a real match or forecast.
ONE EVENT, TWO PERSPECTIVES
Compare estimates, not certainty.
| Outcome | Model | Market | Gap |
|---|---|---|---|
| Home win | 52% | 47% | +5 pp |
| Draw | 25% | 27% | -2 pp |
| Away win | 23% | 26% | -3 pp |
+5 percentage points
For the home-win outcome, 52% minus 47% is a difference in estimates. It is not a 5% return, an accuracy score or a guaranteed advantage.
Interpret the gap alongside freshness, assumptions and calibration. Actual market participation also involves costs and risk.
INTEGRATION CONCEPT
A signal needs its context.
{
"example_only": true,
"event_id": "DEMO-001",
"snapshot": "T-30m",
"source": "synthetic",
"model_version": "illustration",
"probabilities": {
"home": 0.52,
"draw": 0.25,
"away": 0.23
},
"lineup_status": "unconfirmed"
}Before a partner integration
Agree on entity identifiers, update timestamps, source rights, output fields and behavior when an input is missing. Partners should be able to trace an estimate back to a specific model and data snapshot.
This object illustrates information structure only. It is not a published API schema, endpoint or integration contract.
PRODUCT CAPABILITIES
Inside EX Insight.
Bring the match, the model and the market into one research workflow. Understand what changed, why it matters and how different probability estimates compare.
Intended product experience. Contact the team for current availability and access.
01AI match briefings
Review the fixture, recent performance and relevant match context alongside the factors behind a model estimate. Separate observed information from the model’s interpretation.
For a pre-match review, start with the competition, kick-off time, expected lineup and relevant historical window. If a player’s availability is unconfirmed, the briefing should preserve that uncertainty instead of silently treating an assumption as a fact.
EXPECTED OUTPUT
A structured briefing with key factors and update context.
02A consistent sports data view
Bring schedules, results and event state together with sources and timestamps, so comparisons refer to the same event and moment. Missing or delayed inputs should remain visible.
The intended view should make event identifiers, time zones and corrections easy to reconcile. Two records with similar team names are not necessarily the same fixture, and a historical result should not be confused with a current event state.
EXPECTED OUTPUT
Aligned event records, source context and freshness information.
03Probability comparison
Compare model estimates with market-implied probabilities. A 52% model estimate versus a 47% market view is a 5-percentage-point difference, not a guaranteed opportunity. Timing, assumptions and model error matter.
A meaningful comparison uses the same outcome definition and a comparable snapshot. A regulation-time result and an eventual tournament winner are different questions; comparing their percentages would create a misleading gap even if both figures were correctly calculated.
EXPECTED OUTPUT
Side-by-side estimates with an explicitly labeled difference.
04Expert context and change alerts
Combine quantitative outputs with human analysis. Follow material updates and inspect what changed before revisiting a view, rather than reacting to an unexplained score.
A useful alert explains the changed input, its source and how the interpretation differs from the previous version. Readers should be able to revisit the earlier context and decide whether the change is material to their original research question.
EXPECTED OUTPUT
Contextual analysis and a traceable history of meaningful changes.
05Enterprise API services
A planned integration route for structured event information and model outputs. Partners can discuss required fields, update frequency and permitted use for research tools, media products or internal workflows.
An integration discussion should cover event identifiers, timestamp format, versioning, update limits and correction handling. A small agreed sample can help a partner check fit before committing to a broader workflow; no live endpoint or service level is announced here.
EXPECTED OUTPUT
An integration scope agreed with the team; API access is not confirmed here.
USE CASES
From a real question to a useful output.
Pre-match research
A researcher needs to understand which inputs changed before kickoff, without rebuilding a briefing from separate sources.
INTENDED RESULT
A match brief with source timing, model estimates, material changes and unresolved questions.
Media and audience context
An editorial team wants to explain why a probability moved and which assumptions still matter.
INTENDED RESULT
Explainable context for editorial use, with observed facts separated from model interpretation.
Enterprise integration
A data team wants to place structured intelligence inside an existing research or reporting workflow.
INTENDED RESULT
An agreed field mapping, delivery scope and evaluation plan before an integration is enabled.
INTENDED WORKFLOW
A clear path, step by step.
- 01
Choose an event
Start with the match and outcome you want to understand.
- 02
Read the briefing
Review the relevant data, sources and latest update.
- 03
Compare perspectives
Examine model estimates, expert context and market pricing together.
- 04
Revisit after changes
Follow new information and reassess the underlying assumptions.
PREPARE TO CONNECT
Start with a clear brief.
This site demonstrates the information structure, not a connected data feed. Coverage, API documentation, pricing and access arrangements must be confirmed with the team.
Coverage and rights
Specify sports, competitions, historical depth and intended use. Confirm source permissions and permitted redistribution.
Delivery and freshness
Agree on event identifiers, fields, update frequency, freshness indicators and handling of delayed or missing inputs.
Evaluation before expansion
Review data quality and model calibration against agreed criteria. No accuracy, latency or service-level guarantee is specified here.
COMMON QUESTIONS
Know more before you act.
Does a higher model probability guarantee an opportunity?
No. The difference shows that the model and market disagree. Data timing, assumptions, estimation error and trading costs all affect its interpretation. A model is decision support, not a guarantee.
Which sports and competitions are covered?
A confirmed coverage list is not published here. Tell the team which competitions and data you need to discuss the current scope.
Can my team access an API?
API services are part of the product direction. Share your use case, required fields and expected update frequency to discuss access and integration requirements.
EX INSIGHT / NEXT STEP
View inquiry optionsDiscuss EX Insight access
Tell us which sports, analysis workflow or integration you are exploring.
Explore the technical architecture and product principles
FROM EVENT STREAM TO EXPLAINABLE PROBABILITY
Every output should show the probability, the change, the reason and the time.
EX Insight is being developed to turn fragmented sports information into a structured analytical workflow. Rather than offering a black-box score, it connects every probability to its context, update time and model version.
01
Intelligence pipeline
Relevant event and reference inputs are collected, normalized and transformed into model-ready features before an estimate reaches the interface.
- Input normalization
- Feature generation
- Versioned inference
02
Calibration, not certainty
A probability is an estimate. Historical evaluation, error analysis and uncertainty ranges help users understand where the model is strong and where conditions remain unclear.
- Backtesting
- Calibration analysis
- Uncertainty context
03
Multiple delivery modes
The same intelligence can support a concise match briefing, movement alert, scenario comparison or structured response inside a partner workflow.
- Briefings and alerts
- Scenario comparison
- Enterprise APIs
HOW IT WORKS
How an intelligence output is formed
A transparent pipeline makes it possible to inspect the path from raw input to user-facing probability.
01
COLLECT
Gather relevant inputs
Bring schedules, results, event state and context into a consistent stream.
02
TRANSFORM
Create model features
Resolve entities, align timestamps and prepare features for inference.
03
ESTIMATE
Generate probability
Run a versioned model and attach confidence and calibration context.
04
EXPLAIN
Publish the reasoning
Surface important changes, contributing signals and update time.
EX PRINCIPLE / FROM EVENT STREAM TO EXPLAINABLE PROBABILITY
Designed to make a model inspectable.
The product direction prioritizes clear assumptions, visible movement and decision support—not guaranteed outcomes.