◉ SPORTS INTELLIGENCE AI
SPORTS EDITION · V1.0
MATCH CENTRE / FIVE-DAY OUTLOOK
Your five-day sports outlook
Live-source match centre, ten sport channels, player insights, verified outcomes and transparent prediction research in one place. Feeds clearly show freshness and coverage.
Research examples are not live fixtures. Fictional example events are provided to demonstrate the interface. Free scoreboard and event feeds are checked while this browser is open. Coverage varies by provider; player data and injuries are shown only when sourced and available. Sport-specific models are untrained, illustrative outputs, not validated predictive performance, and are not intended for wagering.
Expert intelligence & personal preferences
Independent research layerFollow the commentators, analysts and pundits you trust in each sport. Opinions are recorded with their source and timestamp, and are not silently incorporated into the statistical model. This version supports user-curated entries; it does not claim to retrieve or verify pundit commentary automatically.
Your expert preferences
Add a commentator or analyst
Record a sourced expert opinion
Record only attributable, public sporting analysis. The named event and predicted outcome should match the actual event you later evaluate.
Saved viewpoints
| Sport / Expert | Event | Expert forecast | Published | Original source | Actual result | Action |
|---|
Expert historical scorecard
Expert opinions
0
Scored outcomes
0
Historical hit rate
—
Comparison is descriptive only. Expert hit rates from different events, selections, sports or sample sizes are not directly comparable to model accuracy. No private-life rumours are collected.
Automatic fixtures · Rolling 5-day window
No-subscription public research feeds: ESPN's undocumented scoreboard and TheSportsDB's capped free API. Coverage and availability vary by sport; these are not exhaustive calendars. Availability, usage conditions and response fields can change without notice. Dates are displayed in your browser's timezone. No predictions are generated from live fixtures automatically: select a fixture to create an explicitly experimental forecast.
Not refreshed yet. Live feed updates start when the page opens.
| Local start | Sport | Competition | Fixture | Verified source status | Actions |
|---|---|---|---|---|---|
| Refresh to retrieve public fixtures. | |||||
Feed availability
Event intelligence
Provider-attributed onlyChoose Stats & availability from a fixture to retrieve the provider’s box scores, player statistics, lineup or injury information where available.
Automatic result matching
Audit trailForecasts linked to a verified event can be matched against an official final result when the source updates. Match and score must pass validation.
No feed check yet. Some sports require specialist data licensing or authorised APIs and will remain explicitly unavailable.
An empty list does not mean there are no events globally. Only configured leagues are checked. A feed may fail because of browser connectivity, rate limits, access policy or provider changes. Records are for performance research, not wagering.
◆ MATCH INTELLIGENCE · FIVE-DAY WINDOW
Inside the game.
Beyond the score.
Follow upcoming fixtures, track the big performance factors and compare match perspectives—across 10 dedicated sports.
Research scoreboard
Local forecast recordsRecorded forecasts
0
All categories
Awaiting results
0
Forecasts locked in history
Outcome accuracy
—
Resolved events only
Brier score
—
Lower indicates better calibration
Explore the sports world
🏇 South African Racing
Horse/rider recent form, draw, distance, course, weight, going, trainer trends and verified withdrawals.
🌍 Eight global sports
Each sport has a dedicated dashboard, research focus, forecast register and accuracy scorecard.
⚽ Soccer
Separate three-way outcome model: home win, draw and away win, with expected goals and tactical effects.
All 10 sports
Forecast register
| Date | Sport | Event | Forecast | Probability | Status |
|---|
South African horse racing
Model design: runner-level comparisons including recent speed figures, field strength, distance suitability, draw bias, weight, rider/trainer performance, course conditions and pedigree. For a true field of runners, probabilities should sum to 100% across the full declared field. The current experimental horse-racing model compares TWO runners only and cannot estimate full-field win probabilities. A full-field model is a planned extension.
Racing forecast history
| Event | Forecast | Probability | Status |
|---|
🏉 Rugby intelligence
Sport-specific research focus: Team ratings, scrum and set-piece form, squad availability, home advantage, recovery and travel. A dedicated experimental model now applies sport-specific factor weights and a separate probability transform. Inputs remain manual and require historical calibration.
Forecasts
0
Resolved
0
Outcome accuracy
—
Brier score
—
Rugby forecast history
| Event | Forecast | Probability | Status |
|---|
🏏 Cricket intelligence
Sport-specific research focus: Format-specific form, batting/bowling matchups, pitch, toss, weather and player availability. A dedicated experimental model now applies sport-specific factor weights and a separate probability transform. Inputs remain manual and require historical calibration.
Forecasts
0
Resolved
0
Outcome accuracy
—
Brier score
—
Cricket forecast history
| Event | Forecast | Probability | Status |
|---|
🎾 Tennis intelligence
Sport-specific research focus: Surface-adjusted Elo, serve/return performance, fatigue, recent opponents and verified injuries. A dedicated experimental model now applies sport-specific factor weights and a separate probability transform. Inputs remain manual and require historical calibration.
Forecasts
0
Resolved
0
Outcome accuracy
—
Brier score
—
Tennis forecast history
| Event | Forecast | Probability | Status |
|---|
🏈 American football intelligence
Sport-specific research focus: Quarterback availability, offensive and defensive efficiency, injuries, rest, travel and weather. A dedicated experimental model now applies sport-specific factor weights and a separate probability transform. Inputs remain manual and require historical calibration.
Forecasts
0
Resolved
0
Outcome accuracy
—
Brier score
—
American football forecast history
| Event | Forecast | Probability | Status |
|---|
🏀 Basketball intelligence
Sport-specific research focus: Net ratings, lineup combinations, player minutes, rest, back-to-backs and pace. A dedicated experimental model now applies sport-specific factor weights and a separate probability transform. Inputs remain manual and require historical calibration.
Forecasts
0
Resolved
0
Outcome accuracy
—
Brier score
—
Basketball forecast history
| Event | Forecast | Probability | Status |
|---|
🏐 Volleyball intelligence
Sport-specific research focus: Set-winning rates, serve/receive efficiency, blockers, rotations and player availability. A dedicated experimental model now applies sport-specific factor weights and a separate probability transform. Inputs remain manual and require historical calibration.
Forecasts
0
Resolved
0
Outcome accuracy
—
Brier score
—
Volleyball forecast history
| Event | Forecast | Probability | Status |
|---|
🤾 Handball intelligence
Sport-specific research focus: Goalkeeper save percentage, scoring efficiency, turnovers, depth and rest. A dedicated experimental model now applies sport-specific factor weights and a separate probability transform. Inputs remain manual and require historical calibration.
Forecasts
0
Resolved
0
Outcome accuracy
—
Brier score
—
Handball forecast history
| Event | Forecast | Probability | Status |
|---|
🏸 Badminton intelligence
Sport-specific research focus: Singles/doubles form, rallies, matchup tendencies, workload and competition conditions. A dedicated experimental model now applies sport-specific factor weights and a separate probability transform. Inputs remain manual and require historical calibration.
Forecasts
0
Resolved
0
Outcome accuracy
—
Brier score
—
Badminton forecast history
| Event | Forecast | Probability | Status |
|---|
Dedicated soccer engine
Soccer forecasts use a three-way outcome distribution (home win / draw / away win). A calibrated production model should also incorporate non-penalty expected goals, shot quality, squad availability, opposition-adjusted recent form, venue effects and fixture congestion. The simple manual prototype is for workflow testing, not professional match forecasting.
Soccer forecast history
| Match | Predicted outcome | Probability | Status |
|---|
Forecast lab
Enter a research forecast
Model preview
Enter factor assessments and click Preview probabilities.
Each sport now has an independent, transparent experimental rule set. Weights and probability transforms have NOT been estimated from historical results. Do not interpret their percentages as validated win probabilities.
Sport-specific model:
🏇 Full-field South African horse racing
MULTI-RUNNER MODEL · EXPERIMENTALEnter every declared runner in a race. Each runner gets four comparable research scores from −5 to +5: recent form, jockey/trainer/fitness, distance/draw/pedigree, and going/weight/course. Softmax normalisation makes the field probabilities sum to 100%. Scores and weights are assumptions until validated on independent historical races. Scratchings require a new forecast version; saved predictions are immutable except for the official result.
Race setup and runner field
| Runner | Form | Jockey / fitness | Fit | Conditions |
|---|
Race probabilities
Calculated for all currently entered runners; not calibrated or live.
Research only. No claimed edge over the field or market. Complete runner cards, official result verification and backtesting are necessary before judging predictive quality.
Historical backtesting & model validation
Metrics update when official outcomes are recorded or historical forecast CSVs are imported. They test already-locked probabilistic forecasts, not a freshly fitted model. To avoid hindsight bias, evaluate in date order using only information available before each event. Full-field races count as one event each.
Resolved forecasts
0
Multiclass Brier
—
Mean log loss
—
Top-choice accuracy
—
Results by sport
| Category | Completed | Correct | Accuracy | Brier | Log loss | Uniform baseline Brier |
|---|
Probability reliability
| Predicted top-choice confidence | Forecasts | Average stated probability | Observed correct rate | Difference |
|---|
Important testing controls
Rolling five-day windows
Analyze forecasts by scheduled event date and retain their original pre-event timestamp.
Comparable baselines
A simple equal-probability baseline is displayed. Future versions should include league strength, surface/Elo and bookmaker-free historical form baselines.
Out-of-sample checks
Avoid training or tuning on the same events used to declare accuracy. Compare consecutive held-out event windows and report sample counts.
Automated settlement audit
Source-event ID / exact team matchOnly completed events with a valid score and unambiguous competitors are auto-scored. Forecasts recorded after kickoff, fictional samples, ties unsupported by the model, and ambiguous fixtures remain unresolved.
Accuracy, scoring & audit
Record outcomes only after an event concludes. Metrics exclude pending results. Brier score is computed using the full categorical probability vector; for binary events it ranges from 0 (perfect) to 2 (worst), and for soccer's three-way outcomes the same formula applies. No event should be used to train a model if it is also in that model's held-out test set.
| Event | Predicted probabilities | Recorded result | Brier | Action |
|---|
Testing design
Baseline
Compare against a naive model, historical performance-only model and sport-specific statistical model.
Calibration
Group predictions by probability and check observed outcome rates. Track by sport and time period.
Incremental features
Add factors one by one and test whether each improves performance on unseen events.