Premier League12:30 BST
Arsenal vs Liverpool
Not yet analysedAnalyze →
EdgeXI · Quantitative sports intelligence
Find the edge.
Every number shows its work — the model trace, the evidence, the sources, and the markets where the answer is no bet.
DEMO FIXTURESEngine not connected · billing closed
- Data
- Model
- Probability
- Market
- Edge
Live edgeMOCK DATA — not live prices
- SOCTotal Goals61%+8.6 pp
- NFLGame Total60%+7.6 pp
- SOCCorners58%+5.1 pp
- NFLSpread57%+5.7 pp
- SOCShots on Target57%+4.4 pp
- NFLPassing Yards59%+6.4 pp
- SOCAsian Handicap54%+2.7 pp
- NFLTeam Totals56%+4.2 pp
- SOCDouble Chance82%+3.9 pp
- NFLRushing Yards56%+3.7 pp
- SOCTeam Goals63%+4.8 pp
- NFLMoneyline57%+3.0 pp
- SOCShots52%+0.3 pp
- NFLReceiving Yards51%+0.2 pp
- SOCTotal Goals44%-4.8 pp
- NFLTeam Totals45%-3.1 pp
How an edge is found
One path · no shortcuts- 01
Data
Fixtures, form, minutes, venue.
- 02
Model
Baseline, then matchup adjustment.
- 03
Probability
One number, assumptions attached.
- 04
Market
Where the price sits.
- 05
Edge
The gap, in points.
The work happens before the whistle.
Training load, minutes, venue, conditions — the inputs the model is actually built on.
Product tour
One frame. Five surfaces.
Terminal surface · demo sessionMOCK DATA — demo fixtures
03 — Full scan
Full scan: every evaluated market, ranked by edge.
One table per game: RANK · MARKET · PICK · LINE · ODDS · MODEL · PROB. · EDGE · CONF. · RATING. The ranking is EdgeXI's. The values shown on this page are demo fixtures, and the table says so on its own header.
| Rank | Market | Pick | Line | Odds | Model | Prob. | Edge | Conf. | Rating |
|---|---|---|---|---|---|---|---|---|---|
| 01 | Total Goals | Over 2.5 | 2.5 | 1.91 | 2.87goals | 61% | +8.6 pp | 78% | STRONG |
| 02 | Corners | Over 10.5 | 10.5 | 1.87 | 11.4corners | 58% | +5.1 pp | 70% | MODERATE |
| 03 | Shots on Target | Over 9.5 | 9.5 | 1.90 | 10.8shots | 57% | +4.4 pp | 68% | MODERATE |
| 04 | Asian Handicap | Arsenal -0.5 | -0.5 | 1.95 | -0.7 | 54% | +2.7 pp | 66% | MODERATE |
| 05 | Double Chance | Arsenal or Draw | — | 1.28 | — | 82% | +3.9 pp | 74% | MODERATE |
| 06 | Team Goals | Arsenal Over 1.5 | 1.5 | 1.72 | 1.9goals | 63% | +4.8 pp | 71% | MODERATE |
| 07 | Shots | Over 24.5 | 24.5 | 1.93 | 25.1shots | 52% | +0.3 pp | 58% | NEUTRAL |
| 08 | Total Goals | Under 3.5 | 3.5 | 2.05 | 2.87goals | 44% | -4.8 pp | 64% | AVOID |
8 evaluated markets · ranked by edge in percentage points · missing values render “—” and are never substituted
| Rank | Market | Pick | Line | Odds | Model | Prob. | Edge | Conf. | Rating |
|---|---|---|---|---|---|---|---|---|---|
| 01 | Game Total | Over 47.5 | 47.5 | 1.91 | 51.2points | 60% | +7.6 pp | 75% | STRONG |
| 02 | Spread | BUF -3.5 | -3.5 | 1.95 | -5.1points | 57% | +5.7 pp | 71% | MODERATE |
| 03 | Passing Yards | Josh Allen Over 267.5 | 267.5 | 1.90 | 284.2yards | 59% | +6.4 pp | 72% | MODERATE |
| 04 | Team Totals | Bills Over 26.5 | 26.5 | 1.88 | 27.4points | 56% | +4.2 pp | 69% | MODERATE |
| 05 | Rushing Yards | James Cook Over 68.5 | 68.5 | 1.92 | 73.8yards | 56% | +3.7 pp | 67% | MODERATE |
| 06 | Moneyline | Bills | — | 1.85 | — | 57% | +3.0 pp | 68% | MODERATE |
| 07 | Receiving Yards | Khalil Shakir Over 54.5 | 54.5 | 1.95 | 56.1yards | 51% | +0.2 pp | 60% | NEUTRAL |
| 08 | Team Totals | Chiefs Under 23.5 | 23.5 | 1.98 | 24.1points | 45% | -3.1 pp | 65% | AVOID |
8 evaluated markets · ranked by edge in percentage points · missing values render “—” and are never substituted
Markets evaluated for these two modules: Total Goals · Corners · Shots on Target · Asian Handicap · Double Chance · Team Goals · Shots · Game Total · Spread · Passing Yards · Team Totals · Rushing Yards · Moneyline · Receiving Yards. NO BET is a row like any other, and it is never hidden on a lower tier.
02 — The workspace
One workspace per sport. One scale across all of them.
Soccer and NFL are modules inside one application, not two products. Choose the sport once and every surface follows: games, leagues, markets, terminology, metrics and team cards. The scan table is the same table. The rating scale is the same scale. The trace format is the same format — what changes is the analysis.
Terminal previewMOCK DATA — demo fixtures
The same shell, the same cards, the same scale — the sport is a choice, not a different product.
NFL · Week 716:25 ET
Buffalo Bills @ Kansas City Chiefs
Not yet analysedAnalyze →
Also inside: the ranking scale STRONG / MODERATE / NEUTRAL / AVOID / NO BET is one scale with thresholds in the shared domain layer, so a new sport module needs no new mental model.
Planned modules
Declared in the registry, not yet routable.
NBA · not registeredMLB · not registeredNHL · not registeredCollege Football · not registeredCollege Basketball · not registered
04 — Model trace
Every number shows its work.
Show the model, or the number means nothing. Every headline estimate carries a WHY?. Open it and the estimate is rebuilt in front of you: the inputs, the league baseline, the matchup adjustment, the final estimate, then probability, market line, edge and sources. The same shape holds in every sport.
Josh Allen passing yards — model trace
formula shown
- QB passing average281.4yards
- Opponent passing allowed254.2yards
- League baseline267.8yards
- Matchup adjustment+5%
Final estimate281.2yards
(QB passing average × opponent passing allowed) ÷ league baseline → matchup adjustment
Assumptions stated by the model
- Primary receivers active
- Neutral game script (no large early deficit)
WHY?show the chainshrink
- 01Prediction
- 02Historical data
- 03Statistics
- 04Baseline
- 05Matchup adjustment
- 06Final estimate
- 07Probability
- 08Market line
- 09Edge
- 10Sources
The intermediate steps are the engine's, not the interface's. If a step is missing, the number is not shown.
FixtureMOCK DATA — demo fixture
A passing projection, traced: quarterback passing average × opponent passing allowed → league baseline → matchup adjustment → final estimate. The same structure carries a corner count, a game total or a goals model.
05 — Evidence and inference
Evidence is not inference. The interface says which is which.
The first is data. The second is the model's reading of that data — an interpretation, with conditions attached, and it is styled as one. Nothing the model infers is ever dressed as something that was measured.
| Evidence | QB passing average 281.4yards | Opponent passing allowed 254.2yards | Market line 267.5 |
|---|---|---|---|
| Inference | The matchup adjustment lifts the baseline by +5%, and that adjustment assumes a normal snap distribution and no material change in participation. | ||
Recorded facts supplied by the analytical layer, each with its category, sample size and source reference.
[01]Allen has cleared his passing line in 4 of the last 5 games.
4/5player-trendN=5Source src-provider-stats[02]Bills have scored 27+ in three of their last four games.
3/4team-formN=4Source src-league-official[03]Chiefs allow 254.2 passing yards per game.
254.2defensive-profileSource src-provider-stats[04]No precipitation forecast; wind under 15 km/h.
<15 km/hweatherSource src-weather
Interpretation, not fact. Each item states the weight the model gives it and what would falsify it.
Neutral or positive game script keeps passing volume near season levels.
Weight58%Falsified ifA two-score Bills lead in the third quarter would reduce pass attempts.
Pace may increase if both teams trade early scoring drives.
Weight52%Falsified ifSensitive to turnover outcomes.
Removing colour does not remove the distinction: a solid rail with numbered citations against a dashed rail with a MODELLED stamp.
06 — Sources
Every analysis names its sources.
Source, data, last updated, link and a reliability class. If a source is stale, contradictory or unavailable, the panel says so rather than dropping the row. An analysis you cannot trace back is not an analysis; it is a claim.
| Source | Data | Updated | Link |
|---|---|---|---|
League official statistics official-leagueprimary | Goals, shots, corners per match | 19 Sept | — unavailable |
Statistical provider feed statistical-providersecondary | Team and player per-90 baselines | 19 Sept | — unavailable |
Market odds sample sportsbooksecondary | Decimal prices across tracked books | 19 Sept | — unavailable |
Stadium weather forecast weatherprimary | Wind, precipitation, temperature at kickoff | 19 Sept | — unavailable |
“—” means the fixture supplies no link. A live row states an unavailable source explicitly rather than omitting it. Reliability classes: official league data · team data · statistical provider · sportsbook · weather · injury report · news.
07 — No bet
The model is allowed to say no bet.
No statistical edge detected — that is a valid analytical result, and the product treats it as one. No edge is a finding rather than a failure: it is not hidden in a smaller font, and it is never replaced with a lesser pick. A tool that can only ever produce a selection is not telling you anything about the market.
Best edge · Total Goals
STRONGOver 2.5
Model probability advantage is large relative to variance.
- Probability
- 61%
- Odds
- 1.91
- Edge
- +8.6 pp
- Confidence
- 78%
Line2.5CONF HIGH
EDGE+8.6 ppMOCK DATA
Best edge · none
NO BETNO BET
No statistical edge detected.
Modelled probabilities do not exceed market-implied probabilities by a material margin in any evaluated market. No statistical edge detected.
Valid analytical result — not an errorMOCK DATA
No statistical edge detected — valid result
NO BET appears on every tier, including Free: it is not a coverage limit and it is not styled as a failure. It is the same neutral card, with the same reasoning available behind WHY?.
Scrolls inside the frameOpen the full terminal →
Every evaluated market, ranked by edge — including the ones rated no bet.
One workspace, two sport modules: the same shell, cards and rating scale.
Every estimate rebuilt step by step, from raw inputs to market edge.
Measured facts kept apart from modelled judgement, each line sourced.
No statistical edge detected is a valid result, not a hidden failure.
Coverage
Every surface, drawn to scale.
Night and day
Night or day. Same instrument.
Dark by default, inverting to a daylight palette for desk work — same density, same colour logic, same tabular numerals. STRONG stays green, MODERATE amber, NEUTRAL grey, AVOID red.
NIGHT58% modelled
Corners · Over 10.5MODERATE
- Line
- 10.5
- Odds
- 1.87
- Prob.
- 58%
- Edge
- +5.1 pp
DAY58% modelled
Corners · Over 10.5MODERATE
- Line
- 10.5
- Odds
- 1.87
- Prob.
- 58%
- Edge
- +5.1 pp
Token set · identical names, two palettes
NIGHT
- --color-base
- --color-surface
- --color-panel
- --color-sunken
- --color-line
- --color-line-strong
- --color-ink
- --color-muted
- --color-accent
DAY
- --color-base
- --color-surface
- --color-panel
- --color-sunken
- --color-line
- --color-line-strong
- --color-ink
- --color-muted
- --color-accent
Rating colours re-derived, semantics unchanged
NIGHT
- STRONGthreshold in core
- MODERATEthreshold in core
- NEUTRALthreshold in core
- AVOIDthreshold in core
DAY
- STRONGthreshold in core
- MODERATEthreshold in core
- NEUTRALthreshold in core
- AVOIDthreshold in core
Colour is never the sole carrier: every state also carries a glyph and a label.
FixtureMOCK DATA — demo fixture
Applied before first paint — no flash, no layout shift.
Pricing
Three tiers. One is enough to start.
Free reads the headline layer. Pro opens the analysis. Elite removes the ceilings. Billing is not open yet — nothing here can be purchased and no card is requested anywhere.
Free
START HERE
$0
The headline layer: today's games, the best edge on each, and enough of the scan to judge whether the method is worth your attention.
Ideal for First-time readers evaluating EdgeXI before paying, and occasional readers who want one ranked number per game.
Free with an account. No card, no trial clock, no expiry — Free keeps the headline layer for as long as the account exists.
Create a free account
Opens sign-in. Nothing is charged and nothing is stored beyond your account.
- Coverage
- Dashboard with today's games across Soccer and NFL
- Game pages with the headline assessment and team projections
- Best Edge card: market, pick, probability, odds, edge, confidence
- Live / Games board, global search and the sport selectors
- Analysis
- Full Scan ranked table — top 5 selections per game
- Overview market page only; other markets per sport are locked
- Rating scale STRONG / MODERATE / NEUTRAL / AVOID / NO BET, identical across sports
- NO BET results are never hidden on a lower tier
- Trace and provenance
- Model trace headline: final estimate and probability shown
- Baseline, matchup adjustment and stated assumptions locked
- Evidence vs inference and SOURCES panels locked
- History
- Session only — runs are not saved and do not carry across devices
- Edge AI
- 5 questions per day, counted server-side per account
- Prompt suggestions are scoped to the game and market in view
- Engine
- Standard scan queue
- Results may be served from a shared cache for up to 10 minutes
- Not included
- Player props and per-player projections
- Saved history and exports
- Priority engine queue
- New sport modules on arrival
Pro
RecommendedMOST USED
$29 / month
The full analysis on both sports, for readers who check the work before they use the number.
Ideal for Regular readers who work one or both sport modules several times a week and want the reasoning, the sources and the saved runs.
Billed monthly when paid plans open. Not purchasable in this release — no card is requested and no checkout exists on any EdgeXI page.
Request Pro access
Opens the access request — an email address, and nothing else. It is a mailing, not a sale.
- Everything in Free
- All Free coverage, analysis states and rating behaviour
- Coverage
- All Soccer markets: Total Goals, Double Chance, Asian Handicap, Team Goals, Corners, Shots, Shots on Target
- All NFL markets: Spread, Moneyline, Game Total, Team Totals, Player Props, Passing, Rushing, Receiving
- Full Scan across every evaluated market with no row cap
- On-demand refresh when a cached result is older than the refresh window
- Trace and provenance
- Full MODEL TRACE: baseline, matchup adjustment, final estimate, stated assumptions
- WHY? on every headline estimate, expanding the full chain to sources
- EVIDENCE VS INFERENCE panels, visually separated
- SOURCES table: source, data, updated, link, reliability class
- Players
- Player analysis cards with recent-performance samples
- Passing / rushing / receiving / targets, each with its own line and edge
- History
- 100 saved analysis runs, searchable, newest first
- CSV export of the full scan table
- Edge AI
- 200 questions per day
- Follow-ups keep the game and market context
- Not included
- Unlimited saved history and full-payload JSON export
- Priority position in the scan queue when the live engine is enabled
- Three seats and shared workspace history
- Future sport modules at no extra cost
Elite
ANALYST
$79 / month
No ceilings: the whole history, the whole payload, and the front of the queue when the live engine is running.
Ideal for Analysts and small teams who archive the analysis, compare runs across a season, and want the structured payload itself.
Billed monthly when paid plans open. Display-only today: no checkout, no card field, no charge, no trial clock running against you.
Request Elite access
Opens the access request — an email address, and nothing else. It is a mailing, not a sale.
- Everything in Pro
- All Pro coverage, trace, provenance, player and Edge AI capabilities
- History and export
- Unlimited saved analysis runs per workspace
- Full-payload JSON export of the structured response behind each run
- Edge AI and refresh
- Unlimited Edge AI questions, subject to a fair-use rate limit shown in Settings
- Longer-lived results with full refresh history per market
- Engine
- Priority position in the scan queue when the live engine is enabled
- Engine-incident notices, so a failed scan is explained rather than silently retried
- Coverage as it ships
- Every new sport module at no extra cost as it registers — NBA, MLB, NHL, College Football, College Basketball
- Workspace
- Three seats in one workspace with shared saved history
- Audit view of who ran which scan and when
- Support
- Direct email line for methodology and data-provenance questions
- Not included
- Anything not yet built is listed on the roadmap, not sold here as included
Paid plans are display-only in this release. Nothing on this page takes a payment. No card is requested anywhere.
Ask a question instead →Access
Read the numbers first.
Walk one game end to end — the scan, the trace, the evidence, the sources, and the markets where the model says no bet. An email and a password, not a card.
Inspect before identifying
The terminal is reachable directly: sign in and read a full scan, a model trace and the sources table before you tell us anything.
Dashboard
Live / Games
Full Scan
Analysis
History
ENGINE — DEMO FIXTURES · BILLING — CLOSED · SEATS — EARLY ACCESS
