SPX Engine Room · Confidence Bands
The Engine's highest-confidence reads
Every forecast is sorted by model confidence. The Engine's elite reads — the Platinum band — land ~89% of the time pre-kickoff historically, each one SHA256-stamped and model-versioned before kickoff, with zero bookmaker odds as input.
Audited track record — by confidence band
♛ Platinum
top prob ≥ 88%
93.4%
58,540 of 62,699 correct
💎 Diamond
top prob 84%–88%
84.8%
19,936 of 23,508 correct
⭐ High Confidence
top prob 78%–84%
80.4%
37,699 of 46,898 correct
🔹 Strong
top prob 65%–78%
71.5%
125,365 of 175,383 correct
Honest pre-kickoff hit-rates over the full graded corpus, current model — graded on the locked pre-match call, not refined after kickoff. Each band lands close to its stated probability (a calibrated ladder), and in-play updates refine these up as a match unfolds. These are strong favourites; the Engine's edge is being curated, calibrated and cryptographically auditable with no odds input — an independent second opinion, not out-pricing the bookmaker.
✓ correct · ✗ missed — the receipts behind the headline. Click any to open its locked report.
Tournament Guarantee — the confident reads
On international tournaments — World Cup, Euro, Copa, AFCON, Nations Leagues— the draw wall that caps club football doesn't bind: knockouts are a binary who-advancesread and group reads carry the double-chance hedge. Give the Engine 100 tournament matches and let it commit only where it's confident — at top probability ≥ 50% it has been right 77.3% of the time across 70% of fixtures. Sharpen the floor for a higher guarantee. Pre-kickoff, SHA256-stamped, zero bookmaker odds.
Wide net
top pick ≥ 50%
77.3%
1,549 of 2,003 correct
covers 70% of matches
Balanced
top pick ≥ 55%
78.4%
1,363 of 1,739 correct
covers 61% of matches
Selective
top pick ≥ 65%
80.2%
1,013 of 1,263 correct
covers 44% of matches
Elite
top pick ≥ 78%
87.3%
585 of 670 correct
covers 23% of matches
Graded on the locked pre-kickoff read with the tournament double-chance + who-advances verdict, over 2,855 graded tournament matches, current model. A selectiveguarantee: the Engine commits on the matches it's confident about and stands aside on the genuine coin-flips — the honest way to clear 70%.
Upcoming reports
Engine Reads — Engine Room
The Engine's highest-confidence upcoming reads — Platinum, Diamond, High-Confidence and Strong bands, each SHA256-stamped and model-versioned before kickoff.
All reports independent — no bookmaker odds used as input
Live now — open the cockpit
Projection movement — what changed
The upcoming reads whose probabilities moved most between the T-48h cold report and the latest pre-kickoff refresh — where new information (form, availability, lineups) actually changed the Engine's mind.
Projection Movement
The upcoming reads whose probabilities moved most across the pre-kickoff ladder — where new information changed the Engine's mind, each stage SHA256-stamped.
All reports independent — no bookmaker odds used as input
Upset Radar — where the obvious story is fragile
Every favourite carries an obvious story. These are the upcoming matches where the Engine trusts that story least— the favourite's probability is soft, a second outcome is genuinely live, or the Engine's own review flags elevated upset risk. Use it to route your attention to the fragile results: a probability assessment, not a stake — with zero bookmaker odds as input.
Upset Radar
The upcoming matches where the Engine most disagrees with the obvious favourite story — elevated-risk reads, each SHA256-stamped and model-versioned before kickoff, with zero bookmaker odds as input.
All reports independent — no bookmaker odds used as input
Who Advances — knockout ties
For knockout matches the only question that matters is who goes through — a two-way read (extra time and penalties folded in), not a three-way result. The Engine reads knockout ties far more accurately than league fixtures because the draw is no longer an answer.
Audited final-verdict record: across 8,983ties that were still level at 90', the pre-kickoff read identified the eventual advancer 57.1% of the time (two-way Brier 0.250) — extra time and penalties included, graded on who actually went through.
Who Advances
The Engine's projected team to progress in every upcoming knockout tie — extra time and penalties included, with zero bookmaker odds as input.
All reports independent — no bookmaker odds used as input
Likely Draw — the level calls others miss
No model ever picksa draw to win the three-way — yet draws are ~24% of all matches, and the Engine's draw probability is honestly calibrated. These are the upcoming matches it reads as genuinely level, with the historical hit-rate of each lean. A probability assessment, not a stake — zero bookmaker odds as input.
Likely Draw
The matches the Engine reads as genuinely level — the calibrated draw calls a single-pick model can't surface, with each band's pre-kickoff hit-rate. Zero bookmaker odds as input.
All reports independent — no bookmaker odds used as input
Outcome coverage — strong two-way reads
When no single result dominates, the Engine often has a confident two-outcome read — a side ~80%+ to avoid defeat, or a match very unlikely to end level. A probability assessment, not a stake — and it surfaces calls the single-outcome bands miss.
Outcome coverage
The Engine's strongest two-outcome reads for matches without a single dominant result — pure probability assessment, zero bookmaker odds as input.
All reports independent — no bookmaker odds used as input