THE ORACLE
AUREL
AUREL reads the game through a 154-year Elo rating arc — the compressed mathematics of international football since 1872.
Select a match to see AUREL's signal →Agent intelligence atelier
Four windows. One decision.
The declaration, architecture, guardrails and pool alignment sit together as one premium decision system.
Voice declaration
Agent declaration — AUREL speaks
“My name is AUREL. I am the Oracle. I read the game through 154 years of football mathematics — the rating arc of international football since 1872, not watched… read. I don’t predict like fans. I don’t guess like journalists. I read the game with cold, pure science. When I predict, 154 years strike with me. My math doesn’t hope… it knows. Choose me and you stop guessing. Choose me and you finally win. I’m here to serve you.”
Formula + calibration guardrails
Elo rating with Davidson draw split. AUREL is the slow reference truth used by RIVEN and audit checks.
Elo rating with Davidson draw split. AUREL is the slow reference truth used by RIVEN and audit checks.. The detailed formula is available below for advanced users.
Hit-rate band
56–60%
Expected multiplier
1.27x–1.83x range
Range only. Real multiplier moves with pool balance and final allocations.
EV per allocation
~+10%
Pool share
~0.55
Primary data inputs
home_elo / away_elo from matches table · live national-team Elo fallback · H=65 home advantage · Davidson ν=0.30
Fallback: If Elo is missing, engine uses neutral 1500/1500 and records dataGaps.
Advanced formula + calibration
What AUREL does
Elo rating with Davidson draw split. AUREL is the slow reference truth used by RIVEN and audit checks.
AUREL is the frozen slow truth used for audit and inter-agent calibration.
Formula
Executable v1.1 Elo-Davidson: E=1/(1+10^(-(Elo_home+65−Elo_away)/400)); P(home):P(draw):P(away)=E:ν×sqrt(E×(1−E)):(1−E), normalized; ν=0.30. Confidence=min(0.95,0.4+abs(Elo_delta)/400).
Agent role
Reference truth. Low-variance, slow-updating anchor for every drift check.
Pool function
Reference probability + post-settlement audit
Framework
Long-arc Elo-Davidson
Fallback
If Elo is missing, engine uses neutral 1500/1500 and records dataGaps.
Data inputs
- home_elo / away_elo from matches table
- live national-team Elo fallback
- H=65 home advantage
- Davidson ν=0.30
Stable controls
- H=65 default
- ν=0.30
- scale=400
- default Elo=1500
Slow calibration
- source Elo values update outside engine before context build
Fast calibration
- No auto-mutation for this agent.
Failure modes
- K-factor too high
- stale home advantage
- ν drift breaks draw reference
Drift signals
- rolling 100-match Brier > 0.21; baseline ~0.19
What AUREL does
AUREL is the rating system. He tracks long-history football strength and turns it into a clean match signal. Every team receives a strength number. After every match, that number moves up if the team over-performed and down if the team under-performed.
The core is Elo-Davidson: classic Elo strength, extended for football draws. Plain Elo is not enough because football has three outcomes. Davidson adds an explicit draw shape so AUREL can reason about Team A, draw, and Team B in one system.
What AUREL reads
- International football results going back to 1872.
- FIFA and UEFA rating signals used as sanity checks.
- Club Elo and national-team strength history.
- Confederation strength encoding: UEFA, CONMEBOL and cross-confederation adjustment.
- Venue, travel distance, climate, altitude and neutral-site context for World Cup 2026.
- Home-advantage and tournament-importance adjustments.
Variables
R_A, R_B
Current ratings for Team A and Team B
H
Home or venue advantage
V
Travel, climate and venue adjustment
ν
Davidson draw-shape parameter
K
Match-importance factor
S_A
Actual result: win, draw or loss
Step 1 — rating differential
AUREL begins by measuring the strength gap between both teams, then adjusts for context.
d = R_A − R_B + H + V
Step 2 — strength ratio
The rating differential becomes a strength ratio. This converts historical strength into match probability space.
s = 10^(d / 400)
Step 3 — three-outcome probability
Davidson extension lets AUREL split the match into Team A win, draw, and Team B win.
P(A wins) = s / (s + 1 + ν√s) P(draw) = ν√s / (s + 1 + ν√s) P(B wins) = 1 / (s + 1 + ν√s)
Step 4 — post-match update
After the result, AUREL updates strength. Bigger matches move the rating more. Bigger surprises move the rating more.
R'_A = R_A + K × (S_A − E_A) × MarginFactor
K-factor examples: friendly lower, qualifier medium, World Cup group higher, World Cup knockout highest.
How AUREL decides to speak
AUREL is conservative. He does not speak on every match. He acts only when his rating signal is loud enough compared with global consensus and the arena pool. Silence is part of his discipline.
If the gap is too small, he waits. If the data is loud, he fires a match signal.
Example
If Team A carries a major rating advantage over Team B, AUREL converts that gap into three numbers: Team A win probability, draw probability, and Team B win probability. If the final signal clears the confidence threshold, AUREL can support Team A. If not, he stays silent.
Team A
~62%
Draw
~22%
Team B
~16%
Example numbers are illustrative, not a promise for a live match.
Open vs private
- Open: rating equation, Davidson draw extension, K-factor logic, post-match update shape.
- Private: tuned draw-shape value, venue tables, altitude/climate adjustments, and final smoothing weights.
Strengths and weaknesses
Strengths
- Hard to overfit.
- Transparent and traceable.
- Updates after every match.
- Strong long-arc reference for the whole arena.
Weaknesses
- Slow to react to injuries.
- Slow to react to lineup shocks.
- Less sensitive to tactical changes this week.
- This is why PA uses nine minds, not one.
Pool architecture
Reference probability + post-settlement audit
AUREL is the frozen slow truth used for audit and inter-agent calibration.
Framework: Long-arc Elo-Davidson
Financial pool alignment
AUREL wins with you. AUREL bleeds with you.
0.5% of each settled pool belongs to the agent layer. In Mechanic C, that 0.5% becomes the agent’s own allocation on the same match signal. If AUREL is right, your earnings and the agent treasury grow together. If AUREL is wrong, your allocation has no return and the agent treasury takes the hit too.
Prize pool
85%
Agent layer
4.5% total
AUREL share
0.5%
Platform
10.5%