The method · nothing hidden
How the T&G Score works
Most tools hand you a number and tell you to trust it. We publish the equation. The T&G Score is a
fixed, transparent 0–100 verdict built from the same dealer-positioning math the street runs
internally — same inputs, same output, every time. Here's exactly what goes into it.
The equation
One line, on purpose. The Score is a weighted blend of four components, each normalized to a
0–1 scale so no single input can run away with the number.
T&G Score = 100 × ( Σ weightᵢ × componentᵢ ) ÷ Σ weightᵢ
The four components
Weights reflect where the real edge is. Dealer positioning carries the most because it's the
part retail never gets to see.
40%Dealer positioning
The street's real edge: the gamma regime (are dealers amplifying moves or dampening them),
how close price sits to the flip (the tipping point where their hedging behavior changes), and how near
the nearest wall (defined support/resistance from dealer gamma).
Inside it: regime 45% (negative gamma = accelerant scores high, positive = pinned scores lower)
· distance-to-flip 35% (within 3% = maximal) · wall proximity 20% (within 4%).
30%Setup structure
Is there an actual, tradeable geometry here — a real flag/pole setup with favorable
reward-to-risk — or just noise?
Inside it: flag quality 60% · reward-to-risk 40% (2R saturates the credit).
15%Vol context
Are options cheap or rich for this name right now? Cheap premium favors the buyer.
IV rank 0 → full credit, 100 → 0.15 floor. Cheap vol scores higher.
15%Flow confirmation
Does order flow agree with the setup's direction? This is the confirmation layer — it can
nudge the Score up when flow agrees and down when it fights the math, but the verdict stands without it.
Agree → 1.0 · fights → 0.0 · no flow read → simply not counted.
Two layers — and why they're separate
The number you see is deliberately NOT a machine-learning black box. It's a stable equation,
with a learning model riding on top as a confidence read.
Layer 1 — what you see
The T&G Score
A fixed, transparent, deterministic equation. Same inputs → identical output, forever. The weights are
versioned (tg_score_v1) and only change on purpose — never silently week to week. This is the
number we publish and stand behind.
Layer 2 — riding on top
Win-probability model
A calibrated model trained on our graded track record: "the equation says 82; setups scoring like this have
historically hit ~X%." It refines trust — it never overrides the Score. It activates once the ledger
has enough graded trades to be honest, and sharpens every week after.
Why keep them apart? A machine-learning model retrains and its weights drift — great as an
odds overlay, disqualifying as the headline. The verdict has to be stable; the probability rides on top.
Worked example
A real setup: negative gamma 0.5% from the flip, a bull flag with 2.1 reward-to-risk, mid-range IV,
and flow agreeing.
Dealer positioning (×0.40)0.885
Setup structure (×0.30)0.844
Vol context (×0.15)0.560
Flow confirmation (×0.15)1.000
→ 100 × weighted blend84 · A
Grades
The distribution is intentionally selective — most setups land C or below on any given day. A
high grade is meant to be rare, not flattering.
When we're flying blind, we say so
Some names (or historical dates) have no live dealer-gamma chain. Rather than fake a number,
the Score drops the dealer-positioning component and renormalizes the rest — and the card tells you
"gamma-blind: dealer positioning unavailable." Honesty beats false precision.
The T&G Score is a structural, mathematical read — not a prediction, not financial advice. Live
prices come from a real-time feed; the options chain is ~15-min delayed on our current tier. Trade your own plan.