Kinetic Alpha Risk Cockpit
companion to "Margin Is the Product" · illustrative venue, twenty markets
design study · not a live system

The launch schedule, twenty illustrative markets

model · calibrated 2026-09-23external OI · CoinGlass / CoinGecko 2026-09-23
Maintenance margin covers the 99% adverse intraday move scaled to the tier's close-out horizon; initial is twice maintenance; leverage rounds to the grid and caps at 20x. Tier colour on the left edge: A majors · B large alts · C long tail.

Leverage steps down with size

model
Base rate on the first 10% of the market's cap held by one account, 1.5× on the next 10%, 2× on the next 10%; 30% of the cap is the account limit (20% tier C, 10% new). Rates are marginal, so the requirement is continuous in size.

Where the maintenance margin sits against the tail

model
maintenance set99% loss at horizon99.9% loss at horizon

The one dial: fund balance → open-interest caps

model · λ from the cascade simulation
Total capacity
open interest the fund covers at target
Scaling factor θ
<1 fund binds · =1 liquidity binds
Design-stress loss at caps
Σ λ × cap
Volume this supports
at 0.6 open interest per unit of daily volume

Loss rates λ (99th pct bad debt ÷ OI)

model
The fund must cover 125% of Σ λm × OIm. Every market first gets its floor ($2M A, $500k B, $250k C, $100k new) so it stays tradable; the remaining coverage scales the liquidity ceilings by one factor θ.

How the fund grows

model · 30% fee sweep + liquidation fees
$25M/day$100M/day$250M/daysolid capacity · dashed open interest carried

Portfolio margin calculator

model · correlations since 2023, stressed on BTC's worst 5% days
Gross maintenance
sum of each leg
Portfolio maintenance
Reduction
initial margin is twice maintenance
Carried at gross
long-tail or unmeasured legs
Pairwise coefficient used (1 = adds, −1 = adds at gross, between = credited hedge at stressed correlation)

Rules

Same-direction pairs: coefficient 1 (no diversification credit).

Opposite-direction pairs: the stressed correlation, only if both normal and stressed correlation ≥ 0.50 and the committee has recorded a theoretical basis; otherwise −1, which makes the legs add.

Floors: never below the largest single leg's margin; never below 25% of gross.

Tier C and unmeasured markets (HYPE is carried by proxy until its own history is loaded) are always carried at gross. Eligibility: opt-in, ≥ $100k equity, whitelist for the first ninety days.

Equal-notional hedge credit, % of gross

model

Where a position stands

model
Equity
Health
equity ÷ maintenance requirement
State
Every step is a price the trader can compute in advance. Fills are limit orders inside the tier's band; the fee is charged on the tranche and only out of equity that exists; residual equity is returned.

The worked example, baseline against framework

model · $100k BTC long at 20x, 7% decline over 45 minutes with a 1.5% opening gap
flat-leverage baselineframework

Cascade simulation: flat-leverage baseline against the framework, 200 runs per market

model · simulation method in the piece
baseline: one flat 20x for every market, a 10/30/60% ladder of the original size, market-order fills, no backstopframework: tiered schedule, 25/50/100% of what remains, banded limit fills, backstop at 67% health

Pre-launch scenario list

acceptance test · run per market before it opens

Loss waterfall

Framework: trader margin → backstop programme / vault (takes the position at 67% health, keeps the margin) → insurance fund → ranked, disclosed, reimbursed auto-deleveraging. Socialised loss and protocol halt leave the loss waterfall; a halt remains a tool for engine, settlement or collateral failure only.

Market and venue states

A price move alone never halts a market. Order of tools for a fast market: bands → margin → reduce-only.