Margin, Max Loss and Liquidation: Field Guide
The practical rules for levering fortress structures without surprise
liquidations, plus the pre-trade ladder. Numbers from 20260801_2058.html at $700,000 NLV.
Generated 2026-08-04 12:59 by
module_liq_ladder.py. New to this? Start with the animated
dumbed-down explainer.
Updated 21 Oct 2026. The rules below are unchanged. The
addendum at the bottom adds how to check an account against IBKR, and what that check shows for today's four accounts, which are not built to this guide's standard: for them the ~$100k floor is reachable (Neville about an 18.5% S&P drop, Main about 28%). Live verdict per account: the Desk's Repair tab; dated record:
Crash Liquidation.
The three numbers everyone confuses
| Number | What it is | How it behaves in a crash |
|---|
| Max Loss | Worst case if held to LEAPS expiry. Fixed by your
strikes: net debit + (short put − hedge put) gap. | Never moves.
The broker never looks at it. |
| Margin requirement | What IBKR sets aside today: worst loss in a
±15% scan from the current price. | Shrinks as the market falls
(deep-ITM options lose gamma; the scan band narrows in dollars). |
| NLV (your equity) | Live mark-to-market account value. |
Falls hard. This is the only number that actually collapses. |
The one rule: a margin call fires when NLV < margin
requirement. It is an equity event, a race between two falling numbers:
your equity falls at the book's leverage (fast), the requirement falls slowly, and
liquidation is where equity dives through it from above. "Margin ballooning in a
crash" is a myth for broad index books; the requirement works in your favor.
The rule that makes liquidation impossible
If total Max Loss ≤ your cash, you cannot be margin-called,
verified in simulation to a −95% crash. When the worst case is smaller than
what you hold, there is nothing for a call to bite on.
The one giant asterisk: this only holds when the hedge is OWNED, not rented.
An owned hedge is a held LEAPS put: paid once, protects for the whole term, costs $0
on the path. A rented hedge is a 1-3 month put re-bought forever: keeping the max-loss
cap alive costs cash every cycle, and in a slow bear that rent drains NLV until the
call fires anyway. Proven on the 2022 path: same book, same capped max loss, rent
charged = margin call Jun 2022 ($234k of hedge rolls vs $172k income); rent set to
zero = survives. The roll cost IS the liquidation mechanism.
Predicting your liquidation level upfront
The best single predictor is the Max-Loss-to-Cash ratio. For this
structure family (deep-ITM LEAPS call + OTM LEAPS short put + far-OTM held hedge):
liquidation drawdown ≈ −90% +
(Max Loss ÷ Cash − 1.0) × 60 pts
Check: 1.5x max-loss book → −90 + 30 = −60%, matching
the simulated −61%. Below 1.0x the answer is NEVER. Verify the exact level in
IBKR Risk Navigator (beta-weighted What-If chart at Range 100%, see the 21 Oct addendum) before
executing: that is the broker's own model and the final word.
Delta leverage vs ML/cash: two dials, two risks
| Dial | Measures | Drives |
|---|
| Delta leverage | Live market exposure per account dollar. Sum
each leg's share-equivalence (deep LEAPS call ~0.9, short put adds +0.2-0.3, hedge
~0), × spot × 100 × contracts, ÷ NLV. |
Day-to-day P&L per 1% move; your upside. |
| Max Loss ÷ cash | Worst-case loss vs what you hold. |
Liquidation risk. The ladder keys off this one. |
They rise together as you add contracts but are not the same thing. A
fortress contract carries MORE than 1.0 delta at entry (the short put adds bullish
exposure), and the delta breathes: toward 1.0 per contract in a rally, above it in a
selloff while dollar leverage self-de-levers toward 1x.
SPY + QQQ ladder (DIAMOND)
Income = fully covered at the run's SC delta.
| Book | Max Loss | ML ÷ Cash | Delta Leverage | Margin (t0) | Cushion | Net Inc/mo | E[Inc]/mo | Liquidates At |
|---|
| 3 SPY / 4 QQQ | $459,465 | 0.66x | 0.82x | $89,293 | $610,707 | $7,807 | $3,221 | NEVER |
| 4 SPY / 5 QQQ | $593,370 | 0.85x | 1.05x | $114,967 | $585,033 | $9,962 | $4,113 | NEVER |
| 5 SPY / 6 QQQ | $727,275 | 1.04x | 1.28x | $140,642 | $559,358 | $12,117 | $5,005 | NEVER |
| 6 SPY / 7 QQQ | $861,180 | 1.23x | 1.52x | $166,316 | $533,684 | $14,271 | $5,897 | -82% |
| 7 SPY / 9 QQQ | $1,052,835 | 1.50x | 1.86x | $204,260 | $495,740 | $17,769 | $7,334 | -61% |
| 8 SPY / 10 QQQ | $1,186,740 | 1.70x | 2.10x | $229,935 | $470,065 | $19,924 | $8,226 | -49% |
| 9 SPY / 11 QQQ | $1,320,645 | 1.89x | 2.33x | $255,609 | $444,391 | $22,079 | $9,118 | -40% |
IBIT ladder (DIAMOND)
Income zeroed for: IBIT
| Book | Max Loss | ML ÷ Cash | Delta Leverage | Margin (t0) | Cushion | Net Inc/mo | E[Inc]/mo | Liquidates At |
|---|
| 90 IBIT | $265,950 | 0.38x | 0.50x | $52,235 | $647,765 | $0 | $0 | NEVER |
| 120 IBIT | $354,600 | 0.51x | 0.66x | $69,647 | $630,353 | $0 | $0 | NEVER |
| 150 IBIT | $443,250 | 0.63x | 0.83x | $87,058 | $612,942 | $0 | $0 | NEVER |
| 180 IBIT | $531,900 | 0.76x | 0.99x | $104,470 | $595,530 | $0 | $0 | NEVER |
| 230 IBIT | $679,650 | 0.97x | 1.27x | $133,490 | $566,510 | $0 | $0 | NEVER |
| 250 IBIT | $738,750 | 1.06x | 1.38x | $145,097 | $554,903 | $0 | $0 | NEVER |
| 300 IBIT | $886,500 | 1.27x | 1.66x | $174,117 | $525,883 | $0 | $0 | -85% |
| 350 IBIT | $1,034,250 | 1.48x | 1.93x | $203,136 | $496,864 | $0 | $0 | -72% |
Why the same liquidation level means opposite things:
a −70% point is bulletproof on SPY/QQQ (worst ever: 2008 at −57%; 2020
−34%; 2022 −25%) and a coin-flip on IBIT (BTC: 2018 −84%, 2021-22
−77%, Mar-2020 −50% in days). Rule: set the liquidation point ~15 pts
deeper than the worst crash the underlying has actually done. For SPY/QQQ that
allows Max Loss up to ~1.2-1.3x cash (1.5x puts the liquidation point at −61%,
right at the edge of a 2008 repeat, acceptable only with eyes open); for IBIT it
means Max Loss ≤ cash, always (≤230 contracts at current strikes on a
$700k account).
Practical checklist before levering
- Size by Max Loss, never by margin. Margin looks cheap (15-20% here) and
shrinks when you fall; it tells you nothing about survival.
- Own the hedges (held LEAPS puts). Rented hedges void the max-loss-=-cash
guarantee.
- Max Loss ≤ cash = ironclad. Above that, each +0.1x of ML/cash pulls the
liquidation point ~6 pts shallower.
- Match the level to the underlying's real history (box above).
- Confirm in Risk Navigator with live prices (beta-weighted What-If, Range 100%: the line must stay above NLV minus $100k to −30%); model tolerance here is a few
points and IBKR can add discretionary house margin in a panic.
Methodology and regeneration
Engine: Black-Scholes leg marks vs entry mids; TIMS-style ±15%
scan margin (validated within ~4% of the fortress_v28 margin_proxy at t0);
crash vol multiplier 1 + 3.3×|drawdown| with sqrt(30/DTE) tenor damping on the
margin surface; crossover = first 1% step where NLV − margin ≤ 0, searched
to −95%. Income columns are the dashboard's net (headline) and calibrated
E[net] rates and are NOT credited inside the crossover (conservative). Levels carry
±3-4 pts of model tolerance. Regenerate:
cd ~/fortress/scripts && python3 module_liq_ladder.py <dashboard.html>
--nlv 700000 --ladder "SPY,QQQ:3/4,...,9/11" --ladder "IBIT:90,...,350" --no-income IBIT
--html <out.html>
Addendum 2026-08-21: the ML ≤ cash rule, settled exhaustively (proposed 3-account book)
Claim. If every position is a fully hedged fortress (LC + SP + HP, matched counts), total Max Loss ≤ cash, and the hedges are always rolled, no IBKR liquidation can occur. Proven against the proposed Main ($2.0M, ML 0.94x) and Joint ($1.0M, ML 0.94x) books with the module_risk_cockpit TIMS-race engine, run live 2026-08-21, 1% steps to −95%.
- One trigger only. IBKR liquidates when Excess Liquidity < 0 (maintenance > NLV). So the claim holds iff NLV > maint on every path. It is a race between two falling numbers.
- Equity has a hard floor. With the hedge alive the worst terminal loss is ML by construction, and pre-expiry MTM loss is smaller than ML (hedge puts carry time value plus crash-vol premium when spot falls). So NLV ≥ cash − ML > 0 always. Worst simulated path NLV: Main $349,752 (at −66%) vs its cash−ML floor $110,344; Joint $272,024 (at −95%) vs $55,053.
- The requirement has a falling ceiling. Maintenance is the worst loss in a ±15% scan around current spot, not worst case to zero; as spot falls the legs go intrinsic-dominated and the requirement shrinks (to ~$9k at Main's worst point, against ~$350k of equity). The race is not close.
| Attack (crossover = first point where excess ≤ 0) | Main | Joint |
|---|
| Base, flat vol | NONE past −95% | NONE past −95% |
| Crash-vol ramp (vol ~2x at −30%, skew-aware) | NONE | NONE |
| Ramp + house margin hike x1.6 | NONE | NONE |
| Ramp + house margin hike x2.0 | NONE | NONE |
| Ramp + house margin hike x3.0 | NONE | NONE |
| t+12 months decay + ramp + hike x2 | NONE | NONE |
| ~$100k portfolio-margin floor race | not within −95% | not within −95% |
- PM-floor cliff cleared. Below ~$100k NLV IBKR reverts the account to Reg-T and requirements jump; the equity floor keeps both accounts above ~$270k on every path, so the cliff is unreachable. This is why 0.94x is the calibration: ML ≤ cash is what keeps the NLV floor above the PM cliff.
- Early assignment is not a liquidation vector. A short put deep enough ITM to be assigned has share-equivalent delta ≈ +1; assignment swaps it for actual shares (identical exposure, NLV unchanged, plus the time value the exerciser forfeited). The long puts are the disposal mechanism: sell the stock at market (realizing only the already-marked intrinsic) or exercise the HP against it, same day. Even a broker auto-liquidation in that transient window executes the same stock sale you would do voluntarily. Do NOT passively hold assigned stock: that state (not assignment itself) is where margin balloons.
- Conditions (operational, not structural). (1) The premise: the guarantee is "ML ≤ cash AND hedge alive", never ML ≤ cash alone; rent must always be paid (the 2022 replay above: rent IS the liquidation mechanism when rolled blindly with no income; the proposed book's coverage is 5-7x, so income must collapse >80% before rent touches principal). (2) Roll before expiry; never let a hedge lapse even one session. (3) Broker discretion beyond modeling: tested to a tripled requirement; expiry "anticipated deficiency" liquidations are covered by the HP; the ritual final check remains Risk Navigator (slack here is 30+ pts vs a few pts of model tolerance).
- 2026-08-24 re-validation (3-account restructure). RetiredInc is closed; its index book and cash merged into Joint as one $1.35M account (index fortresses IC ~$544k + ~$806k cash; ML $944,947 = 0.70x cash, deeper than the original 0.94x sizing the table above was run at). Battery re-run at the merged sizing: crossover NONE past −95% under every attack incl. house hike x3; worst path NLV $622,024 at −95%; cash−ML floor $405,053. Every guarantee above holds with MORE slack than the printed Joint figures.
- Verdict. On every price path to −95%, under crash vol, a tripled house requirement, twelve months of decay, and full early assignment, neither account can be margin-called. The residual risk is whether the two operational rules are kept, not whether IBKR can liquidate.
Addendum 2026-10-21: checking against IBKR, and today's accounts
Scope of the guarantee. Everything above is about a book built to the rule: Max Loss ≤ cash with owned hedges. Today's four accounts are not built that way (rented 1-3 month hedges, legacy BBC sleeves, stock no fortress covers, Max Loss above cash), so for them the ~$100k portfolio-margin floor is reachable, and it, not the excess-liquidity crossover, is what binds.
| Account (21 Oct) | NLV | IBKR: floor reached at | Left at −30% |
|---|
| Main | $1.37M | about −28% (line ends at −26%, extended) | below the floor |
| Joint | $702k | not to −30% | ~$87k |
| Neville | $628k | about −18.5% | below the floor (with 12x SPY 730P: ~$38k) |
| RetireInc | $331k | not to −30% | ~$20k (thin) |
- IBKR's yardstick is the beta-weighted What-If, volatility unchanged. It matched the model's volatility-held-still path within 2-11% on all four accounts, and for those accounts the crash-vol ramp in this guide's methodology was the optimistic end, so for an account check quote the volatility-held-still number. The ML ≤ cash guarantee does not depend on the ramp: it holds on flat volatility too (first row of the attack table above), because the most a fully hedged book can lose is its Max Loss.
- Not the yardstick: IBKR's Stress Test PDF moves every stock by the same % (vol +300% at −30%), which is kinder to a high-beta book. Do not judge the floor from it.
- How to check an account:
- TWS → Risk Navigator → the account → What-If portfolio; chart Beta weighted (SPX), Plot = Equity Portfolio Value Change.
- Range = 100%, then Update (about a minute). Range sets how far each stock moves; 40% only reaches about −11% on the S&P axis.
- Read the line at −10 / −20 / −30% from its own 0% point. The account is safe to that depth while the loss stays under NLV minus $100k.
- To test a hedge, add it to the What-If portfolio and redraw.
- Hedges against the floor: a whole-account SPY put is the cheapest lever. In the What-If, IBKR credited an added SPY put about 85% of the model's crash gain, and the plan must still hold after a 5% bad day; the Repair tab prices and sizes it that way. A 3-month put is a rented hedge: the rent rule above applies, and a longer-dated put is the owned version.
- Where the verdict lives: the Desk's Repair tab, per account, the worse of the model and the latest IBKR check (trusted 45 days; refresh monthly). Data:
scripts/ibkr_stress_calibration.json; bundles _md_files/repair/2026-10-21_ibkr_calibration/ and _md_files/portfolio/2026-10-21_liquidation_check/.