Fortress Core
Two sleeves with one job each: an Engine that only ever gets bought, and a Fortress reserve held specifically to be spent into it during crashes. Normal months split the contribution 50/50. When SPY closes below its 200-day moving average for five consecutive days, the Fortress starts converting into the Engine: an amount equal to twice the usual contribution immediately, and after that only when VIX closes at or above 40. There is no monthly deployment schedule. Those amounts are measured in contributions but raised by SELLING the reserve; the contribution itself never changes. The phase ends when QQQ closes back above the -7% line, which is where the selling stops and is not the same as recovered: the contribution then goes wholly to rebuilding the Fortress until it is back in its target band. A second, independent trigger at -30% swaps the Engine from 2× to 3× instruments. Built for a taxable account: within the Fortress the cheapest gains are realised first, which is what makes the whole mechanism affordable outside a shelter.
Not investment advice. This page shows the methodology's transparent logic, not a proven or guaranteed return. Trading leveraged ETFs and individual equities involves substantial risk of loss. The methodology also has a known structural limitation, described in full below: the reserve can decay to near zero as the portfolio grows, which weakens the crash protection this design depends on.
Evaluated once daily after the close. Closing prices only: an intraday dip does not count toward a confirmation.
Pre-filled with an example; replace either figure with your own. Everything is computed in your browser: nothing is sent anywhere, nothing is saved, and this is arithmetic on the published rules rather than a plan for an account.
Measured in units of the monthly contribution, but the money is raised by selling Fortress lots, lowest unrealized percentage gain first. Your contribution is not increased, and no new cash is required.
The opening deployment costs two months' worth, then one per month. This is also why the design is self-correcting: as the portfolio grows against a fixed contribution, the same Fortress percentage funds progressively more months without changing any rule.
Illustrative arithmetic on the published rules, not investment advice and not a recommendation to buy or sell anything. It assumes the structure above; a portfolio shaped differently would produce different figures.
Tickers are examples, not the rule. Any 2× vehicle serves. QLD (2× Nasdaq-100) and SPUU (2× S&P 500) are simply the reference implementation, one on each index. What the system watches is always an unlevered index, never your holdings: QQQ for the drawdown lines, SPY for the deployment trigger. What it holds is your choice. Everything the system does is aimed at owning more of this. It is not a funding source for the reserve ladder at any point: not at the trigger, not at the bottom, not on recovery. The one exception is Trigger 3, which swaps the instruments themselves; see the tax note below. Deployments buy pro-rata to the Engine's current weights, so they change its size and never its shape: an 80/20 Engine takes an 80/20 deployment. Same principle as the 3× swap, which moves each holding to its counterpart at the weights it already had.
You choose what the reserve holds. Four workable shapes, in rising order of effort: plain cash or a T-bill ETF like SGOV; one low-volatility defensive ETF like DHS; a basket of individual defensive names; or any other ETF you would rather hold as ballast. Any mixture of those, or just one of them, is valid.
It is ONE pool of structured cash: the deployment rules size against the reserve's TOTAL value and never care what is in it, and the sell order handles the sequencing on its own. There are no special rules for ETFs versus stocks versus T-bills, and no requirement to hold individual names.
If you do pick individual names: defensive, low-correlation holdings such as consumer staples, waste services, defense, pharmaceuticals. Or just DHS for simplicity, if picking a dozen names is not worth the effort: it is the closest single fund to what this sleeve is for, with a positive total return through 2022 and genuine business inelasticity across its holdings rather than a screen that merely correlates with defensiveness. A T-bill ETF works too if you would rather hold plain cash, giving up the yield to keep the stability. Think of it as structured cash: capital that earns while it waits, rather than idling at the short-term rate. Because it is bought near current prices and held as ballast, selling it realises very little EITHER WAY: small gains, and in a tech drawdown small losses rather than large ones. Both routes are chosen for the same property. The dozen names are defensive businesses whose demand does not track the Nasdaq; the one-ticker versions are picked for holding value through a tech decline, not for yield. A sleeve that fell with the index would be unable to do the one job it has, which is to still be worth something at the moment it has to be spent.
The 50/50 split above is a starting condition, not a steady state. The Engine compounds far faster than the reserve, and the rule that rebuilds the Fortress after a deployment draws on contributions only. Contributions are fixed; the portfolio is not. Restoring a 50% reserve on a $1,000,000 portfolio takes $500,000, which a monthly contribution cannot supply. Once the portfolio is large relative to what you add each year, the target band stops being reachable in any practical timeframe.
In backtesting, the portfolio reaches more than 99% Engine within roughly a decade at every opening split tested, and contributions do not arrest it. A severe decline arriving after that point meets no reserve to deploy and no exit to take, because the Engine is only ever bought. Modelled against a spliced dot-com magnitude decline from that state, the portfolio fell about 99.8% and had not recovered seven years later. That is a single simulated path rather than a forecast, but the mechanism behind it is structural.
This is a gap in the methodology, not a tuning problem. Closing it requires a rebalancing rule or an exit condition, and neither is currently part of this strategy. Anyone running it should understand that its crash protection weakens as the portfolio grows, and is weakest precisely when the portfolio is largest.
This conflict resolves itself naturally below approximately 25x your annual contribution: at that scale contributions can maintain the reserve. Above that threshold a rebalancing rule or exit condition is required to preserve crash protection, at the cost of realizing gains.
Contributing more raises that threshold proportionally, since it sits at roughly 300 times your monthly contribution. It extends the timeline by less than that implies, because larger contributions also compound the portfolio faster, so past a point the additional runway is small. What contributions change is when the reserve stops being maintainable, not whether it does.
Converts the Fortress into the Engine during a drawdown. Sizes are measured in monthly contributions but raised by selling.
- Phase 1 · Normal50% of the contribution into the Engine, 50% into the most beaten-down Fortress holding, meaning the one furthest below its OWN 52-week high in percentage terms. Not the cheapest name, and not the one you are most underwater on: cost basis is the sell-side metric and depends on when you happened to buy, so two people holding the same things would otherwise get different instructions. The 52-week high is tracked as a rolling window that expires, not a high-water mark that only ever rises. This is the resting state, and the only phase in which a new QQQ high resets the cycle peak. "New high" means above the latched cycle peak, which is the highest close the system has recorded, not an independently sourced all-time high.
- DeploySPY closes below its 200-day moving average for five consecutive days. Five closes, not one: a single-day crash does not start a deployment, and the count resets the moment price closes back above the line. A trend signal rather than a depth signal, so it fires on the market turning rather than on a fixed distance from a peak, and it does not need a cycle peak to measure from. The cycle peak is still latched, but it is now used only by the 3x swap and by the stop.
- Opening sizeAn amount equal to twice the normal monthly contribution, deployed at once, and raised by selling, not contributed. The contribution is a unit of measurement here, not the source of the money. Fires once per cycle: a price oscillating around the line cannot repeat it, and it re-arms only on a new cycle peak.
- Phase 2 · DeployingContributions switch to 100% into the Engine while the phase is active. The Fortress is selling into the Engine during the phase, so it is not also receiving: it would be funding and refilling the same sleeve at the same time.
- VIX accelerationAfter the opening deployment, this is the ONLY thing that deploys. If VIX closes at or above 40, 25% of the current reserve goes into the Engine immediately. This can repeat if VIX falls below 35 for five days and rises above 40 again. It cannot start a phase and does nothing in Phases 1, 3 or 4. There is deliberately NO monthly deployment schedule: tested 2026-08-16 over 2000-2026, adding one cost 2.35 CAGR points and 15 points of drawdown, because a monthly cadence spends the reserve into grinding declines that keep declining. The threshold is inherited from the already-validated arming gate rather than fitted here, and it has not been independently optimised.
- Sell orderLowest unrealized percentage gain first, across the WHOLE reserve, using fractional shares to raise exactly the amount being deployed. One sort, no categories: an ETF, an individual stock and a T-bill fund all enter the same ordering, and no instrument type gets special treatment. Raising a fixed sum realises tax in proportion to the gain fraction, not to position size, so the percentage is the only figure that matters. If the reserve happens to be all SGOV or all DHS there is nothing to sort: sell the required dollar amount directly. The Engine is never in the sort, because it is what is being accumulated.
- StopOne close above 7% below the cycle peak ends the phase. Deliberately a different measure from the entry: deployment starts on a TREND signal (SPY below its 200-day average) and stops on a DEPTH signal (QQQ recovered to within 7% of its cycle peak). Using one line for both would let price cross it back and forth and switch the phase on and off repeatedly; requiring a real recovery in price to stop selling means the phase ends because the drawdown healed, not because the trend briefly ticked up.
- Phase 3 · RecoveryStop selling. Contributions go 0% into the Engine and 100% into the Fortress until it is back INSIDE its 45-50% target band, i.e. on reaching 45%, not 50%. The band is a range, so "rebuilt" needs a specific edge or Phase 3 has no defined end. Not back to 50/50 yet: the sleeve that just funded the whole ladder is the one that needs refilling, and sending half the contribution to the Engine while the Fortress is depleted leaves nothing to fund the next drawdown. Phase 3 ends on a BALANCE, not a price.
- Phase 4 · RebuiltOnce the Fortress is back at target, contributions return to the Phase 1 split. The cycle peak still does not move: Phases 2, 3 and 4 all measure from the same peak, because a drawdown that has not been recovered is still the same drawdown. Only a new QQQ high above the latched peak confirms full recovery and re-arms the opening deployment.
Independent of the rules above: its own threshold, its own confirmation window, its own exit. It can fire while a deployment is running, or never fire at all.
- Engine 3×A second, fully independent trigger, not a deeper rung of the ladder above. When QQQ closes 30% or more below its cycle peak for five consecutive days, the ENTIRE PORTFOLIO is sold and moved into TQQQ. Every holding, with no exceptions: the Engine, the reserve, and the specialist sleeve. The portfolio becomes one ticker, so this is a concentration increase as well as a leverage increase, taken at the bottom of a drawdown. The composition held immediately before the switch is recorded, because a single destination cannot be reversed by rule -- that snapshot is what the switch back rebuilds from. Since the reserve is spent here, the deployment phase and the VIX gate stop while this is active: there is nothing left to deploy.
- The -7% stop does not end the 3×That line ends the deployment selling only. The 3× position has its own exit and is deliberately allowed to outlive the deployment phase; holding the leveraged recovery is the entire point of it.
- Tax treatment of the 3× swapNothing is sold in normal markets or during the reserve deployment phase, other than the reserve that funds it. At Trigger 3 the whole portfolio is sold into TQQQ: this realizes gains on every position, not just the Engine, and resets cost basis across the book. The 365-day holding requirement is specifically designed to qualify these gains for long-term treatment. Stated plainly because the taxable case for this methodology rests on not realizing Engine gains, and this is the one mechanism that does. What that costs depends entirely on the circumstances of whoever holds it in the year it fires, which this page does not attempt to estimate. Not tax advice.
- Back to 2×Both conditions, never either: QQQ at least 50% above the cycle low AND the 3× held at least 365 days. The cycle low is the lowest close from the FIRST day of the -30% confirmation window onward, not from the day the swap fired: the 3× confirms on the fifth consecutive close, so days one to four are part of the same decline and any of them can be the bottom. Measuring from the confirming close would compute the recovery off a higher floor and unwind the 3× earlier than this rule says. A bear-market rally can produce +50% before price makes new lows, and a year can pass while still deeply underwater, so requiring both means the switch back needs a recovery that has also lasted. The 365-day floor also makes the swap a long-term capital gain, which matters because this is a taxable account and the exit is a sale.
The Fortress is sized to outlast a typical bear market. How many months that actually buys depends on the size of the portfolio relative to the monthly contribution; the calculator above computes it for any pair you enter, rather than this page asserting a single figure that would only be true at one account size. The design is also self-correcting: as the portfolio grows against a fixed monthly contribution, the same Fortress percentage buys progressively more months of deployment, with no change to any rule.
The 3× line sits at -30% on QQQ because that is roughly a 55-60% decline in a 2× fund: a leveraged ETF falls close to twice as far as the index it tracks. Expressing the threshold on the unlevered index keeps it measuring how far the market has fallen rather than how much decay the instrument has accumulated, which is the same reason every other line on this page is set on an unlevered index rather than on whatever leveraged fund you happen to hold.
The confirmation window and the gap between the two lines exist for the same reason: the system should be difficult to start and difficult to stop, so that noise cannot switch it on and a single green day cannot switch it off.