Final strategy • tested Nov 2000 to Sep 2026 • 25.8 years

The Volatility Gate

One rule, three conditions, roughly five trades a year. This is what survived after testing moving averages at six lengths, SuperTrend at sixteen parameter pairs, dip-buying, faster exits, and inverse funds at three leverage levels.

The rule

Hold TQQQ when QQQ is above its 200-day average and 10-day volatility is under 30%. Otherwise hold cash, except that while price is below the 200-day, a close back above the 50-day average with volatility under 45% buys in early, and a close back under the 50-day sells again.

Check once a day at the close. Act only when the state changes. Across the full history that meant about five round trips a year, and being invested 77% of the time.

Invest starting then add

Total paid in

—

The volatility gate

—

Just hold TQQQ

—

QQQ, buy and hold

—

Every line starts at 0% at the left edge of whatever you are looking at, so every window reads as its own comparison. Zoom or drag and the baseline moves with you. Log scale. The % button switches the axis to dollars, which also draws the money-paid-in line. The dashed line is QQQ's 200-day average, drawn so it crosses the blue QQQ line exactly when price crosses the average. The band underneath marks the periods the gate was holding, and the panel below tracks the volatility condition. Every line, both averages, the holding band and the volatility panel can be switched off by clicking its name in the legend below the chart, and your choices are remembered. With contributions on, the lines include new deposits as well as growth.

Start investing:

Click any item above to hide or show it. The two averages are drawn against the QQQ line, so they only appear while QQQ is shown.

Does trading beat holding?

It depends on when your money goes in, which is exactly what the controls above let you test.

Start Nov 2000 with a single deposit and the gate ends at $7.63M against $22,875 for just holding, because holding walks into the dot-com crash and never recovers. Start Dec 2010, just after a crash with none ahead for a decade, and holding wins outright. Start Sep 2021 and the gate is ahead again, because 2022 lands early in that window.

Switch contributions on and the gap narrows sharply, because every dollar added during a crash buys at prices that later multiply. Steady contributions are what rescue buy and hold from catastrophe. What they do not fix is the drawdown you have to sit through while making them.

The 50-day early re-entry, and how much of it to believe

The gate's weakness was that the 200-day is slow. In a long grinding bear market it kept you in cash through rallies worth catching. The clause added in September 2026 fixes that: while price is below the 200-day, buy back as soon as QQQ closes above its 50-day average with volatility under 45%, and sell again if price closes back under the 50-day. It fired on 408 days out of 6,490, so it is rare. The green line on the chart is the rule without it.

Fast average used	2001-06	2007-12	2013-18	2019-26	25-year
none (the plain gate)	3.6%	8.6%	31.3%	47.1%	201x
40-day	-6.6%	17.1%	35.9%	50.8%	227x
45-day	-0.3%	17.7%	34.5%	51.4%	340x
50-day	6.0%	23.5%	37.0%	53.1%	802x
55-day	-1.1%	18.7%	32.3%	49.9%	313x
60-day	2.0%	19.9%	31.1%	47.1%	327x

Read this table carefully. The robust finding is that every fast average improves three of the four periods, most dramatically 2007-12, where the plain gate earns 8.6% a year and every variant earns 17% or more. That consistency across six independent parameter choices is what makes the idea credible. The 802x for the 50-day specifically is not credible on its own: its neighbours score 340x and 313x, and the gap comes almost entirely from the noisy 2001-06 column. The honest expectation is the family, roughly 300x, not the winner. The 50-day is chosen because it is the most widely watched average in the market, a reason that exists independently of this table.

It won in both halves, which is the part that matters

Period	The gate	QQQ	Gate drawdown
Crash decade, 2000 to 2010	$37,332	$6,655	-57%
Bull decade, 2010 to 2026	$2,044,596	$151,814	-47%
Full period	$7,632,907	$101,035	-57%

Fixed record: a single $10,000 deposit on the dates shown, no contributions, regardless of the settings above. Note the crash decade: QQQ itself lost a third over those ten years, ending at $6,655, while the gate nearly quadrupled. That is the whole case for the rule. A strategy that only wins in bull markets is just leverage wearing a disguise.

Why this works when faster rules did not

Leveraged funds are killed by chop, not by declines. A 3x fund that falls 10% then rises 10% ends up behind, not level. Repeat that enough and it dies even if the index went nowhere. So the thing worth detecting is turbulence, not direction.
The two conditions catch different dangers. The 200-day average handles the slow grind down. The volatility reading handles the violent break. Neither alone was enough: the average missed March 2020 entirely, and volatility alone would hold through a slow bleed.
Faster price signals lose as exits, but a fast average earns its keep as an entry. Replacing the 200-day with a 50-day for selling fires on ordinary dips and grinds the account down. Using the 50-day only to get back in while still below the 200-day is the one change that helped, because the risk it takes is bounded: it sells again the moment price slips back under.
Speed helped only on the volatility side, and only so far. Moving from a 20-day reading to 10-day improved reliability. Going to 5-day made it worse again.
Shorting the downturns lost at every leverage level. Cash beat -1x, which beat -2x, which beat -3x, cleanly, in both decades. When the gate shuts, the next month was up 57% of the time.

What you are accepting

A 57% drawdown. Not hypothetical. Every version of leveraged Nasdaq has one, and you have to still be holding at the bottom for any of these numbers to be yours.
A known failure mode. A bear market that bleeds slowly without a volatility spike keeps the gate open the whole way down.
Chosen settings, not discovered laws. 200 days and 30% are two numbers on a grid. The 10-day window was picked because its neighbours all performed similarly, not because it scored highest. The top-scoring cell returned $2.48M while its immediate neighbours returned a third of that, which is what a lucky result looks like. The same caution applies to the 50-day, as the table above spells out.
A simulated instrument before 2010. TQQQ did not exist until February 2010. Everything earlier models a 3x fund from QQQ returns, less fees and financing. Checked against the real fund over the overlap, the model understates returns by about 20%.

Operating mandate

This is how the rule is actually run in Robinhood account ••••7916, the one account Claude can trade. It replaces the earlier sleeve mandate of 17 September 2026, which was written before this strategy existed and contradicted it.

Position sizing

When the gate is open, hold 65% of the sleeve in TQQQ and leave 35% in cash. When it is shut, hold 100% cash.

This is the only risk dial that works. At 100% TQQQ the worst historical drawdown is 57% and the return is 29.4% a year; at 65% it is roughly 44% and about 24%. Stops were tested and every one of them made the result worse without reducing the drawdown, because the drawdown is not one bad trade. The deepest one ran from October 2007 to May 2009 and contained twelve separate gate cycles. Change this number whenever you like; it is the decision, not a detail.

Operating rule	Why
The signal reads price and volatility only	QQQ daily closes, the 200-day and 50-day simple averages, and 10-day realised volatility. No news, no sentiment, no discretion, including Claude's. The morning brief reports the gate's state and never feeds it.
Checked once a day, near the close	Act only when the state changes. About five round trips a year. Intraday moves are ignored entirely.
Reconcile, never remember	Every run compares the actual position against the target and corrects the difference. A missed day heals itself the next day rather than compounding.
No stop orders	The gate is the stop. Robinhood also rejects stop orders on fractional positions, so a stop would force whole-share sizing for no benefit.
Fractional orders, regular hours, market	Fractional is how the 65% is sized precisely. Pre-market and overnight books are too thin to be worth it.
Settled cash only	This is a cash account on T+1. Buying with unsettled proceeds and selling before settlement is a good faith violation, and three in twelve months restricts the account for 90 days. A buy waits rather than risks one.
Bad data means do nothing	If the historicals call fails or returns values outside sane bounds, no order is placed and the previous position stands. Never trade on a guess.
No drawdown halt	Deliberate reversal of the old mandate, which halted at a 20% loss. A drawdown halt would have taken you out at the bottom of every crash in the record and never put you back. The sizing decision above is where risk is set, not a panic switch.
Only this account	The main account, the Roth, the traditional IRA and the joint account are unreachable and stay that way.
Kill switch	Say halt the sleeve and no further orders are placed until you say otherwise. Revoking the agentic permission in Robinhood is the harder and more reliable version of the same thing.

Before it runs on real money: a notify-only period of two to three weeks, where the daily task computes the signal and sends it to you but places no orders, so the arithmetic can be checked against a live market before it is trusted with orders. And the thing worth more than any rule above: this strategy realises nearly every gain as a short-term gain. Taxed annually at roughly 35%, the 26-year simulation drops from $58.8M to $7.0M. If this money can ever sit in a tax-sheltered account, that is worth more than every refinement on this page combined.

Data: split-adjusted daily closes for QQQ from the Robinhood market data feed, 21 November 2000 (the first date with both a 200-day average and a volatility window) through 16 September 2026. Signals use the prior day's close, so no future information enters any decision. Costs: a 0.95% annual expense ratio, financing at twice the fed funds rate of the day on the leveraged position, and 2 basis points per round trip. Idle cash earns the fed funds rate.

A historical simulation is not a projection, and past results do not predict future returns. Leveraged ETFs reset daily and are not designed to be held long term. Nothing here is a recommendation or financial advice.