Backtest · Nasdaq-100 and TQQQ · 1986 → 2026

How to Hold TQQQ With the Least Risk: 72 Rules Tested on 40 Years of Nasdaq-100 Data

TQQQ moves three times the Nasdaq-100 each day. Since it launched in 2010 it has made 42.9% a year — and fallen 82% along the way, in 2022. We asked how to hold it with the least risk: 72 rules on 40 years of data, from trend filters and stop-losses to volatility sizing, each ranked on 1986–2015 and then judged on the ten years it had not seen.

Published October 14, 2026 · Stable Income Strategy research

The short answer

Holding TQQQ (the 3× Nasdaq-100 fund) outright is never low risk: a 3× fund built from the index lost 100% from March 2000 to March 2009, and the real TQQQ fell 82% in 2022. Of 72 holding rules ranked on 1986–2015 and judged on 2016–2026, the one family that held up was sizing by volatility: hold 20% ÷ TQQQ’s 20-day volatility of the account in TQQQ (at most all of it) and keep the rest in Treasury bills. Since 2016 that made 23.4% a year with a −32% worst fall, against 20.3% and −35% for holding QQQ and 39.9% and −82% for holding TQQQ; over the full 40 years it made 18.3% a year with a −50% worst fall. A 200-day moving-average filter alone did not protect a 3× fund (−89% worst fall), and the backtest ignores taxes on about 125 short-term trades a year.

a year since 2016 with volatility sizing, on years it was not ranked on: 23.4%
its worst fall since 2016, against −82% for TQQQ held: −32%
worst fall of TQQQ with only a 200-day filter, since 1986: −89%
holding rules tested on 40 years of data: 72

Key takeaways

Holding TQQQ outright is never low risk: a 3× fund built from the index lost 100% from March 2000 to March 2009 and never got back, and the real TQQQ fell 82% in 2022.
Sizing by volatility — 20% ÷ TQQQ’s 20-day volatility of the account in TQQQ, the rest in T-bills — made 18.3% a year since 1986 with a −50% worst fall, against 14.9% and −83% for holding QQQ.
It is the one family that held up on years it was not ranked on: 23.4% a year with a −32% worst fall in 2016–2026, against 20.3% / −35% for QQQ and 39.9% / −82% for TQQQ.
The calmest setting — a third of the account in TQQQ, only while QQQ is above its 200-day average — made 16.1% a year since 2016 with a −25% worst fall, and lost 40% at worst in the dot-com crash.
A 200-day filter alone does not protect a 3× fund: it sells after QQQ has fallen 25–30%, about −70% in TQQQ. Worst fall −89%, −62% in Covid, 14 years under water.
Fixed shares rebalanced monthly, trailing stops, moving-average crosses and momentum rules all had worst falls between −69% and −99% over 40 years.
Most rankings did not survive: across the 72 rules, the order on 1986–2015 barely predicted 2016–2026 (a rank correlation of 0.09).

What we tested

Seven families of rules, 72 rules in all, every one long only with the money not in TQQQ earning the Treasury-bill rate. The real TQQQ from 2010; before that a 3× fund built from the Nasdaq-100 index, back to 1986.

1Buy and hold

Tested: QQQ (1×), a 2× fund (QLD, built from the index) and TQQQ (3×), held through everything.

Found: Over 40 years 3× made the least: 14.5% a year, below QQQ’s 14.9%, with a −100% worst fall against −83%. 2× made 18.2% but fell 99%.

2A fixed share in TQQQ

Tested: 25%, 33%, 50% or 67% of the account in TQQQ, the rest in T-bills, set back to the target every month.

Found: 12.2% to 18.6% a year, but worst falls of −69% to −99%: topping back up each month keeps buying a falling fund.

3Trend filters

Tested: TQQQ only while QQQ is above its 50-, 100-, 150-, 200- or 250-day average (with bands of 1–5%), or TQQQ above its own 200-day; moving-average crosses; 3- to 12-month momentum; November to April only.

Found: Up to 25.6% a year, but worst falls of −77% to −99%: they are all in or all out, and they get out late.

4Trailing stops

Tested: Sell TQQQ after a 15%, 20%, 25% or 30% fall from its high; buy back at a 20- or 50-day QQQ high.

Found: 12.3% to 23.7% a year, worst falls of −85% to −96%: the re-entries buy every dead-cat bounce.

5Volatility sizing

Tested: Hold a target ÷ TQQQ’s 20-day volatility of the account in TQQQ (at most all of it), with targets from 20% to 60%, resized when the share moves 10 points.

Found: At 20%: 18.3% a year and a −50% worst fall since 1986; 23.4% and −32% since 2016. Higher targets earn more and fall further (60%: 27.9% a year, −94%).

6A trend filter with a smaller position

Tested: A third, half or two-thirds of the account in TQQQ, or volatility sizing, only while QQQ is above its 200-day average (2% or 3% band).

Found: A third with a 3% band: 12.5% a year, −40% worst fall, −16% worst year. Adding volatility sizing shrank the dot-com fall to −30%, but whipsaw years cost −49% (2015–16).

7Dips and nights

Tested: Buying sharp QQQ dips (2-day RSI under 10, above the 200-day); holding only overnight, from the close to the next open (from 1999).

Found: The dip rule made 8.0% a year since 2016. The overnight rule had the best Sharpe ratio (return per unit of swing: 1.09) but needs about 750 fills a year and lost 50% in Covid.

The rules that mattered, side by side

The whole 40 years, then the decade the rules were judged on after being ranked on 1986–2015. In every row the money not in TQQQ earns the Treasury-bill rate.

Yearly return, worst fall and worst year of each way of holding TQQQ, 1986–2026 and 2016–2026
Way of holding	Per year, 1986–2026	Worst fall	Worst year	Per year, 2016–2026	Worst fall, 2016–2026
Hold TQQQ (3×)	14.5%	−100%	−92%	39.9%	−82%
Hold QQQ (1×)	14.9%	−83%	−42%	20.3%	−35%
TQQQ only while QQQ is above its 200-day average (3% band)	20.6%	−89%	−70%	38.0%	−62%
TQQQ sized to 30% volatility	21.5%	−75%	−44%	33.1%	−43%
TQQQ sized to 20% volatility	18.3%	−50%	−27%	23.4%	−32%
A third in TQQQ while QQQ is above its 200-day average (3% band)	12.5%	−40%	−16%	16.1%	−25%

“Per year” is the compound yearly return; “worst fall” is the account’s deepest drop from a peak to a low. 2016–2026 is real TQQQ with real opening prices throughout; before 2010 TQQQ is a fund built from the index.

The rule that held up: size TQQQ by its volatility

One number a day decides how much of the account sits in TQQQ. When the fund swings hard the share shrinks; when it is calm the share grows, up to the whole account.

1
Every trading day, after the close
Measure TQQQ’s 20-day realised volatility — how much the fund has been swinging lately: the standard deviation of its last 20 daily log returns, times √252 to turn it into a yearly figure.
2
Then work out the target
Divide 20% by that volatility. That is the share of the account to hold in TQQQ, capped at 100% (never borrowed money on top).
3
If the target is 10 percentage points away from what the account holds
Buy or sell the difference at the next day’s open (a market-on-open order). If it is closer than that, do nothing.
4
All the time
Keep the rest of the account in Treasury bills — a T-bill fund such as SGOV or BIL — so idle cash still earns interest.

Today’s reading (close of October 13, 2026)

TQQQ’s 20-day volatility was 44% a year, so a 20% target meant 45% of the account in TQQQ (a 30% target: 69%; a 40% target: 91%). QQQ was 10.6% above its 200-day average, so the one-third trend rule was in.

About 125 small orders a year since 2016. On average the account held 40% in TQQQ over the 40 years. Fractional shares make the sizing exact.

Crash by crash

The account’s deepest fall between the Nasdaq-100’s peak and its bottom in each crash. A rule that was already out shows a small number.

Deepest fall of each way of holding inside each Nasdaq-100 crash since 1987
Crash	QQQ	TQQQ	200-day filter alone	Sized to 20% volatility	A third, above the 200-day
1987 crash (Aug – Dec 1987)	−40%	−84%	−74%	−37%	−32%
Dot-com (Mar 2000 – Oct 2002)	−83%	−100%	−89%	−50%	−40%
2008 (Oct 2007 – Mar 2009)	−53%	−95%	−56%	−39%	−22%
Summer 2011 (Jul – Oct 2011)	−16%	−44%	−26%	−16%	−9%
Late 2018 (Aug – Dec 2018)	−23%	−58%	−33%	−26%	−11%
Covid (Feb – Mar 2020)	−29%	−70%	−62%	−22%	−25%
2022 (Nov 2021 – Dec 2022)	−35%	−82%	−39%	−32%	−14%
2025 tariffs (Feb – Apr 2025)	−23%	−57%	−34%	−22%	−12%

Before 1999 the QQQ column is the Nasdaq-100 index itself; before 2010 TQQQ is the fund built from the index. The 200-day filter holds all of TQQQ while QQQ is above its 200-day average (3% band).

Why a 200-day filter is not enough on a 3× fund

The popular idea: own the fund only while QQQ is above its 200-day moving average (the average close of the last 200 trading days), with a 3% band so it does not flip back and forth. On QQQ itself it cut the worst fall from −83% to −42%. On TQQQ it did not, because the exit comes only after QQQ has already fallen 25–30% — about −70% in TQQQ.

The same 200-day rule on 1×, 2× and 3× funds
Fund	Per year, 1986–2026	Worst fall	Per year, 2016–2026	Worst fall, 2016–2026
QQQ held, no filter	14.9%	−83%	20.3%	−35%
QQQ (1×) with the filter	12.9%	−42%	17.0%	−25%
2× (QLD) with the filter	18.2%	−73%	28.5%	−46%
TQQQ (3×) with the filter	20.6%	−89%	38.0%	−62%
TQQQ held, no filter	14.5%	−100%	39.9%	−82%

On TQQQ the filter still lost 89% in the dot-com crash and 62% in Covid, and 45% in the flat 2015–16 market from whipsaws: it sold after the August 2015 drop, missed the rebound and bought back near the top, four times over. It spent 14 years under water.

Did it hold up? How we checked

Test 72 rules on one stretch of history and some will look good by luck. These checks are what make volatility sizing stand out.

Ranked on old years, judged on new ones

All 72 rules were ranked on 1986–2015 by yearly return per unit of worst fall, then run on 2016–2026. Sizing to 20% ranked third, then made 23.4% a year with a −32% worst fall: 0.73 of return per unit of fall, against 0.58 for QQQ and 0.49 for TQQQ.

Most rankings did not carry over

Across the 72 rules the order on 1986–2015 barely matched the next ten years (a rank correlation of 0.09). The volatility-sizing rules are the family that stayed at the top.

The built fund against the real TQQQ

Over the 16.7 years both exist, the 3× fund built from the index made 43.1% a year and the real TQQQ 42.9%, with a daily correlation of 0.999.

Nearby settings

Targets from 20% to 60% were tried. 30% made more and fell further: 32.6% a year since 2010 with a −43% worst fall. The 20-day window and 10-point step were not tuned beyond that grid.

Cash that earns nothing

Interest on the idle cash is worth 2–3% a year to these rules. With it set to zero, 20% sizing made 16.1% a year since 1986 (−54% worst fall) and 21.8% since 2016 (−33%).

Every window

Sized to 20%, TQQQ beat holding QQQ with a smaller worst fall in each window: 18.3% vs 14.9% since 1986, 23.5% vs 19.6% since 2010, 23.4% vs 20.3% since 2016.

What did not reduce the risk

Ideas tested for holding TQQQ with less risk, and what happened
Idea	What happened
A 200-day moving-average filter alone	−89% worst fall over 40 years, −62% in Covid. It sells after QQQ has already fallen 25–30%.
A 200-day filter plus volatility sizing	The dot-com fall shrank to −30%, but whipsaws cost more: −49% in 2015–16 and −60% from 1987 to 1989. It ranked below sizing alone on the unseen years.
A fixed share, rebalanced monthly (25–67% in TQQQ)	Worst falls of −69% to −99%. Topping back up keeps buying a falling fund; in 2022 half the account in TQQQ still fell 51%.
Trailing stops of 15–30%	Worst falls of −85% to −96%. Getting back in at a 20- or 50-day high bought every dead-cat bounce — a short rally inside a fall.
Moving-average crosses and 3- to 12-month momentum	Worst falls of −87% to −96%. They only trade the whole position, all in or all out.
November to April only	12.2% a year with a −99% worst fall.
Higher volatility targets (40–60%)	25.4% to 28.3% a year since 1986, but worst falls of −84% to −94%.
Overnight only (buy the close, sell the next open, while QQQ is above its 200-day)	The best Sharpe ratio (return per unit of swing: 1.09 since 1999) and 29.5% a year since 2016, but −50% in Covid and about 750 fills a year. At 0.05% slippage a side instead of 0.02% it gives up about 15% a year. A trading system, not a low-risk holding.

The risks

Even the best rule fell 50% once: from March 2000 to October 2002, and it was not back at its peak until July 2007 — 7.3 years under water.
Taxes are not counted. Volatility sizing is about 125 small orders a year, all short-term; in a taxable account that is the real cost.
Idle cash has to earn interest. That is worth 2–3% a year here; with cash earning nothing, 20% sizing made 16.1% a year since 1986 instead of 18.3%.
Before 2010 TQQQ did not exist: the 1987, dot-com and 2008 numbers come from a fund built from the index. It tracked the real fund closely where both exist (43.1% against 42.9% a year), but it is a model.
The index has no real opening prices before 1999, so trades in 1986–1998 fill at the next close, a day late. The 2016–2026 numbers use real TQQQ with real opens and are the ones to trust most.
Rankings shift: across 72 rules, the order on 1986–2015 barely predicted 2016–2026 (rank correlation 0.09). One good unseen decade is not a promise.

Questions

Is TQQQ safe to hold long term?

Not on its own. In our backtest a 3× Nasdaq-100 fund built from the index lost 100% from March 2000 to March 2009 and never got back to its peak, and the real TQQQ fell 82% in 2022. Over 40 years (1986–2026) holding it made 14.5% a year — less than QQQ’s 14.9% — with a −92% worst year. Since its 2010 launch it made 42.9% a year, with that −82% fall along the way.

Does the 200-day moving average work for TQQQ?

It helps QQQ but does not protect a 3× fund. Holding TQQQ only while QQQ is above its 200-day average (3% band) made 20.6% a year since 1986, but its worst fall was −89% (dot-com), it lost 62% in Covid and 45% in the 2015–16 whipsaws, and it spent 14 years under water. The exit comes after QQQ has fallen 25–30%, which is about −70% in TQQQ. On QQQ itself the same filter cut the worst fall from −83% to −42%.

What is the least risky way to hold TQQQ?

In this test, sizing by volatility: hold 20% ÷ TQQQ’s 20-day volatility of the account in TQQQ (at most all of it), the rest in Treasury bills, and resize when the target moves 10 percentage points. Since 1986: 18.3% a year, −50% worst fall, −27% worst year. Since 2016, years it was not ranked on: 23.4% a year, −32% worst fall. For smaller falls at a lower return, a third of the account in TQQQ while QQQ is above its 200-day average made 12.5% a year with a −40% worst fall (16.1% and −25% since 2016).

How much of my account should be in TQQQ?

This is not advice, but here is what the backtest shows. Fixed shares rebalanced monthly (25% to 67% of the account) had worst falls of −69% to −99% over 40 years. Letting the share move with volatility did better: at a 20% target the account held 40% in TQQQ on average since 1986, less when the fund swung hard. At the October 13, 2026 close TQQQ’s 20-day volatility was 44% a year, so the 20% target meant 45% of the account in TQQQ (69% at a 30% target).

What is volatility targeting for a leveraged ETF?

Holding less of a fund when it has been swinging a lot and more when it is calm. Here: each day after the close, measure TQQQ’s 20-day realised volatility (how much it has swung over the last 20 days, as a yearly figure), hold 20% divided by that number of the account in TQQQ (never more than 100%), and trade at the next open only when the target is 10 percentage points away. In the 2022 bear market that kept the account’s fall to −32%, against −82% for TQQQ held.

How did TQQQ strategies do in the 2008 and 2022 crashes?

From the October 2007 peak to the March 2009 bottom, a 3× fund built from the index fell 95%, QQQ 53%, TQQQ with a 200-day filter 56%, volatility sizing 39% and a third in TQQQ above the 200-day 22%. From November 2021 to December 2022, TQQQ fell 82%, QQQ 35%, the 200-day filter 39%, volatility sizing 32% and the one-third rule 14%.

Do stop-losses work on TQQQ?

Not in this test. Trailing stops of 15% to 30%, buying back at a 20- or 50-day QQQ high, had worst falls of −85% to −96% over 40 years and made 12.3% to 23.7% a year. Each re-entry tended to buy a short rally inside a falling market.

Is TQQQ better than QQQ long term?

Over 40 years, no: 14.5% a year against QQQ’s 14.9%, with a −100% worst fall against −83%. Since 2016 TQQQ made more (39.9% against 20.3% a year) but fell 82% against 35%. TQQQ sized to 20% volatility beat holding QQQ in every window with a smaller worst fall: 18.3% against 14.9% a year since 1986, 23.4% against 20.3% since 2016.

Is this investment advice?

No. It is a backtest of past prices for education. Leveraged ETFs can lose most of their value quickly, taxes are not counted, and past results do not promise future ones.

How the test works

Prices: the Nasdaq-100 index from 1985 (price only), QQQ from March 1999 (with dividends), the real TQQQ from February 2010, and the 13-week Treasury bill rate for cash — 10,338 trading days in all.
Before February 2010 a 3× fund is built from the index: 3× its daily move, less a 0.9% yearly fee and financing of 2 × (the T-bill rate + 0.5%). Over the 16.7 years both exist it made 43.1% a year against the real fund’s 42.9%, with a daily correlation of 0.999.
Signals use yesterday’s close; trades fill at the next open with 0.02% slippage a side. The index has no real opens before 1999, so those trades fill at that day’s close instead, a day late.
Long only. Money not in TQQQ earns the T-bill rate (a T-bill fund such as SGOV or BIL does the same job). Taxes are not counted.
72 rules, ranked on September 1986 – December 2015 by yearly return per unit of worst fall, then judged on January 2016 – October 2026, which played no part in the ranking.
Volatility is the standard deviation of TQQQ’s last 20 daily log returns × √252. “Per year” is the compound yearly return; “worst fall” is the account’s deepest drop from a peak to a low.

Not investment advice

This is a study of past prices, published for education. Leveraged funds, options and futures can lose most of their value quickly, and a pattern seen in the past can break. Nothing here is a recommendation to buy or sell anything. Decide with your own judgment, or with a licensed adviser.