Open the analyzer →

Home › Explore Investing Strategies › Leveraged ETF strategies

Leverage cross-rebalancing: 3× on drops, 2× on breakouts, 1× when overheated

A community cycle: TQQQ bought in steps on a big drop, down to QLD at a new QQQ high, QLD trimmed into QQQ when RSI overheats. Structure and risk.

In one line

"Buy TQQQ on the drop and switch to QLD on the way up" and "sell QLD on the way up and add TQQQ on the drop" describe the same strategy in a different order. The rule needs a judgment about where the market is, so it is used with only a portion of capital, not all of it.

Backtest results for this strategy: TQQQ, QLD and QQQ cross-rebalancing: how does it compare with simply holding?

Article

Simply holding TQQQ means losing more than 80% of the account in a decline like 2022. Holding only QQQ leaves you wishing for leverage in rising markets. In between, the idea of raising leverage only after a fall and lowering it after a rise keeps coming up in forums, and this article covers one rule built on it. The heart of the rule is whether the moments to change leverage can be fixed in numbers in advance.

Community posts describe this method in two ways. One says "buy TQQQ on the drop, sell it on the way up and buy QLD". The other says "trim QLD into QQQ on the way up, and add TQQQ on the drop". Both describe the same strategy in a different order. This article is the operator's own rewrite of the strategy's structure and risks, based on a rule Gyuni posted on Threads.

The strategy

  • When the market drops hard, buy TQQQ (3×) in steps.
  • When QQQ breaks its prior high, sell all TQQQ and buy QLD (2×). Holding 3× for a long time makes volatility decay worse, so leverage steps down.
  • When a pre-set signal such as RSI overheating arrives, trim QLD and buy QQQ (1×).
  • On the next drop, start over from the top.
The strategy. Buy TQQQ on a big drop, sell all TQQQ and buy QLD when QQQ breaks its prior high, trim QLD into QQQ when RSI overheats. Repeat on the next drop. ① Strategy: 3× on the drop, 2× on the breakout, 1× on overheating X is time, Y is QQQ (1×) price. Leverage goes up at one moment only: a big drop. It comes down at two: prior-high breakout and overheating. So 3× is held only briefly. prior high Buy TQQQ big drop · buy in steps Buy TQQQ next drop · same rule Prior-high breakout sell all TQQQ → buy QLD RSI overheated trim QLD, buy QQQ

Leverage goes up at only one moment, after a big drop, and comes down at two: the prior-high breakout and overheating. So most of the time leverage is low, and 3× is held only briefly. That is the biggest difference from holding TQQQ all the time.

Why it is hard

The rule only works if you judge whether this is a big drop and whether the market is overheated. That means it contains a forecast of where the market is. During a drop you cannot buy because it looks like it will fall further, and during a rise you cannot trim because it looks like it will rise further. That is the human default (Investor psychology). The rule looks simple, but it is hard to keep in practice.

Numeric rules

Here is a simple example. The yardstick is the QQQ (1×) price, not the 3× ETF.

  • When QQQ is 10% below its high, move 10% of total assets into TQQQ. At 20% down move 20%, at 30% down 30%, and at 40% down 40%. Sell QQQ first, then QLD if QQQ is not enough.
  • When QQQ breaks its prior high, sell all TQQQ and buy QLD.
  • When a pre-set signal such as RSI overheating arrives, trim QLD and buy QQQ.
A simple numeric example. Move 10% of total assets into TQQQ at -10% from the QQQ high, 20% at -20%, 30% at -30%, 40% at -40%. Sell all TQQQ for QLD on a prior-high breakout; trim QLD into QQQ when RSI overheats. ② Numeric rules: write down entry, switch and trim in advance (simple example) Drawdown ladder: simple example. Measured on QQQ (1×), not on the 3× ETF QQQ -10% from high10% of total → TQQQQQQ -20% from high20% of total → TQQQQQQ -30% from high30% of total → TQQQQQQ -40% from high40% of total → TQQQ Switch: QQQ breaks prior high → sell all TQQQ, buy QLD Trim: RSI overheated → reduce QLD, buy QQQ An eye for channel tops and bottoms helps; if not, write numeric rules and follow them strictly.

The numbers are only an example. The point is to set in numbers, in advance, when you buy, when you switch and when you trim. An eye for whether price is near the top or the bottom of its channel helps. If that is hard, write numeric rules and follow them as written (Plan before entry).

Allocation

This is what the original stressed most. Running all of your capital this way is very hard. The base is an S&P 500 or Nasdaq 1× ETF at the core, with this rule used only for the portion aimed at extra return. Apply a rule that contains a forecast to everything, and one wrong call shakes everything (Asset allocation).

Keep most of the capital in a 1× index ETF; run this strategy with a portion only. ③ Allocation: a portion, not the whole Core: S&P 500, Nasdaq 1× ETF held without predicting the market Satellite: this rule TQQQ ⇄ QLD ⇄ QQQ The proportions are not an example; they only mean "a portion". The rule does not say how much. This rule reads the market's position (how far down, how overheated) and changes leverage. That is a forecast. Run on all your capital, one wrong call shakes everything.

Results from this site's engine (as of September 2026)

This site turned a rule Gyuni posted on Threads into numbers. Holding cash was left out on purpose to keep the rule simple: the money is always fully in TQQQ, QLD or QQQ. Each time QQQ falls 10, 20, 30 and 40% from its high, 10, 20, 30 and 40% of total assets move into TQQQ, from QQQ first and then from QLD. When QQQ regains its previous high, all TQQQ is switched to QLD. When QQQ's 14-day RSI reaches 70 or more, half of the QLD moves into QQQ. This trim happens once per recovery. The RSI 70 line and the half are the operator's values. The run starts with $10,000 and no further deposits, reinvests dividends and charges 0.15% per trade.

PeriodCross-rebalancingHold QQQHold QLDHold TQQQ
5 years (Oct 2021 to Sep 2026)25.1% a year
max −50.4%
16.2% a year
max −35.1%
21.5% a year
max −63.7%
21.2% a year
max −81.7%
10 years (Oct 2016 to Sep 2026)30.1% a year
max −55.0%
20.9% a year
max −35.1%
33.5% a year
max −63.7%
40.9% a year
max −81.7%
15 years (Oct 2011 to Sep 2026)30.4% a year
max −55.0%
20.4% a year
max −35.1%
34.7% a year
max −63.7%
45.4% a year
max −81.7%

In all three periods the max drawdown was smaller than holding QLD (−63.7%) or TQQQ (−81.7%) and larger than holding QQQ (−35.1%). The annual return beat holding QQQ in all three periods. Against holding QLD or TQQQ it was higher over 5 years and lower over 10 and 15 years.

Moving the start to each January from 2012 to 2021, ten runs in all, the annual return beat holding QQQ in all 10 and beat holding QLD and TQQQ once each (January 2021). Max drawdowns ranged from −52.1% to −55.0%. The year-by-year numbers are on the cross-rebalancing question, which is recalculated every month.

When the rule worked and when it did not

It came closest to holding QLD when a large fall came soon after the start. The January 2021 start, when QQQ fell 35.1% from its high in the first two years, is the only one that beat both QLD and TQQQ (24.8% a year against 23.3% and 24.7%). The January 2020 start, with a 28.6% fall, was also close (30.5% a year against 31.2% for QLD). Those are the stretches in which the drop ladder actually bought TQQQ.

When the first years brought only small falls, it trailed holding QLD by more. The January 2013 start (27.5% a year against 33.7%) and the January 2014 start (27.1% against 31.1%) saw QQQ fall only 8.2% and 13.9% in their first two years. Because half of the QLD moves into QQQ when overheated, the rule rises less than QLD alone while prices climb.

It beat holding QQQ all 10 times and holding QLD only once. The Nasdaq 100 rose for most of the period since 2011, and in such a period staying leveraged with no cash helps. With no cash, the max drawdown was larger than holding QQQ every time, though smaller than holding QLD. This rule gives up part of the return of holding QLD in exchange for a smaller drawdown.

Related concepts and search terms

This rule also goes by a community name, but each part belongs to a known idea.

  • Dynamic leverage, tactical asset allocation: the whole family of methods that change weights or leverage with market conditions. This is the parent concept, and moving between 3×, 2× and 1× by regime is one form of it.
  • Scaling in, averaging down: buying in steps depending on how far the market has fallen. This article's drawdown ladder is this.
  • Volatility targeting: lower leverage when volatility rises and raise it when volatility falls. It points the same way as going to 1× when overheated, but the yardstick is volatility, not RSI.
  • Risk parity, HFEA: the best-known leveraged strategy in the Bogleheads community, which mixes 3× stocks (UPRO) with 3× long-term Treasuries (TMF) to spread volatility. It differs from this rule in that it handles risk through the asset mix instead of changing leverage (original Bogleheads thread). This site's calculation is in HFEA (UPRO, TMF).
  • CPPI (constant proportion portfolio insurance): add more of the risky asset as the cushion above a loss floor grows. That is the opposite direction from this rule, which lowers leverage as the market rises (Wikipedia).
  • Volatility decay: the loss of value a leveraged ETF suffers from repeated ups and downs when held for a long time. It is the term that explains why this rule holds 3× only briefly.

Search terms: dynamic leverage, tactical asset allocation, scaling in, volatility targeting, HFEA, leveraged ETF decay, 무한매수법, 밸류 리밸런싱.

How it differs from similar methods

The "James Lee" method (a YouTuber). It is known for keeping a 1× index ETF as the base position and adding 2× and 3× positions as the index falls. The skeleton, raising leverage as the market falls, is the same as this article's rule. The difference is how you get out. This rule also sets in numbers when to step 3× down to 2× on a prior-high breakout and 2× down to 1× when overheated, and then repeats that cycle. If only the way up is defined and the way down is loose, you end up holding 3× for a long time during the recovery.

The casino Martingale. After every loss, you bet double the previous stake. The drawdown ladder (10% at -10%, 20% at -20%, 30% at -30%, 40% at -40%) looks similar because it buys more the further the market falls. But the premise is different. In Martingale every round is independent and the expected value is negative, so a long losing streak runs into the table limit or your bankroll first, and one loss wipes out all the earlier gains. This rule rests on the premise that a broad index eventually gets back to its prior high, and it sets the end of the ladder (100% of total assets) in advance. The resemblance is in the shape. The difference is in the premise and the limit. Still, in a long slump where that premise fails, this rule shows the same weakness as Martingale, because there is no rule below the last rung of the ladder.

Drawbacks and risks

  • In a correction that never reaches the first rung (-10% in the example), nothing happens. If the market rises in the meantime, this portion stays mostly in 1× or 2×.
  • There is no rule below the last rung. At -40% this whole portion is in the 3× ETF, with no cash cushion. If the market falls further, the 3× ETF ends up far deeper than that.
  • A prior-high breakout can take years. That means holding 3× the whole time, and in a sideways market only volatility decay piles up.
  • Overheating measures (RSI, the top of a channel and so on) give different values depending on the indicator and how it is calculated. In a rally to new highs the overheated state lasts a long time, so you end up trimming early.
  • Every switch triggers capital gains tax. The more often you trade, the more tax cuts into the return.

Operator's assessment

The inputs this rule reads are the same ones Confluence Zone reads through regime and RSI. What differs is how they are used. This site treats regime as a gate and uses the score only for timing inside it. This rule uses that judgment to change leverage itself. The basket method ignores regime and runs on one target amount, so it is the opposite of this rule. The more a rule tries to read the market, the harder it is to keep. In exchange, this rule gets entries in downturns and trimming in overheated stretches.

What this article does not cover

This article does not recommend any ETF or set drawdown or weight figures. The numbers in the article are the source's simple examples. This site's Backtesting tab has an experimental model that turns these rules into numbers, but it does not reproduce the source's results, and holding cash was left out on purpose to keep the rule simple. The overheating line (RSI 70) and the share moved (half of QLD) were chosen by the operator. Volatility decay, tax, currency and product-structure risks of leveraged ETFs are yours to check. Everything here is reference information based on past, public data and is not investment advice.

Articles and questions on the same topic

Leveraged ETFs (TQQQ, QLD, SOXL)

Asset allocation and rebalancing

Related

Cross-rebalancing backtest results · The basket method · Infinite buying · Value rebalancing · Asset allocation · Investor psychology · Plan before entry · Reading indicators

Basics and trading strategies

← The basket method: fill 240 slots, then open one a month · Contents · Infinite buying: buy in 40 steps, sell at 10% above the average cost →

See these metrics on a real stock →

Confluence Zone is an analysis tool for information and research; it is not investment advice, a solicitation, or a trading instruction. Investment decisions and their outcomes are your own responsibility.
About & methodology · Explore Investing Strategies · Glossary · Questions and answers · Rule status · Sample analysis · Pricing · Terms · Investment disclosure · Privacy policy · Open the analyzer · 한국어 · 日本語 · 繁體中文 · 简体中文