How 3× Leveraged ETFs Actually Behave Over Time

Suppose one investment turned $100,000 into roughly $1.3 million, while a leveraged version of the same exposure grew to more than $14 million. The obvious conclusion is that the leveraged fund was better. Maybe the only additional risk was having to sit through much larger crashes along the way.

That is close to what happened with QQQ and TQQQ over the 15 years ending December 2025. It is real history, and the leveraged result was extraordinary. But it does not prove that the long-term risk was merely a test of nerve. The period began after the global financial crisis and contained a remarkably strong and persistent technology-led bull market.

A 3× ETF does not promise three times the return of its index over a year, a decade or your entire holding period. It targets approximately three times the index’s daily return. Because the multiplier starts over each morning, the eventual result depends on how smooth or volatile those daily returns are, not only whether the market is higher years later.

I walked through the same idea using Canadian-listed HEQL and HEQT in this video comparison.

What happened over 15 years?

The charts below compare TQQQ with QQQ and UPRO with SPY. QQQ and SPY are practical unleveraged proxies for the exposures targeted by the 3× funds; they are not promises of what an investor could have earned with a frictionless three-times strategy.

December 2010–December 2025

Growth of $10,000

Monthly adjusted prices on a logarithmic scale. Distributions and splits are reflected in the data.

QQQ vs. TQQQNasdaq-100 exposure
QQQTQQQ
QQQ and TQQQ: growth of $10,000 Growth of $10,000 in QQQ and TQQQ from December 2010 through December 2025 on a logarithmic scale. $10k$25k$50k$100k$250k$500k$1m$2m201120142017202020232025 QQQ and TQQQ: growth of $10,000 Growth of $10,000 in QQQ and TQQQ from December 2010 through December 2025 on a logarithmic scale. $10k$100k$1m2011201620212025

$128,602 in QQQ · $1,434,513 in TQQQ

SPY vs. UPROS&P 500 exposure
SPYUPRO
SPY and UPRO: growth of $10,000 Growth of $10,000 in SPY and UPRO from December 2010 through December 2025 on a logarithmic scale. $10k$25k$50k$100k$250k$500k$1m$2m201120142017202020232025 SPY and UPRO: growth of $10,000 Growth of $10,000 in SPY and UPRO from December 2010 through December 2025 on a logarithmic scale. $10k$100k$1m2011201620212025

$70,887 in SPY · $430,541 in UPRO

The leveraged funds produced extraordinary ending values in this particular period. This is the strongest argument for holding them long term, and it should not be waved away. But it is an outcome from one starting date, one ending date and two unusually favourable underlying markets. It is not an expected-return estimate.

An actual 3× fund that lost the long-term comparison

The QQQ and TQQQ result came from an unusually favourable market. EDC provides a different real-world case. It launched in December 2008 and still targets three times the daily return of the MSCI Emerging Markets Index. EEM is a practical unleveraged proxy for the same broad exposure.

Start with the fund’s own record: Direxion reports that EDC returned 0.02% annualized from its December 17, 2008 inception through August 14, 2026. That is the appropriate figure when describing EDC’s since-inception return.

To compare EDC directly with EEM, the chart below uses their common Yahoo adjusted-price window. It starts on December 30, 2008 and ends on December 31, 2025, so it answers a different question from Direxion’s current since-inception figure.

Yahoo window: December 30, 2008–December 31, 2025

A surviving 3× fund that underperformed

Growth of $10,000 from the available common daily adjusted-price history, shown at month-end on a logarithmic scale. This begins 13 days after EDC’s official inception, so it is not a since-inception return.

EEM vs. EDCEmerging-markets exposure
EEMEDC
EEM and EDC: growth of $10,000 Growth of $10,000 in EEM and EDC from December 30, 2008 through December 31, 2025 on a logarithmic scale. $100$250$500$1k$2.5k$5k$10k$25k$50k20092013201720212025 EEM and EDC: growth of $10,000 Growth of $10,000 in EEM and EDC from December 30, 2008 through December 31, 2025 on a logarithmic scale. $100$1k$10k$100k200920172025

$31,763 in EEM · $9,352 in EDC

FundTotal returnCAGRMaximum drawdown
EEMUnleveraged proxy+217.63%+7.03%−39.82%
EDC3× daily ETF−6.48%−0.39%−92.54%

Within that common Yahoo window, EEM gained about 218% while EDC lost about 6%. The point is the relative result, not reconciling this particular ending value with a different since-inception period: EDC survived, yet substantially underperformed the unleveraged exposure.

When does the 3× version fall behind?

EDC provides one real outcome. To examine the conditions more broadly, I paired every available QQQ return with TQQQ’s return from the same trading day, from February 2010 through December 2025. I kept those paired returns together in 20-trading-day blocks, randomly drew the blocks with replacement, and used them to build 20,000 different 15-year sequences.

When a QQQ block was selected, the actual TQQQ block from those same dates came with it. The table therefore compares resampled histories of the two real funds, including the fees, financing, tracking differences and daily compounding already present in TQQQ’s adjusted returns.

Two paths assembled from actual paired return blocks

Similar QQQ growth, different TQQQ outcomes

The first chart puts the 19% and 24% volatility QQQ samples on the same tighter scale so the path difference is visible directly. The second chart shows the TQQQ outcomes produced by those exact paired return blocks. These are individual bootstrap samples, not forecasts.

QQQ: smoother path vs. choppier pathNearly identical QQQ CAGRs; the choppier sample had about 24% more annualized volatility.
19.4% vol24.1% vol
Comparison of smoother and choppier sampled QQQ paths Growth of $10,000 in QQQ over two 15-year sequences assembled from paired actual QQQ and TQQQ return blocks. $5k$10k$25k$50kStartYr 3Yr 6Yr 9Yr 12Yr 15 Comparison of smoother and choppier sampled QQQ paths Growth of $10,000 in QQQ over two 15-year sequences assembled from paired actual QQQ and TQQQ return blocks. $5k$10k$25k$50kStartYr 5Yr 10Yr 15

QQQ finished at nearly the same value: $44,560 versus $44,368. Volatility was 19.4% versus 24.1%.

What those same paths did to TQQQEach TQQQ line uses the returns paired with the corresponding QQQ sample above.
Paired with 19.4% QQQPaired with 24.1% QQQ
Comparison of smoother and choppier sampled TQQQ paths Growth of $10,000 in TQQQ over two 15-year sequences assembled from paired actual QQQ and TQQQ return blocks. $5k$10k$25k$50k$100k$250kStartYr 3Yr 6Yr 9Yr 12Yr 15 Comparison of smoother and choppier sampled TQQQ paths Growth of $10,000 in TQQQ over two 15-year sequences assembled from paired actual QQQ and TQQQ return blocks. $5k$10k$25k$50k$100k$250kStartYr 5Yr 10Yr 15

TQQQ finished at $82,053 alongside the smoother QQQ sample, but only $24,600 alongside the choppier one.

20,000 paired bootstrapped 15-year sequences

How often TQQQ underperformed QQQ

Rows show QQQ’s realized compound annual return. Columns show QQQ’s realized annualized volatility. Each cell reports how often the resampled TQQQ history finished lower than the matching QQQ history.

QQQ CAGRQQQ volatility below 20%QQQ volatility 20%–22%QQQ volatility above 22%
Below 5%100.0%2 samples100.0%42 samples100.0%47 samples
5% to 10%50.0%50 samples77.8%473 samples100.0%311 samples
10% to 15%0.0%552 samples0.3%2,355 samples14.3%775 samples
Above 15%0.0%5,435 samples0.0%8,593 samples0.0%1,365 samples

The 10%–15% QQQ return row is especially useful. When QQQ’s volatility stayed below 20%, TQQQ underperformed in 0.0% of those samples. At 20%–22% volatility, it underperformed in 0.3%. Above 22% volatility, it underperformed in 14.3%. QQQ finished in the same broad return range; the amount of volatility changed the TQQQ comparison.

Across all 20,000 sequences, $10,000 had a median ending value of $137,601 in QQQ and $1,823,665 in TQQQ. TQQQ underperformed QQQ in 4.6% of samples and finished below its original $10,000 in 1.1%.

This is not a forecast. QQQ and TQQQ share only a strong, technology-led market history beginning in 2010, and resampling it does not invent market conditions absent from that period. The exercise changes the order and frequency of their actual paired return blocks so we can see which sampled combinations were unfriendly to TQQQ.

What the evidence supports

The conclusion is not that leveraged ETFs must deteriorate over time. TQQQ and UPRO plainly did not. Persistent trends can make daily compounding extraordinarily favourable, while reversals and volatility can make the same structure punishing. Fees, financing and tracking differences add another layer.

Long-term holding can work. What the evidence does not support is assuming that a large eventual gain is assured as long as the investor refuses to sell. The underlying market, the amount of volatility, the sequence of daily returns and the investor’s actual horizon still have to cooperate.

A positive underlying return is not enough. The return must be strong enough relative to the volatility and costs for a 3× daily ETF to outperform over time.

For the product language itself, see ProShares’ TQQQ page, ProShares’ UPRO page and Direxion’s EDC page. Each describes a daily investment objective rather than a promise of three times the benchmark’s cumulative long-term return.

How the figures were calculated. Actual-fund comparisons use Yahoo Finance daily adjusted-close data through December 31, 2025, reflecting distributions and splits. The TQQQ and UPRO comparisons begin December 31, 2010. Yahoo’s common EEM/EDC window begins December 30, 2008, which is 13 days after EDC’s official inception. That display is therefore not a since-inception return. Growth charts use month-end observations for readability; summary figures use daily observations. The bootstrap uses the common QQQ and TQQQ adjusted-return history from February 2010 through December 2025. Their returns are paired by trading date, and 20-trading-day paired blocks are sampled with replacement until each of 20,000 sequences contains 15 years of returns. TQQQ’s sampled returns are its own actual adjusted returns, so the comparison retains the fund’s historical fees, financing, tracking differences and daily compounding. The two illustrated paths were selected from those samples to contrast QQQ CAGR of 10%–15% at below-20% volatility with the same return band at above-22% volatility; the full table shows frequency rather than two selected examples. Block sampling preserves short stretches of actual paired returns but not complete market regimes, and the results remain conditional on the funds’ unusually strong shared source period. QQQ, SPY and EEM are practical unleveraged ETF proxies rather than the funds’ formal benchmarks. Taxes and an individual investor’s currency experience are not modelled.
✉️ Get new posts by email

When I publish something new, an article, a guide update, or a video worth your time, I’ll send one short email. No spam, no schedule, unsubscribe anytime.