If you've ever held a 2× or 3× leveraged ETF for more than a few weeks and felt something was off — like the math didn't add up — you weren't imagining it. That gap has a name, and it's predictable enough that you can estimate it before you put a dollar in.
This post is about volatility decay in leveraged ETFs: what it is, where it comes from, and how to read it. We'll cover volatility drag, compounding decay, beta slippage, path dependency, and the surprising flip side — reverse slippage — using real data from NVDL, TSLL, TQQQ, and the inverse fund NVD.
The simple version
Imagine you have $100 and a magic stock. The stock goes up 10% one day, then down 10% the next day.
Regular stock:
- Day 1: $100 → $110 (up 10%)
- Day 2: $110 → $99 (down 10%)
- You end with $99. Lost a buck.
Okay, weird, but not a disaster.
Now here's a 2× leveraged ETF that promises to double whatever the stock does each day:
- Day 1: stock goes up 10%, so you go up 20%. $100 → $120
- Day 2: stock goes down 10%, so you go down 20%. $120 → $96
- You end with $96. Lost four bucks.
The regular stock lost $1. The "double power" version lost $4 — way more than double the loss! 😱
That's it. That's slippage. The "more leverage = more reward" intuition is correct on any single day, but over many days the math compounds against you whenever the price bounces around.
So what is a leveraged ETF actually promising?
Leveraged ETFs (like NVDL, TQQQ, SOXL) promise to deliver L times the daily return of the underlying — not L times the cumulative return over weeks or months.
That word — daily — is the entire story. Every evening, the fund rebalances back to the target leverage. Tomorrow it starts fresh at 2× (or 3×, or −2×). It does not track 2× of the year-to-date move.
This is called a daily reset, and it's why the long-term return of a leveraged ETF is almost never L × (underlying return). It's path-dependent.
The volatility drag formula
For a daily-rebalanced leveraged ETF, the expected drag from compounding is:
Drag ≈ −½ × L × (L − 1) × σ² × T
Where:
- L = leverage (2 for NVDL, 3 for TQQQ, −2 for NVD, −3 for SQQQ)
- σ = daily volatility of the underlying (standard deviation of daily returns)
- T = number of trading days you hold
Some quick intuitions you can read off the formula:
- Drag grows with the square of volatility. A stock that moves 3% a day produces ~4× the drag of one that moves 1.5% a day. This is why volatile single-stock ETFs (NVDL, TSLL, MSTU) bleed faster than broad-index ETFs (TQQQ, SPXL).
- Drag is non-linear in leverage. A 3× ETF doesn't suffer 1.5× the drag of a 2× ETF — it suffers 3× the drag (because
L(L−1)goes from 2 to 6). - Drag is roughly linear in time. Hold twice as long, lose roughly twice as much to drag.
- Inverse ETFs aren't safe. Plug in L = −2:
−½ × (−2) × (−3) × σ² × T = −3σ²T. Still negative. We'll come back to this.
This formula is the well-known "volatility drag" or "beta slippage" approximation from the leveraged-ETF literature (Cheng & Madhavan, 2009 — link at the end). It's an approximation, not a law: the real-world slippage also includes the expense ratio, swap financing costs, and execution slippage from daily rebalancing. More on that later.
A real example: NVDL vs NVDA
NVDL is the GraniteShares 2× Long NVIDIA Daily ETF — the most popular 2× NVDA fund. Here's the trailing year (May 22, 2025 → May 20, 2026, 249 trading days):
The numbers:
- NVDA underlying: +68.2%
- Naive 2× expectation: +136.5% (what most people assume they're buying)
- NVDL actual return: +126.7%
- Slippage: −9.7%
So someone who bought NVDL a year ago captured roughly 1.86× the underlying's move, not 2×, despite NVIDIA being in a strong uptrend.
The volatility drag formula predicts −11.3% (NVDA's daily σ ≈ 2.13%, over 249 trading days, with L=2). Almost dead-on. The small remaining gap (~+1.6%) is the trend-bonus / fee tradeoff, which we'll get to in a moment.
Why volatility eats leveraged ETFs
The cleanest way to see this is a fully synthetic example.
Imagine a stock that goes +10% one day, then −9.1% the next day, alternating forever. After each pair of days, the underlying is back to exactly where it started (1.10 × 0.909 ≈ 1.00).
Now what happens to a 2× ETF holding the same stock?
- Day 1: +20% → $120
- Day 2: −18.2% → $98.18
Already 2% in the hole. And it loses 2% every single round trip. Over 60 trading days of this seesaw:
A flat market → 1× ETF flat, 2× ETF down ~42%, 3× ETF down ~79%. Nobody got the market direction "wrong" here. The volatility alone did the damage.
This is compounding decay (also called volatility decay, path dependency, or daily-reset effect — the terminology is interchangeable). It's not a fee, it's not a bug, it's not the fund manager's fault. It's the math of compounding daily resets in a volatile market.
Reverse slippage: when leverage beats L×
Here's the part most articles skip: the volatility drag formula assumes the underlying mostly bounces around. In a smooth, sustained uptrend, daily compounding actually works for you.
This is sometimes called the trend bonus or reverse slippage. Each daily gain compounds onto a larger base, so by the end of a strong trend, a leveraged ETF can deliver more than L × the underlying's total return.
Real example — TQQQ (3× QQQ) over the trailing year (May 22, 2025 → May 20, 2026):
- QQQ: +38.7%
- Naive 3× expectation (arithmetic, i.e. 3 × QQQ's cumulative return): +116.2%
- TQQQ actual: +124.1%
- Slippage: +7.8% (positive!)
QQQ is much less volatile than NVDA or TSLA (daily σ ≈ 1.0% vs 2.1–2.9%), so volatility drag was only about −7.7%. The relatively smooth uptrend produced enough positive compounding to overcome the drag and add a small bonus on top.
Takeaway: Slippage isn't always negative. In low-volatility, trending markets, leveraged ETFs can outperform their stated leverage. In high-volatility, choppy, or sideways markets, they bleed.
Inverse ETFs: a different kind of pain
Inverse leveraged ETFs (NVD = −2× NVDA, SQQQ = −3× QQQ, SOXS = −3× semis) suffer the same structural drag as long ETFs — the formula is symmetric:
For NVD (L = −2): Drag ≈ −½ × (−2) × (−3) × σ² × T = −3σ²T
For SQQQ (L = −3): Drag ≈ −½ × (−3) × (−4) × σ² × T = −6σ²T
But inverse ETFs in an uptrend have a unique problem on top of drag: they're constantly betting against a winning trade and rebalancing the loss into a smaller base each day.
The good news: a daily-rebalanced inverse ETF mathematically cannot lose more than 100%. The bad news: it can come painfully close.
NVDA went up +68% over the past year (May 22, 2025 → May 20, 2026). A naive "short 2× NVDA" position would imply a −136% return, which is impossible for a fund. NVD's actual return was −76%:
So in one sense NVD did "better" than the naive expectation (−76% vs −136%). But the holder still lost three-quarters of their money. This is why inverse leveraged ETFs are generally considered trading instruments, not hedges — over any meaningful timeframe, the decay overwhelms whatever directional protection you thought you were buying.
When slippage is worse than the formula predicts
The volatility drag formula is a floor for what your slippage should be in expectation. Real slippage can be much worse, because the formula doesn't include:
- Expense ratio (NVDL: 1.15%/yr, TSLL: 0.95%/yr, TQQQ: 0.84%/yr)
- Swap financing costs (the embedded cost of borrowing — roughly SOFR + spread, easily 5–6% annually right now)
- Rebalancing execution slippage at the close
- Tracking error from imperfect swap pricing or cash drag
Real-world example — TSLL (2× TSLA) over the same trailing year (May 22, 2025 → May 20, 2026):
| Metric | Value |
|---|---|
| TSLA (underlying) | +22.4% |
| Naive 2× expectation | +44.7% |
| TSLL actual return | +0.3% |
| Total return gap (actual − naive) | −44.4% |
| Theoretical volatility drag | −21.7% |
| Residual (gap − drag) | −22.7% |
That ~22% residual is mostly the operational piece — fees, financing, swap costs, plus TSLA's particularly violent intraday whipsaws that make daily rebalancing expensive. Strictly speaking the residual also includes higher-order compounding terms the linearized drag formula misses, so it's not a pure operational measure. The mean daily tracking error (next section) is the cleanest read on pure operational cost. Either way, TSLL is a cautionary tale: even when Tesla was up over a year, holders made nothing.
How to read daily tracking error
One last useful metric. For each day, compute:
Daily tracking error = ETF_daily_return − L × Underlying_daily_return
This isolates the fund's operational slippage from compounding. NVDL averages about −4.4 basis points per day of tracking error (≈ −11% annualized), with very low day-to-day variance (std ≈ 12 bps). Translation: NVDL hits its daily 2× target tightly, but there's a small consistent drag from fees and financing that compounds to about 10% over a year. NVD (−2× NVDA) has much higher daily tracking error variance (~73 bps std) because shorting a high-volatility single stock is operationally messy.
TL;DR
| Concept | What it means | Sign |
|---|---|---|
| Volatility drag / decay | Structural loss from daily compounding in volatile markets | Always negative |
| Beta slippage | Same thing — gap between L× and the ETF's actual return | Usually negative |
| Trend bonus / reverse slippage | Positive gap from sustained smooth uptrends | Can be positive |
| Daily tracking error | How well the fund hits its daily target | Small, mostly fees |
| Path dependency | Same start and end price ≠ same ETF return — the path matters | — |
The headline rule: leveraged ETFs deliver L× daily, not L× over time. Whether that gap (slippage) helps you or hurts you depends on the volatility and the path. In most markets, most of the time, it hurts.
If you want to see slippage in action across different stocks, our single-stock ETF roundups are a good place to start: NVIDIA, Tesla, Microsoft, Bitcoin, QQQ. Or compare any two ETFs directly.
I worked hard to keep the numbers, formulas, and references in this post accurate, but if you spot a factual error — a mis-stated expense ratio, a return that no longer matches the data, a broken citation, anything — please reach out at contact@leveragedposition.com. Much appreciated!
References & further reading
- SEC Investor Bulletin — Leveraged and Inverse ETFs. The regulator's own warning on how compounding affects long-term returns: https://www.sec.gov/investor/pubs/leveragedetfs-alert.htm
- FINRA Investor Alert — Leveraged and Inverse ETFs: https://www.finra.org/investors/insights/leveraged-and-inverse-etfs-specialized-products-extra-risks-buy-and-hold-investors
- Cheng, M. & Madhavan, A. (2009). "The Dynamics of Leveraged and Inverse Exchange-Traded Funds." The foundational academic paper deriving the volatility drag formula and path dependency results. Journal of Investment Management, Vol. 7, No. 4.
- ProShares — Understanding Leveraged & Inverse ETFs: https://www.proshares.com/our-etfs/strategies/geared-investing
- GraniteShares NVDL prospectus (the source-of-truth on NVDL's daily objective and the exact risk disclosures around path dependency).