Metadata
  • Date
    10.31.25
  • Author
    Thelonious Casey
  • Reading Time
    10mins.
  • Categories
    Leverage
Article

What a difference a day makes

Nearly all of Tesla’s appreciation happens during extended hours. That’s true of US stocks in general, which is more likely to be for benign reasons than evidence of a grand conspiracy. Tesla just happens to be a particularly extreme example:

Monaco GP
Unlike its cars, the manufacturing of Tesla's leverage is anything BUT uniform.

In a 12-month sequence where Tesla shares started at $211 and ended at $334, an investor’s $1,000 staked at 9:30am and sold at 4pm every day would have made $25. Had they flipped it and bought at 4pm and sold the following 9:30am every one of those 249 days, they’d have made $547.

The gap between Tesla’s regular-session and night-mode performance makes it a case study for the fastest growing and least understood subsets of the ETF market: leveraged single-stock funds. LETFs use derivatives to amplify the returns of an underlying stock, usually with a daily reset. Want to go long or short Tesla’s daily performance with leverage of up to 3x? Take your pick of products.

Source: equityquant.dev. Data as at September 3, 2025. “Close to close” calculated using first and last available price printed “opening bell-to-closing bell”, from 9:30am-4pm EST.

By “daily performance” a buyer might assume they’re getting a whole 24 hours. But is that what they’re getting in practice?

As the above chart shows, Tesla delivered a 59 per cent close-to-close gain in the 12 months to end September. What happens when leverage is added? We’ll run the numbers using three of the more established 2x long LETFs, from Direxion (TSLL), GraniteShares (TSLR) and T-Rex (TSLT).

Two times 59 equals 118. Did any of these 2x levered ETFs rise 118 per cent? No — but that’s to be expected. Leverage amplifies volatility, so underperformance to the ETF’s reference asset builds over time, as we’ve been writing about here for approximately two decades.

A fairer question: did the leveraged ETFs hit the promised rate? To work that out, we’ll need to quant. Only a nerd will want to see all the working, and since that describes a significant proportion of our audience, here it is:

To arrive at a promised rate we use the actual leverage ratio, derived from the slope of the LETFs daily excess returns against the underlying daily excess return. Assuming that rebalancing only introduces minor deduction from borrowing, using the corresponding rates listed in SEC filings (for example, 1-month average SOFR), the promised rate can be derived from the following formula:

F^t+1−F^tF^t=β(St+1−StSt)−(β−1)rt,F^0=F0\frac{\hat{F}_{t+1} - \hat{F}_{t}}{\hat{F}_{t}} = \beta \left( \frac{S_{t+1} - S_{t}}{S_{t}} \right) - (\beta - 1) r_{t}, \quad \hat{F}_{0} = F_{0}

Where β\beta is the actual leverage ratio, SS is the price of Tesla, the underlying, and F^\hat{F} is the promised price, FF is the price of the real LETF, and rr is the risk-free rate.

Working on that formula, the promised rate for a Tesla 2x long pops out at around 48 per cent. Did any of the ETFs get near that? Also no. Actual performance for TSLR was 39 per cent, for TSLL it was 36 per cent, and for TSLT it was 31 per cent. Sure, in isolation it looks bad. But in context it also looks bad.

Alibaba and Nvidia shares were similarly strong in the same 12 months and the performance of 2x long ETFs referencing their shares was relatively OK:

Source: equityquant.dev. Data as at September 3, 2025. Table highlights exclude ASTS, CRWD and ORCL whose stocks had a similar return but whose LETFs were not trading through the same 249 period. Nvidia has a fourth 2x long LETF which was excluded as it also was not trading through the same 249 period.

That’s in spite of Alibaba and Nvidia enduring the same fun-house mirror outperformance as Tesla during extended hours:

Stock Extended hours gain Intraday gain
Alibaba 60.11% 4.22%
Nvidia 41.45% 14.06%
Tesla 54.74% 2.45%

So given the same underlying gain, why do those other stocks’ LETFs outperform Tesla’s point-to-point over the same time period?

Our first problem is to define what’s meant by a day, because the rules are not clear.

The SEC’s ruling 6c-11 in 2019 that paved the way for LETFs has numerous references across the 259 pages to “daily return”, “daily objectives” and “daily reset”, but doesn’t offer any specifics as to when days should start and end. References elsewhere in the document to SEC rule 22c-1 — which says an investor gets today’s price if the order arrives before 4pm, otherwise they get the next day’s price — is the closest it comes to offering a definition.

Today, most funds calculate NAV per share at the close on the major US stock exchanges, which typically means 4pm Eastern Time. GraniteShares, in its LETF documentation, cites the same point for its set and reset:

“The Fund aims to generate 2 times the daily performance of the Underlying Stock for a single day. A “single day” is defined as being calculated “from the close of regular trading on one trading day to the close on the next trading day.”

The reality is not so simple. LETFs mostly use swaps, which are agreements with terms, usually involving a benchmark plus a spread. Different ETFs and their counterparties may have different rebalancing thresholds and different rebalancing frequencies, with no explicit SEC legislation to keep all issuers on the same page.

It’s naive to assume LETFs sell their entire position at the end of a trading day and buy at the beginning of the next. Within reason, they can press the reset button whenever they wish. Filings for Graniteshares’ TSLR suggest ETF flow executing between 4-5pm, in other words, well after market close.

The marked divergence between the three 2x long Tesla ETFs shown above can also in part be explained by their different swap benchmarks, spreads and counterparties. There’s more to underperformance than “leveraged volatility drag”.

Another problem is unique to Tesla, at least in public markets: Elon Musk.

The CEO and Technoking’s tweet-first-think-later tendencies have contributed to Tesla’s high and persistent average volatility through the year. Higher than even Alibaba or Nvidia. Imagine trying to roll contracts in perpetuity with such a turbulent underlying.

As a general rule, high volatility is a bad thing for higher-leveraged ETFs. The optimal leverage ratio has sigma as its denominator — which is a fancy way of saying it goes up when expected returns rise, and goes down as volatility grows. We could show you that as an algorithm, so will:

The optimal leverage ratio — a generalised Kelly criterion — is obtained by taking the optimal L to maximise the drift term of the log price process, can be written as:

L∗=arg max⁡L∈R(r+L(μ−r)−12L2σ2)=μ−rσ2L^{*} = \argmax_{L \in \mathbb{R}} \Big(r + L(\mu - r) - \frac{1}{2} L^2 \sigma^2\Big) = \frac{\mu - r}{\sigma^2}

Where rr is the risk-free rate, (𝜇−r)(𝜇-r) is the risk premium, and σ\sigma is the volatility of the reference price.

The key point is that optimal leverage scales with volatility squared. Volatility-sustaining means the best amount of leverage is lower. Often lower than 2x. For that reason alone, the performance of the 2x LETF is on shaky ground.

The 3x LETFs? Even shakier. Over the period that Tesla rose 59 per cent, 3x leveraged Tesla fell between 25 and 29 per cent. As products, they are bets on volatility that have almost nothing to do with Tesla’s EV sales.

Of course, the end investor will often hold for a matter of days so doesn’t necessarily feel all this pain. Drag, spreads, attrition and vol-associated underperformance are effects that build over time. And LETFs are, to their credit, unambiguously marketed as short-term trading tools.

From this marketing, a person might think single-stock LETFs are niche products, yet they make up a quarter of all new ETFs this year. And the sluice of new ones being approved by the SEC is wide open.

Tesla’s 2x and 3x LETFs aren’t so much a cautionary tale as they are a perfect short — if only the short could be executed frictionlessly. Maybe inverse-double-short-leveraged ETFs will be the next big thing?