How I Ran the Same Strategy Across Three Leveraged ETFs and Walked Away With a 96% Win Rate
A story about volatility harvesting, Beta adaptability, and why “a little SOXL” became the final answer
If you like to receive the real-time performance of the strategy discussed in this article, please join the
Over the past six months, I’ve been refining an hourly trading strategy built on the S&P 500. It started as a simple idea: let the hourly chart define the trend, and use the 15-minute chart to harvest counter-trend moves.
When I ported the exact same signals to SPXL, TQQQ, and SOXL, the numbers started telling a story I never expected.
Here’s the spoiler: this isn’t a story about which asset is best. It’s about how a strategy adapts itself across different Beta environments. The final answer wasn’t a single pick. It was a triangle.
The Starting Point: A Modest Validation on SPY
The core logic is simple. v1 is the hourly strategy — it sets the directional bias and carries the main position. v2 layers a 15-minute counter-signal on top: **only when the 15-minute signal goes against v1’s direction does the overlay trigger**. Once the overlay signal ends, the position snaps back to v1’s original track.
What I didn’t fully appreciate until I dissected the data was this: **the overlay condition itself acts as a natural quality filter.** The 15-minute strategy generates 19 signals independently, but only 9 of them actually meet the “against v1” threshold and get adopted into v2. The other 10? They either align with v1 (no overlay needed) or expire too quickly to trigger the condition.
On the S&P 500 index, v2 delivered this scorecard (December 26, 2025 to June 18, 2026, roughly six months):
SPY
- Buy Win Rate: 93.33% | Sell Win Rate: 100% | Total: 96.43%
- Buy Avg Return: 1.76% | Sell Avg Return: 1.67%
- Compound Return: 60.22%
Solid. But the S&P 500 is a low-volatility playground. A 1.76% average buy return felt like a whisper in a world where leveraged ETFs scream.
So I did the obvious thing: map the identical strategy signals — same entries, same exits, same everything — onto 3x leveraged ETFs.
What came back rewired how I think about this entire system.
The Mapping: Same Signals, Different Universes
I picked three underlyings: SPXL (3x S&P 500), TQQQ (3x Nasdaq 100), SOXL (3x Semiconductors). Same 28 signals. The only variable: which asset the orders hit.
SPY
- Buy Win Rate: 93.33% | Sell Win Rate: 100% | Total: 96.43%
- Buy Avg Return: 1.76% | Sell Avg Return: 1.67%
- Compound Return: 60.22%
SPXL
- Buy Win Rate: 93.33% | Sell Win Rate: 100% | Total: 96.43%
- Buy Avg Return: 5.42% | Sell Avg Return: 4.93%
- Compound Return: 294.99%
TQQQ
- Buy Win Rate: 100% | Sell Win Rate: 92.31% | Total: 96.43%
- Buy Avg Return: 9.88% | Sell Avg Return: 6.65%
- Compound Return: 723.19%
SOXL
- Buy Win Rate: 86.67% | Sell Win Rate: 84.62% | Total: 85.71%
- Buy Avg Return: 27.88% | Sell Avg Return: 11.28%
- Compound Return: 5,693.63%
Stare at the win rate column for a second. SPXL, TQQQ, and SPY all share the exact same number: 96.43%. But compound returns span two orders of magnitude — from 295% to 5,694%.
Here’s what this is telling you: all three leveraged ETFs use 3x exposure. The multiplier is identical. The return gap comes almost entirely from the underlying index Beta.
The strategy doesn’t change. The Beta does.
What This Strategy Actually Is: A Neutral Beta Amplifier
I started to understand the true identity of this system.
It doesn’t predict direction. v1’s hourly signals follow the trend, but that’s momentum-following, not forecasting. The 15-minute overlays capture counter-trend pullbacks — again, pure following.
The only thing it does is execute buy and sell actions efficiently on assets that have trend and volatility.
The higher the Beta, the stronger the trend, the deeper the pullbacks. The same strategy produces exponentially more output, not because it got better, but because the underlying moved more.
But Beta doesn’t hand out free rides.
The SOXL Temptation — and Its Price Tag
SOXL printed a 5,694% compound return over six months. Average buy return: 27.88%. Average sell return: 11.28%. Those numbers will stop any trader mid-scroll.
But the win rate dropped from 96.43% to 85.71%. Two buy failures. Two sell failures.
This is where a mechanism I discovered during SPY testing becomes critical: **15-minute false signals die fast.**
On SPY, the 15-minute strategy had one sell signal that failed — but here’s the key: **it expired before it could trigger v2’s overlay condition.** Time itself acted as a filter. Only 15-minute counter-signals with enough staying power survived long enough to be adopted by v2.
Think about what this means. The overlay logic isn’t just combining two strategies. It’s selectively importing the 15-minute signals that are strong enough to contradict v1’s hourly bias. The weak ones — the ones that reverse quickly — never meet the “against v1” threshold. They self-eliminate.
On SOXL, because volatility is higher and moves are faster, those two bad sell signals triggered the overlay before they died. Brief detours, quickly corrected.
But here’s the key: the cost was tiny. Sell average return was still 11.28%. That means the two losers were rounding errors, while the 11 winners were massively amplified by SOXL’s violent swings.
This is classic asymmetric bet structure: lose small, win huge.
TQQQ: The Sweet Spot
If SOXL was the stress test, TQQQ is the live-fire sweet spot.
Fifteen buys. Fifteen wins. 100% buy win rate.
In this backtest window, the Nasdaq 100’s trending character was almost impossibly clean. v1’s directional calls were never truly challenged on TQQQ. Every single 15-minute counter-trend buy signal nailed the exact end of a pullback.
One sell failure (92.31%), but the sell average return of 6.65% tells you the loss was minimal.
Final result: identical 96.43% total win rate as SPY, compound return of 723%.
What makes this number special isn’t the absolute figure — SOXL is higher. It’s that the win rate didn’t drop. 723% was achieved while winning nearly every single trade. That matters psychologically when real money is on the line. A strategy you can actually execute is worth more than a strategy that looks better in a spreadsheet.
SPXL: The Underrated Anchor
SPXL looks modest next to TQQQ — 295% versus 723%. But it carries a unique asset:
100% sell win rate.
Under the S&P 500’s index structure, not a single 15-minute counter-sell signal triggered a false overlay. While TQQQ had that one small sell-side blemish, SPXL sailed through the same period untouched.
This isn’t random. The S&P 500’s lower volatility means 15-minute counter-signals need more conviction to trigger against v1’s hourly bias. The overlay filter is stricter. What gets through is higher quality.
This creates a natural hedge. Two high-win-rate assets, responding to the same 15-minute signals, occasionally diverging in the rare edge case. Holding both gives you free diversification — no extra analysis required.
The Architecture Beneath the Surface: Why the Overlay Works
Let me pull back the curtain on something I only fully understood after stress-testing the numbers.
The 15-minute strategy, on its own, has asymmetric directionality. Its buy signals are elite: 100% win rate, 2.56% average return. Its sell signals are weaker: 88.9% win rate, 1.19% average return. One sell failure.
But here’s the counterintuitive part: the overlay doesn’t just add the 15-minute strategy to v1. It filters it.
Of the 19 independent 15-minute signals, only 9 actually trigger the overlay condition. Those 9? They all win. The one sell failure never made it past the filter — it died too fast to contradict v1’s hourly position.
So the 9 signals that enter v2 carry a 100% win rate, superior to the 15-minute strategy’s standalone 94.7%. The overlay logic is doing quality control in real time.
And the increased return isn’t evenly distributed. The sell side drives more of the gain than you’d expect. v1’s sell signals were already perfect (100% win rate, 9 for 9). Adding four more overlay sell signals — all winners — expanded the sell compound return from 19.79% to 23.88%. That’s a +20.7% relative gain on the sell side, outpacing the buy side’s +17.1%.
The overlay isn’t just adding signals. It’s adding filtered, high-conviction, contrarian signals that expand the portfolio’s exposure without degrading its edge.
The Final Architecture: A Triangle
Pull everything together, and the optimal answer isn’t one asset. It’s a three-layer structure.
Core Position: TQQQ + SPXL
Both carry a 96.43% win rate, but different underlying Betas. TQQQ is the offense — it captures the Nasdaq’s higher trend momentum and volatility. SPXL is the stabilizer — its perfect sell record anchors the portfolio when counter-trend signals get noisy.
Together, they give you a Beta ladder and signal-level diversification without degrading the composite win rate.
And there’s an embedded free option: you don’t need to predict whether Nasdaq continues to lead or whether S&P 500 style rotates back into favor. Both are in the portfolio. Whichever runs, you collect.
Satellite Position: A Little SOXL
SOXL’s sell average return of 11.28% is nearly double TQQQ’s and more than double SPXL’s. Its counter-trend harvesting on the short side is in a league of its own.
The tradeoff is an 84.62% sell win rate. But “a little” is how you solve that. A position size small enough that two failures are fully absorbable, in exchange for 11 wins that outpace everything else in the portfolio.
Lose small, win big. It’s the only asymmetric bet in the structure — and it’s the source of the portfolio’s return elasticity.
Three Things That Need to Be Said About Risk
First, the backtest covers roughly six months and 28 signals. The sample is limited.
This strategy performed exceptionally from December 2025 to June 2026. But that period had its own market character — trend persistence, volatility levels, sector rotation patterns. Out-of-sample performance demands ongoing scrutiny.
Second, a 5,694% return will not fully replicate in live trading.
SOXL’s backtest result is more of a limit-case validation of Beta adaptability than a realistic target. Real-world friction — slippage, liquidity constraints during extreme moves, position sizing limits — will compress that number significantly.
Third, leveraged ETFs carry inherent structural risk.
Three-x leveraged products suffer from volatility decay and roll costs. This strategy’s high signal frequency partially mitigates those effects — holding periods are short, avoiding the accumulation of long-term drag — but it doesn’t eliminate them.
What I Actually Learned
The biggest takeaway from this process wasn’t finding a high-return portfolio. It was understanding what this strategy actually is: a volatility harvester with an embedded quality filter.
It doesn’t depend on directional predictions. It doesn’t get attached to a single asset. Its core capability is automatically adjusting its output to the Beta environment it sits in, while the overlay logic selectively imports only the highest-conviction counter-trend signals.
The 15-minute strategy’s one sell failure never entered v2. Time filtered it out. That’s not a bug — it’s the architecture working as designed.
TQQQ plus SPXL for win-rate stability. A little SOXL for return elasticity.
That triangle, as far as the current data can show, has no logical holes.
The market will write its own next chapter.
The strategy’s job isn’t to predict it. It’s to take a seat at the table no matter what’s served.
*Disclaimer: This article is for informational purposes only and does not constitute investment advice. Leveraged ETFs are high-risk products that can result in total loss of principal. Past backtested performance does not guarantee future results. Make independent decisions based on your own risk tolerance.*