I've been playing around with a few portfolio ideas on TestFol.io and ended up with five different versions that combine Fama-French Small Cap Value, leveraged ETFs, and trend following. They're all backtested from 1988-2025, and I'd love to hear what people think.
I've been playing around with a few portfolio ideas on [TestFol.io](http://TestFol.io) and ended up with five different versions that combine Fama-French Small Cap Value, leveraged ETFs, and trend following. They're all backtested from 1988-2025, and I'd love to hear what people think.
**Test:** [https://testfol.io/?s=biOKMnuGve0](https://testfol.io/?s=biOKMnuGve0)
# The idea
The goal wasn't to build the highest CAGR possible. I wanted something that could still compound aggressively while avoiding the massive drawdowns that usually come with concentrated growth portfolios.
The portfolios use four building blocks:
* **3x Nasdaq** (simulated to extend the history beyond TQQQ's inception). This is the main growth engine.
* **Small Cap Value (Fama-French factor).** You can get close with SLYV, or combine AVUV and AVDV for U.S. + international exposure. This is the part I think gets overlooked most often. Historically, it has tended to perform well during periods when growth struggles.
* **Leveraged Gold (2x) or GDE.** GDE is WisdomTree's Efficient Gold fund, which uses futures to provide roughly 90% equity and 90% gold exposure at the same time.
* **KMLM**, a managed futures/trend-following ETF that can go long or short across commodities, bonds, currencies, and equities depending on market trends.
The overall concept isn't new. It's basically factor investing and risk parity, just using modern leveraged products so the diversifiers don't reduce expected returns as much as traditional bond allocations.
# Results
**STRAT1**
* 10% 3x Nasdaq
* 20% 2x Gold
* 50% Small Cap Value
* 20% KMLM
* CAGR: **14.91%**
* Max Drawdown: **-33.26%**
* Sharpe: **0.74**
This had by far the best risk-adjusted performance.
**STRAT2**
* 10% 3x Nasdaq
* 20% GDE
* 70% Small Cap Value
* CAGR: **16.10%**
* Max Drawdown: **-58.57%**
**STRAT3**
Same allocation as STRAT2, but replacing GDE with 2x Gold.
* CAGR: **15.30%**
* Max Drawdown: **-50.42%**
Interestingly, just changing the gold exposure reduced the worst drawdown by about 8 percentage points while giving up less than 1% CAGR.
**STRAT4**
* 10% 3x Nasdaq
* 30% GDE
* 60% Small Cap Value
* CAGR: **16.23%**
* Max Drawdown: **-57.57%**
**STRAT5**
* 30% 3x Nasdaq
* 30% 2x Gold
* 40% Small Cap Value
* CAGR: **19.79%**
* Max Drawdown: **-66.58%**
Highest return, but also by far the highest risk.
# What stood out to me
The biggest surprise was STRAT1.
Giving up around five percentage points of CAGR compared to the most aggressive version nearly cut the maximum drawdown in half. That seems like a very attractive trade-off, especially over multi-decade investing horizons.
It also looks like KMLM is doing most of the heavy lifting there. Trend following tends to perform well during prolonged bear markets because it can move defensively or even short certain asset classes while equities are falling.
The STRAT2 vs STRAT3 comparison was interesting as well. Holding everything else constant, simply replacing GDE with 2x leveraged gold noticeably reduced drawdowns, which suggests the choice of gold implementation matters more than I expected.
STRAT5, on the other hand, feels much more like adding leverage than adding diversification. It produces the highest CAGR, but the drawdown profile ends up looking pretty similar to a highly leveraged equity portfolio.
# A few caveats
Obviously these results aren't perfect.
* The leveraged ETFs are simulated before their actual inception dates, so they don't fully capture real-world tracking error, leverage decay, fees, or implementation costs.
* Small Cap Value has had a long period of underperformance relative to Growth since roughly 2007, so there's no guarantee its historical premium will look the same going forward.
* Everything assumes annual rebalancing with no taxes, which isn't realistic for taxable accounts.
Still, I found the trade-offs pretty interesting.
Curious to hear if anyone has built something similar or sees any obvious flaws in the approach.
