Full Leverage Spectrum: 1x/2x/3x × B&H/SMA200 across 5 Underlyings
Full Leverage Spectrum: 1x/2x/3x × B&H/SMA200 across 5 Underlyings
Question
The earlier per-underlying analysis covered 1x and 3x only. This study fills in 2x AND adds gold + silver (the metals in a sample heavy-leveraged portfolio) to map the full leverage spectrum. The questions:
- Where's the break-point at which leverage starts hurting B&H returns?
- Does the SMA200 filter help or hurt at each leverage level?
- For the 2x ETFs specifically (USD, QLD, SSO, UGL, AGQ), is there enough vol decay to matter?
- What's the optimal leverage × strategy combo per underlying?
Methodology
- Window: 2000-08-30 → 2026-05-18 (25.7 years, common-start for SPY/QQQ/^SOX/gold/silver)
- Synthetic LETFs:
(1 + leverage × daily_return − expense_ratio/252).cumprod(). ER = 0.91% for 2x and 3x; 0% for 1x. - SMA200 strategy: Long when underlying > 200-day SMA, cash otherwise. Position lagged 1 bar.
- Cost: 1bp per side on position changes.
- Underlyings: SPY (SSO=2x, UPRO=3x), QQQ (QLD=2x, TQQQ=3x), ^SOX (USD=2x, SOXL=3x), GC=F gold (UGL=2x, no 3x ETF), SI=F silver (AGQ=2x, no 3x ETF)
- Starting capital: $10,000
Results
Full table (sorted by underlying, then leverage, then strategy)
| Underlying | Lev | Strategy | CAGR | MDD | Vol | Sharpe | $10k → |
|---|---|---|---|---|---|---|---|
| SPY | 1x | B&H | 8.33% | -55.19% | 19.19% | 0.513 | $77,824 |
| SPY | 1x | SMA200 | 7.42% | -20.85% | 11.02% | 0.704 | $62,698 |
| SPY | 2x | B&H | 12.05% | -83.88% | 38.38% | 0.489 | $185,441 |
| SPY | 2x | SMA200 | 13.29% | -38.56% | 22.04% | 0.677 | $245,790 |
| SPY | 3x | B&H | 12.65% | -96.38% | 57.57% | 0.497 | $212,320 |
| SPY | 3x | SMA200 | 18.80% | -52.92% | 33.07% | 0.688 | $830,761 |
| QQQ | 1x | B&H | 8.68% | -80.45% | 25.59% | 0.453 | $84,631 |
| QQQ | 1x | SMA200 | 9.23% | -32.21% | 14.89% | 0.668 | $96,332 |
| QQQ | 2x | B&H | 9.62% | -97.85% | 51.18% | 0.435 | $105,444 |
| QQQ | 2x | SMA200 | 15.97% | -54.91% | 29.79% | 0.648 | $447,825 |
| QQQ | 3x | B&H | 4.39% | -99.91% | 76.77% | 0.441 | $30,083 |
| QQQ | 3x | SMA200 | 21.16% | -70.58% | 44.68% | 0.656 | $1,377,225 |
| ^SOX | 1x | B&H | 9.44% | -85.14% | 36.27% | 0.429 | $101,117 |
| ^SOX | 1x | SMA200 | 8.48% | -46.04% | 21.56% | 0.486 | $80,808 |
| ^SOX | 2x | B&H | 4.10% | -99.60% | 72.54% | 0.417 | $28,033 |
| ^SOX | 2x | SMA200 | 11.73% | -75.56% | 43.12% | 0.474 | $172,011 |
| ^SOX | 3x | B&H | -12.54% | -100.00% | 108.81% | 0.421 | $321 |
| ^SOX | 3x | SMA200 | 10.35% | -92.00% | 64.68% | 0.479 | $125,266 |
| GOLD | 1x | B&H | 11.62% | -44.36% | 17.79% | 0.707 | $166,878 |
| GOLD | 1x | SMA200 | 8.00% | -39.36% | 14.96% | 0.589 | $71,657 |
| GOLD | 2x | B&H | 19.58% | -74.39% | 35.58% | 0.682 | $972,754 |
| GOLD | 2x | SMA200 | 13.37% | -65.30% | 29.92% | 0.570 | $248,617 |
| GOLD | 3x | B&H | 25.13% | -89.51% | 53.37% | 0.690 | $3,113,159 |
| GOLD | 3x | SMA200 | 17.04% | -80.94% | 44.88% | 0.579 | $561,952 |
| SILV | 1x | B&H | 11.38% | -75.85% | 33.00% | 0.494 | $158,073 |
| SILV | 1x | SMA200 | 5.69% | -70.57% | 26.68% | 0.344 | $41,263 |
| SILV | 2x | B&H | 9.34% | -97.54% | 66.00% | 0.481 | $98,410 |
| SILV | 2x | SMA200 | 2.78% | -94.12% | 53.37% | 0.336 | $20,192 |
| SILV | 3x | B&H | -7.86% | -99.88% | 98.99% | 0.485 | $1,229 |
| SILV | 3x | SMA200 | -11.07% | -99.50% | 80.05% | 0.340 | $495 |
Interpretation
Finding 1: The leverage break-point depends on underlying volatility
| Underlying | 1x vol | Where does B&H leverage start to hurt? |
|---|---|---|
| GOLD | 17.8% | Never in this window. 3x B&H wins. |
| SPY | 19.2% | Marginal at 3x (slight degradation vs 2x) |
| QQQ | 25.6% | At 3x, B&H goes negative (vol decay > underlying return) |
| SILV | 33.0% | At 2x, B&H underperforms 1x. At 3x, near-wipeout. |
| ^SOX | 36.3% | At 2x, B&H halves the 1x return. At 3x, total wipeout. |
Rule of thumb: B&H of a 3x daily-rebalance ETF is structurally negative-EV when the underlying's annualized vol exceeds roughly 25-30%. Below that, leverage compounds. Above that, vol decay compounds against you.
This is purely mathematical (vol decay scales with vol squared) and does not depend on regime. It's why SOXL is structurally a bad B&H choice over 25+ year windows, regardless of how hard the semiconductor sector runs.
Finding 2: SMA200 filter helps equities, HURTS metals
| Underlying | 1x B&H Sharpe | 1x SMA200 Sharpe | Filter effect |
|---|---|---|---|
| SPY | 0.513 | 0.704 | +0.191 (helps) |
| QQQ | 0.453 | 0.668 | +0.215 (helps) |
| ^SOX | 0.429 | 0.486 | +0.057 (helps marginally) |
| GOLD | 0.707 | 0.589 | −0.118 (HURTS) |
| SILV | 0.494 | 0.344 | −0.150 (HURTS dramatically) |
The filter helps on equity underlyings (predictable bear cycles to avoid) and hurts on metals (choppy ranging periods generate too many false exit signals).
Practical implication: the canonical SMA200 trend-following framing should be applied to equity-backed positions, not gold/silver. Holding gold below its SMA200 historically still worked because gold's bull/bear cycles are slower and the filter generates costly whipsaw.
Finding 3: Where leverage + SMA200 produces the money
Top 5 $10k → final value over 25.7 years:
- GOLD 3x B&H: $3,113,159 (25.1% CAGR, but -89.5% MDD)
- QQQ 3x SMA200: $1,377,225 (21.2% CAGR, -70.6% MDD)
- GOLD 2x B&H: $972,754
- SPY 3x SMA200: $830,761
- GOLD 3x SMA200: $561,952
The headline result is uncomfortable: unfiltered 3x gold dominates the dataset. The 25-year window includes both major gold bull markets (2001-2011 and 2019-present) and the filter underperforms by generating false exits.
But — that -89.5% max drawdown on 3x gold B&H means in practice almost nobody would have actually held through it. The "real return" most humans would have captured is much lower because they'd have sold at the bottom.
Finding 4: SMA200 + 2x is the sweet spot for ^SOX
Best ^SOX strategy by absolute dollars: SMA200 + 2x ($172,011) beats SMA200 + 3x ($125,266) and B&H 1x ($101,117).
For semiconductors specifically, 2x with the filter is structurally superior to 3x with the filter over long windows. The extra leverage from 2x → 3x is eaten by vol decay even with the filter on. If you want leveraged semi exposure with a trend filter, USD (2x) > SOXL (3x).
Finding 5: Silver is a bad leveraged play at any level
| Strategy | $10k → |
|---|---|
| SILV 1x B&H | $158,073 |
| SILV 2x B&H | $98,410 |
| SILV 2x SMA200 | $20,192 |
| SILV 3x B&H | $1,229 |
| SILV 3x SMA200 | $495 |
Every leveraged silver variant loses to 1x B&H. And the SMA200 filter actively destroys returns at every leverage level on silver. Silver's combination of high vol (33% at 1x), choppy non-trending behavior, and false-signal-prone price action make it the worst possible asset for trend-following leveraged ETF strategies.
Direct implications for a sample heavy-leveraged portfolio
The portfolio holds 25% in heavily-traded semis (20% USD + 5% SOXL) and 15% AGQ + UGL + SLVP combined. The findings:
| Holding | Verdict |
|---|---|
| UGL (2x gold) at 10% | Strong pick. Gold 2x B&H is 19.6% CAGR with Sharpe 0.68. The portfolio's best asset. |
| TQQQ (3x QQQ) at 15% | Strong if held with discipline. With SMA200: 21.2% CAGR. Without (the portfolio's B&H stance): 4.4% CAGR over 25 years. |
| USD (2x semis) at 20% | Questionable. B&H of 2x ^SOX returned only 4.1% CAGR over 25 years. Without trend filtering, semi 2x exposure is mediocre. |
| SOXL (3x semis) at 5% | Bad over long windows. B&H 3x ^SOX synthetic ended at $321 from $10k. Even small allocations are structurally negative-EV. |
| AGQ (2x silver) at 10% | Bad over long windows. Silver 2x B&H returned 9.3% CAGR over 25 years, less than 1x silver. Worst risk/reward in the portfolio. |
| UPRO (3x SPY) at 5%, QLD (2x QQQ) at 5%, SLVP (silver miners) at 5%, ZROZ (long bonds) at 25% | UPRO and QLD: reasonable. SLVP: questionable (silver-linked). ZROZ: see separate defensive bucket study — actively bad. |
This is one of the most uncomfortable findings on the entire research sheet. A meaningful chunk of the portfolio's holdings (USD, SOXL, AGQ, ZROZ ≈ 60% combined) are structurally weak positions over multi-decade windows. The 2012-2026 bull window flattered them; the 25-year stress test (Sharpe drops from 0.91 to 0.59) reflects this.
This is not a recommendation to change the portfolio. It's framework information. a sample heavy-leveraged portfolio has been deliberately diversified across leveraged assets — the metals and bonds are intended as insurance against the equity sleeve, not as standalone winners. Whether the insurance is worth the cost depends on the regime ahead.
Caveats
-
Synthetic LETFs ignore real borrow cost. Pre-2009 LETFs paid borrow rates of 4-15% on the leveraged portion. The synthetic numbers understate real costs. The directional conclusions hold; absolute numbers are upper bounds.
-
The 25-year window includes 2 commodity bull markets. Gold and silver had massive runs 2001-2011 and gold had another from 2019. A window without these (e.g., 1980s commodity bear) would show metals B&H as much worse.
-
^SOX is the unleveraged semi index, not SOXX (the actual ETF). Tracking error in real LETFs would shave another fraction of a percent.
-
The 3x daily-rebalance assumption. Real LETFs reset at market close; our synthetic assumes the same. Intraday tracking error and rebalance slippage in real LETFs is small but non-zero.
-
No tax modeling. All figures pre-tax. SMA200 strategies generate 2-3 round-trip trades per year of short-term capital gains. After-tax results in taxable accounts would shift further toward B&H variants.
Source
Saved log: /tmp/leverage_spectrum.log.
Inline runner:
import pandas as pd, numpy as np
import sys; sys.path.insert(0, '/Volumes/Mac External/Claudes/trader/src')
from trader.data.yfinance_src import fetch_daily
TD = 252; COST = 1.0/10000.0
ER = {1: 0.0, 2: 0.0091, 3: 0.0091}
underlyings = {
"SPY": fetch_daily("SPY", start="1999-01-01")["close"],
"QQQ": fetch_daily("QQQ", start="1999-01-01")["close"],
"^SOX": fetch_daily("^SOX", start="1999-01-01")["close"],
"GOLD": fetch_daily("GC=F", start="2000-08-30")["close"],
"SILV": fetch_daily("SI=F", start="2000-08-30")["close"],
}
start = max(u.index[0] for u in underlyings.values())
end = min(u.index[-1] for u in underlyings.values())
def synthesize(u, lev):
return (1 + lev*u.pct_change().fillna(0) - ER[lev]/TD).cumprod()
# For each underlying × leverage × {B&H, SMA200}: synthesize, run strategy, compute metrics
Related studies
- Leverage vs volatility per underlying — the earlier 1x vs 3x analysis this extends
- a sample heavy-leveraged portfolio (internal-only study) — the portfolio this informs
- Defensive bucket comparison — companion analysis on defensive bucket choice
This is research output, not investment advice. Backtest results do not predict future returns. Specific portfolio compositions discussed here are illustrative test cases, not allocation recommendations. Do your own research and consult a licensed advisor for personalized advice. Full disclaimer →