CLAUSE · BACKTEST REPORT · HYPOTHETICAL PERFORMANCE

GARCH Volatility Regime Switch: TQQQ/GLD

“fit a garch model to QQQ; hold TQQQ while it predicts volatility below 20 percent, otherwise hold GLD”

Tested over April 2011 to September 2026. On data sealed Sep 6, 2026. Published 6 September 2026.

The contract

1. Universe. Trade only QQQ, TQQQ, GLD.
2. Signal. Each trading day at 09:35 ET, fit a GARCH(1,1) model to QQQ’s returns on a rolling 252-day window (refit every 21 days), predicting next-day volatility from past squared returns at lag 1; its own innovations at lag 1.
3. Entry. If the GARCH(1,1) volatility forecast falls below 20, set the target allocation of TQQQ to 100% and set the target allocation of GLD to 0%.
4. Entry. Otherwise, set the target allocation of GLD to 100% and set the target allocation of TQQQ to 0%.

Equity vs SPY buy-and-hold

STRATEGYSPY BUY-AND-HOLDlog scale

Tear sheet

TOTAL

+2,153%

CAGR

+22.5%

VOL

40.5%

SHARPE

0.67

SPY SHARPE

0.67

BUY-AND-HOLD SHARPE

–

SORTINO

0.94

MAX DD

−58.9%

LONGEST DD

–

BETA

0.38

GROSS

–

GROSS MAX

–

NET

–

LONG

–

SHORT

–

TRADES

583

TURNOVER

–

COST DRAG

–

CAPACITY

–

FINAL VALUE

$2,252,947 from $100,000

MARGIN NOTES

most of the return came from 2023-05-04 to 2026-06-03 — one stretch carries the result

Per-clause attribution

Each number re-runs the backtest with that clause removed and shows what it contributed, so they do not add up to the total.

Clause 3

+1,978%

Clause 4

−9,077%

About these numbers

Window

Tested from Apr 27, 2011 to Sep 4, 2026. On a copy of the market sealed Sep 6, 2026.

Costs

Every fill pays spread and slippage. This run predates the written cost model; run it again to see the assumptions key by key.

Attribution

Each clause's share of the result is measured by re-running the backtest without it.

Hypothetical performance on historical data. The same contract on the same data and cost model reproduces this number exactly. Backtests are not indicative of future results.