Velocity Strength Filter compounded at +17.7% a year from 2016-10-24 to 2026-10-05, with a worst drawdown of -27.4% (high drawdown). It ranks 2 of 6 re-run strategies by drawdown, shallowest first.
The rule in one line: The strategy rotates among eight leveraged index ETFs, such as TQQQ, UPRO and SOXL, and a Treasury bill ETF (BIL).
| Worst drop (max drawdown) | -27.4% |
|---|---|
| Yearly growth (CAGR) | +17.7% |
| Return per unit of risk (Sharpe) | 0.84 |
| Total return | +406.4% |
| S&P 500 (SPY), same period | +322.2% |
| Return ÷ drawdown | 0.65 |
| Period | 2016-10-24 – 2026-10-05 (9.9 years) |
| Risk tier | High drawdown |
| Drawdown rank | 2 / 6 |
Backtest = historical simulation on past prices, not real trading.
Equity curve of a $10,000 start, with SPY dashed.
The strategy rotates among eight leveraged index ETFs, such as TQQQ, UPRO and SOXL, and a Treasury bill ETF (BIL). It uses completed daily closes and trades at the next open. When SPY closes above its 200-day moving average it may hold leveraged funds; when SPY closes below it, everything moves to BIL.
Once a week, on the first trading day, each leveraged fund is scored by how far its price sits above its 63-day average, divided by its recent annualized volatility. Only funds with a positive three-month return and a positive score qualify. The strategy holds the top two at 30% of the portfolio each, with the rest in BIL. To limit trading, a current holding that still ranks in the top three is kept unless a challenger's score beats it by more than 10%.
Two daily safety rules apply. If a held fund closes 15% below its highest close since purchase, it is sold; the money waits in cash until the next weekly review, and that fund is excluded for five trading days. If the whole account falls 10% from its peak, everything moves to BIL for at least 21 trading days. It returns only when SPY is above its 200-day average and has either climbed back to its level at the time of the exit or spent time below that average during the pause.
| Universe | Growth: TQQQ, UPRO, SOXL, FNGU, BULZ, TECL, TNA, SPXL; cash: BIL; market signal and benchmark: SPY. VIX data is loaded but not used for decisions. |
|---|---|
| Rebalancing | Weekly, on the first trading day of each week. The SPY trend exit, trailing stops and account breaker are checked every day. |
| Execution | Signals use completed daily closes only; orders fill at the next session's open and positions are marked at the close. Holdings drift with prices between trades. |
| Trading costs | 0.25% per side on every dollar bought or sold (about 0.50% for a full switch). No separate slippage or spread model. |
| Leverage | Leveraged ETFs fixed at 30% each, at most 60% in total; the rest in BIL. No margin. |
| Risk controls | SPY below its 200-day average: 100% BIL. Fund 15% below its post-entry high: sold, excluded five days. Account down 10%: BIL for at least 21 days, re-entry needs SPY recovery. |
| Price data | Public daily price data adjusted for dividends and splits. BIL earns its adjusted return; proceeds of a stopped fund sit in cash at 0% until the next trade. |
| Known limitations | Days without VIX data are dropped even though VIX no longer drives decisions. Hand-picked surviving ETFs; parameters chosen on the same 10-year window. |
Across the 9 full calendar years in the test, the strategy beat SPY in 4. Its best year was 2023 (+62.4% against SPY's +26.2%) and its weakest was 2018 (-14.1% against -4.6%). It finished 3 full years with a loss (2018, 2022, 2024). In 2018, when SPY fell 4.6%, the strategy returned -14.1%. In 2022, when SPY fell 18.2%, the strategy returned -5.9%.
SPY fell more than 10% from a high 3 times in this window (2018-09-20 to 2018-12-24: SPY -19.4%, strategy -9.4%; 2020-02-19 to 2020-03-23: SPY -33.7%, strategy -10.8%; 2022-01-03 to 2022-10-12: SPY -24.5%, strategy -4.0%). The strategy lost less than the index in 3 of 3 of those declines, so its rules did act as a brake when the whole market sold off.
Labelled by SPY's trend (above a rising 200-day average = uptrend, below a falling one = downtrend, anything else = sideways), the strategy averaged an annualized +21.0% on uptrend days (79% of the test) against SPY's +23.4%, so it trailed the index in steady rising markets. On sideways days (8%), which is where trend breaks and the first leg of sell-offs land, it averaged +23.5% against SPY's -77.2%; on downtrend days (10%), which include the sharp rebounds inside bear markets, -4.3% against +27.1%. Its losses were most concentrated in uptrend stretches: 93% of all losing-day losses came on 79% of the days.
On days SPY rose, the strategy captured 98% of the index's gains; on days SPY fell, it took 94% of the index's losses, a roughly symmetric profile, so its edge did not come from sidestepping down days. 64% of months ended positive; the best month was 2026-05 (+32.3%) and the worst was 2018-02 (-13.9%).
The deepest drawdown on the daily curve ran from a peak on 2025-10-29 to a trough on 2026-03-03 (about 4 months of decline), a fall of -27.4%. Over the same stretch SPY returned -0.7%, so this loss was specific to the strategy rather than a market-wide sell-off. It regained the previous high on 2026-05-05, about 6 months after the peak.
| Year | Strategy | SPY | Difference |
|---|---|---|---|
| 2016 (partial) | +0.7% | +4.6% | -3.9% |
| 2017 | +37.5% | +21.7% | +15.8% |
| 2018 | -14.1% | -4.6% | -9.5% |
| 2019 | +13.8% | +31.2% | -17.4% |
| 2020 | +40.7% | +18.3% | +22.4% |
| 2021 | +21.3% | +28.7% | -7.4% |
| 2022 | -5.9% | -18.2% | +12.3% |
| 2023 | +62.4% | +26.2% | +36.2% |
| 2024 | -5.3% | +24.9% | -30.2% |
| 2025 | +6.9% | +17.7% | -10.8% |
| 2026 (partial) | +39.9% | +14.5% | +25.4% |
| Peak | Trough | Recovered | Fall | SPY, peak to trough |
|---|---|---|---|---|
| 2025-10-29 | 2026-03-03 | 2026-05-05 | -27.4% | -0.7% |
| 2018-01-26 | 2019-10-08 | 2020-02-06 | -27.1% | +4.0% |
| 2024-03-07 | 2025-02-28 | 2025-10-27 | -26.1% | +16.9% |
| SPY trend | Share of days | Strategy, annualized average | SPY, annualized average |
|---|---|---|---|
| Uptrend | 79% | +21.0% | +23.4% |
| Sideways | 8% | +23.5% | -77.2% |
| Downtrend | 10% | -4.3% | +27.1% |
Computed from the daily equity curve of the same re-run (2016-10-24 to 2026-10-05). Trend labels classify each day after the fact (SPY above a rising 200-day average = uptrend, below a falling one = downtrend, otherwise sideways); they describe the past and are not a trading signal. Annualized averages are the mean daily return in that group times 252.
The worst peak-to-trough fall was -27.4%: a $10,000 account would have dropped to about $7,255 at that point before recovering. The +17.7% CAGR is the annual rate that compounds the start value into the end value over 9.9 years; the return-to-drawdown ratio of 0.65 says how many points of annual return each point of worst-case loss bought. Compare strategies on that pair, not on CAGR alone.
Re-run on 2026-10-07 with the strategy's current code on historical prices — a systematic take on The Velocity Strength Filter. Numbers change when the strategy is re-run on newer data. The full write-up explains the rules.
All strategy backtests, lowest drawdown first
Disclosure. Backtests are hypothetical simulations on historical data. They do not include every real-world cost, are not live results, and do not guarantee future returns. Educational research only, not investment advice.