DeployQuantSDS trading strategies, backtested
ProShares UltraShort S&P500: -2x daily S&P 500. Every DeployQuant template run on SDS over 5.7 years of minute data, same engine, same window, sorted by return.
SDS is ProShares UltraShort S&P 500, a fund that targets minus 2x the daily return of the S&P 500. From 2021-01-04 to 2026-10-02 it lost 21.57% a year in the test, and $10,000 became $2,479. It was down in five of the six calendar years. The one up year was 2022, when it gained 30.5%. The drawdown started on the first day of the data and its low came on 2026-08-13, 76.51% below the start, with no recovery by the end of the window.
Twelve templates were run on it. All twelve beat buy-and-hold on CAGR and all twelve had a shallower maximum drawdown. None of them made money. The best was golden cross at minus 2.33% a year, followed by RSI(2) snapback at minus 3.58% and EMA 12/26 trend at minus 6.89%. The worst was monthly cycle at minus 20.54%.
That makes SDS an unusual page. The ranking measures which rule lost least on a fund that fell almost continuously, and all twelve templates were ahead of the fund because each sat in cash for part of the window. The window is one stretch of 5.7 years in which the S&P 500 rose, so it says little about how an inverse fund behaves in a sustained bear market.
| Strategy | CAGR | max DD | Sharpe | trades | win rate | final value |
|---|---|---|---|---|---|---|
| Golden Cross (SMA 50/200) | −2.3% | −33.4% | -0.04 | 2 | 50% | $8,733 |
| RSI(2) Dip Snapback | −3.6% | −47.9% | -0.10 | 169 | 53% | $8,111 |
| EMA 12/26 Trend | −6.9% | −39.4% | -0.19 | 20 | 25% | $6,637 |
| 200-Day SMA Regime Filter | −7.5% | −44.4% | -0.26 | 17 | 0% | $6,381 |
| 20-Day Momentum + Trailing Stop | −9.9% | −48.6% | -0.48 | 23 | 30% | $5,478 |
| SMA-200 Trend + 15% Trailing Stop | −10.8% | −56.4% | -0.39 | 15 | 27% | $5,178 |
| RSI(14) Mean Reversion | −11.0% | −56.8% | -0.47 | 14 | 43% (+1 open) | $5,110 |
| SMA 10/50 Trend | −11.6% | −58.0% | -0.42 | 14 | 21% | $4,935 |
| 3-Month Momentum Switch | −11.6% | −55.4% | -0.46 | 13 | 8% | $4,923 |
| Weekly Entry + 7% Target | −15.5% | −76.2% | -0.49 | 199 | 22% (+1 open) | $3,794 |
| Drawdown Dip Buyer + 8% Target | −19.9% | −75.3% | -0.56 | 1 | 100% (+1 open) | $2,805 |
| First-to-Last Day of Month | −20.5% | −74.3% | -0.60 | 69 | 32% (+1 open) | $2,671 |
| Buy & hold SDS | −21.6% | −76.5% | -0.68 | – | – | $2,479 |
Which rules lost least
Holding SDS lost 42.6% in 2021, gained 28.2% in 2022, then lost 29.7%, 27%, 23.7% and 16.6% in 2023, 2024, 2025 and the part of 2026 in the data. The Sharpe ratio was minus 0.68. Against that, every template cut the loss, and the cut came almost entirely from being absent in the years the fund fell.
Golden cross made 2 round trips. The first ran from 2022-04-25 to 2023-01-04, held 254 days, and gained 8.3%. The second ran from 2025-04-10 to 2025-06-23 and lost 19.62%. Its exposure was 15.5%, and its profit factor was 0.39. In 2021, 2024 and 2026 it did nothing and so finished flat in those years, which is where its lead over the fund comes from. The 2025 trade shows the risk of a slow signal on this fund: it bought on 2025-04-10, the day after SDS fell 18.63% in one session, and lost on the trade. The drawdown of 33.37% ran from 2022-09-30 to 2025-06-12 and had not recovered by the end.
RSI(2) snapback made 169 round trips, won 53% of them and had a profit factor of 0.92. It returned minus 3.58% with a 47.91% drawdown and was invested 47.2% of the time. The average win was 2.35% and the average loss was 2.82%. Its edge was in 2022, when it gained 49.3% against 28.2% for the fund. In the other five years it lost 15.4%, 8.4%, 14.9%, 4.1% and 14.2%. The template beat holding in every one of the six years. A rule that buys short-term weakness in a fund that fell for years keeps buying into a downtrend, and a win rate just above 50% did not offset the drift.
EMA 12/26 trend lost 6.89% with 20 round trips and a 25% win rate. It was invested 27.5% of the time. It made money in 3 of 4 round trips in 2022 and in 1 of 6 in 2023. Its best trade ran from 2025-02-27 to 2025-04-30 and gained 13.06%, while the worst, from 2022-06-13 to 2022-07-21, lost 8.93%. The profit factor was 0.34. In 2022 it lost 2.2% while the fund gained 28.2%, so it was 30.4 points behind in the one year when the fund worked.
The 200-day regime filter had 17 round trips and 0 wins. Its average loss was 2.67% and it lost 20.4% in 2022. Because every trade lost a small amount, the template ended at minus 7.53% a year with a profit factor of 0. The 3-month momentum switch won 1 of 13 round trips for an 8% win rate, with its best trade a gain of 4.6% in March 2022. Trend plus trailing stop won 4 of 15. SMA 10/50 won 3 of 14. Momentum breakout won 7 of 23.
Weekly 7% target made 199 round trips with a 22% win rate and a profit factor of 0.81. Its average win was 7.44% and its average loss was 2.55%. It gained 60.6% in 2022 and lost between 17.2% and 36.2% in each of the other five years. Its exposure was 71.7%, so it stayed close to the fund and its drawdown of 76.18% was about the same as holding at 76.51%. Its worst trade was a loss of 15.77% from 2025-04-07 to 2025-04-10, in the April 2025 reversal.
The drawdown dip buyer made 2 trades. The first gained 8% from 2021-02-09 to 2021-03-04. The second, entered on 2021-03-18 at an adjusted price of 223.21, was still open at the end with a loss of 75.99%. Its exposure was 97.6%, so it behaved like holding after March 2021, and its CAGR of minus 19.86% is close to the fund's. The win rate of 100% counts only the closed trade.
RSI(14) mean reversion lost 11.04% with a 43% win rate and an average loss of 12.68% against an average win of 7.39%. Its worst trade held 203 days from 2025-04-30 to 2025-11-19 and lost 29.42%.
The other inverse funds tell a similar story with different winners. SQQQ held at minus 42.33% and its best template, golden cross, made minus 7.41%. QID held at minus 29.46% and RSI(2) made minus 1.85%. PSQ held at minus 13.76% and RSI(2) made 0.13%, the only positive result of those four. SH held at minus 10% and RSI(2) made minus 0.53%. On SOXS holding lost 48.27% a year and trend plus trailing stop made minus 4.93%, and on EEV RSI mean reversion made 5%. TBF, the short Treasury fund, held at a gain of 12.21%, so a template trailed there.
Months, streaks and the yearly gaps
September 2022 was SDS's best month for holding at 19.06%, and it was the best month for the weekly target at 23.76%, the SMA 10/50 trend at 24.19%, the EMA 12/26 trend at 22.53%, the golden cross at 19.72%, the trailing stop strategy at 19.68%, the dip buyer at 19.61% and the momentum breakout at 11.16%. July 2022 was the worst month for holding at negative 15.9% and for the golden cross at negative 16.4%, the 200-day filter at negative 16.47%, the trailing stop strategy at negative 16.06% and the RSI(14) rule at negative 12.58%. April 2026 was the worst month for the weekly target at negative 16.07%, the monthly cycle at negative 15.18%, the dip buyer at negative 16.41% and the RSI(2) snapback at negative 16.04%, and the RSI(2) snapback's worst trade, a loss of 16.6% from 2026-04-02 to 2026-04-22, is in that month.
The streaks show how the rules lose. The 200-day filter lost 17 round trips in a row and won none. The weekly target lost 19 in a row at its longest and won at most 4. The 3-month momentum switch lost 10 in a row, the monthly cycle 8, and the SMA 10/50 trend, the momentum breakout and the RSI(2) snapback 6 each. The RSI(2) snapback also won 6 in a row at its longest.
The yearly gaps to holding split by rule type. In 2022, the only year SDS gained, the 3-month momentum switch trailed holding by 60.4 points, the 200-day filter by 48.6, the momentum breakout by 46.3, the trailing stop strategy by 43.5 and the EMA 12/26 trend by 30.4. The weekly target led by 32.4 points that year and the monthly cycle by 10.6. In 2021 every template beat holding, from 1.9 points for the monthly cycle to 42.6 for the golden cross, the 200-day filter, the 3-month switch and the trailing stop strategy, which stayed in cash that year. The RSI(2) snapback beat holding in every year, from 27.2 points in 2021 to 2.4 in 2026.
No template recovered from its worst drawdown in the window. The shallowest, the golden cross at 33.37% from 2022-09-30, and the deepest, the weekly target at 76.18% from 2022-11-09, both had no recovery date, and holding's own 76.51% drawdown from 2021-01-04 was still open on the last day. The RSI(14) rule's open trade, entered on 2026-08-13 at 52.61, was up 1.86%, and the monthly cycle entered on 2026-10-01 at 54.31 and was down 1.33%. The weekly target entered on 2026-09-28 at 53.76 and was down 0.32%.
How each strategy traded SDS
| Strategy | Time in market | Avg hold (days) | Best trade | Worst trade | Profit factor | With 10 bps slippage |
|---|---|---|---|---|---|---|
| golden cross | 15.5% | 164 | 8.3% | −19.6% | 0.39 | −2.4% |
| RSI(2) snapback | 47.2% | 6 | 11.1% | −16.6% | 0.92 | −9.0% |
| EMA 12/26 trend | 27.5% | 29 | 13.1% | −8.9% | 0.34 | −7.5% |
| 200-day regime filter | 15.6% | 19 | −0.2% | −6.2% | – | −8.0% |
| momentum breakout | 15.2% | 14 | 23.3% | −10.6% | 0.39 | −10.6% |
| trend + trailing stop | 21.1% | 29 | 16.5% | −14.4% | 0.40 | −11.3% |
| RSI mean reversion | 66.9% | 97 | 13.6% | −29.4% | 0.37 | −11.5% |
| SMA 10/50 trend | 28.0% | 42 | 7.0% | −16.0% | 0.20 | −12.0% |
| 3-month momentum | 19.2% | 31 | 4.6% | −16.4% | 0.08 | −12.0% |
| weekly 7% target | 71.7% | 6 | 13.2% | −15.8% | 0.81 | −21.4% |
| dip buyer | 97.6% | 23 | 8.0% | 8.0% | – | −19.9% |
| monthly cycle | 95.2% | 28 | 18.4% | −15.6% | 0.58 | −22.4% |
Trades, exposure and costs
Exposure separated the templates: monthly cycle 95.2%, the dip buyer 97.6%, weekly 7% target 71.7%, RSI(14) 66.9%, RSI(2) 47.2%, SMA 10/50 28%, EMA 12/26 27.5%, trend plus trailing stop 21.1%, 3-month momentum 19.2%, the 200-day filter 15.6%, golden cross 15.5% and momentum breakout 15.2%. The templates with the lowest exposure lost the least because there was less of the fund to lose. The 200-day filter and golden cross were each in the fund for about one session in six.
Median holding periods ran from 3 days for weekly 7% target and 4 for RSI(2) to 254 days for the one winning golden cross trade. The longest RSI(14) hold was 238 days, from 2021-01-26 to 2021-09-21, and it lost 25.93%.
Costs hurt the busy templates most. RSI(2) snapback went from minus 3.58% to minus 6.41% at 5 basis points per trade and minus 8.99% at 10, and its drawdown rose from 47.91% to 56.07%. Weekly 7% target went from minus 15.53% to minus 17.72% and minus 21.42%, which left $2,506 at 10 basis points against $2,479 for holding, nearly the same. Monthly cycle went to minus 21.64% and minus 22.4%. The slow templates were nearly untouched: golden cross went from minus 2.33% to minus 2.37% and the dip buyer from minus 19.86% to minus 19.91%.
The average daily dollar volume was $226,027,637 and the median minute volume was 2,455 shares. The average overnight gap was 0.87% and the average intraday range was 2.29%. The 10 basis point cost runs are the better guide for templates that trade every few days in a fund with a median minute volume this low.
The prices in the trade lists are adjusted for splits and dividends. The first entry was at 251.51 in January 2021 and an entry in September 2026 was at 53.76, which reflects a fund that lost most of its value.
How SDS behaved
| Measure | SDS |
|---|---|
| Data in this test | 2021-01-04 to 2026-10-02 (1444 sessions) |
| Total return, buy and hold | −79.4% |
| Annualized volatility | 32.8% |
| Deepest drawdown | −79.9% (2021-01-04 to 2026-08-13) |
| Up days | 45.7% |
| Average daily range | 2.29% |
| Average overnight gap | 0.87% |
| Correlation to SPY | -1.00 |
| Correlation to QQQ | -0.94 |
| Correlation to TLT | -0.08 |
| Sessions above the 200-day average | 18.1% |
| Crossings of the 200-day average | 34 |
| Falls of 10% or more from a 20-day high | 44 |
Calendar years
| Year | Return |
|---|---|
| 2021 | −44.5% |
| 2022 | 30.5% |
| 2023 | −31.5% |
| 2024 | −29.5% |
| 2025 | −26.8% |
| 2026 (part) | −19.6% |
Biggest single days
| Best day | Move |
|---|---|
| 2025-04-04 | 12.1% |
| 2025-04-03 | 9.5% |
| 2022-09-13 | 8.7% |
| 2022-05-18 | 8.0% |
| 2022-06-13 | 7.6% |
| Worst day | Move |
|---|---|
| 2025-04-09 | −18.6% |
| 2022-11-10 | −10.9% |
| 2025-05-12 | −6.6% |
| 2022-06-24 | −6.3% |
| 2022-10-04 | −6.2% |
Average return by calendar month
| Jan | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep | Oct | Nov | Dec |
|---|---|---|---|---|---|---|---|---|---|---|---|
| −2.0% | 0.2% | −1.1% | −0.9% | −5.6% | −1.9% | −5.3% | −0.9% | 5.5% | −5.0% | −7.2% | 0.4% |
Most and least correlated funds
| Most correlated | Least correlated | ||
|---|---|---|---|
| SH | 1.00 | SSO | -1.00 |
| SPDN | 1.00 | SPY | -1.00 |
| SQQQ | 0.94 | VOO | -1.00 |
| QID | 0.94 | VV | -1.00 |
| PSQ | 0.94 | SPUU | -0.99 |
Leverage and decay against SPY
| Year | SDS | SPY | -2× SPY, reset daily |
|---|---|---|---|
| 2021 | −44.5% | 30.5% | −44.1% |
| 2022 | 30.5% | −18.2% | 25.5% |
| 2023 | −31.5% | 26.2% | −40.3% |
| 2024 | −29.5% | 24.9% | −38.8% |
| 2025 | −26.8% | 17.7% | −35.0% |
| 2026 | −19.6% | 13.8% | −25.7% |
The last column compounds -2× SPY's daily return with no fees or financing: the return a perfect daily-reset fund would have had. It leaves out the fund's fees, its borrowing costs and the interest it earns on cash, which put a real fund below or above this line. A daily-reset fund does not aim for -2× the underlying's return over a year.
Over 2021-01-04 to 2026-10-02, SDS returned −79.4% while SPY returned 125.3% and a perfect daily-reset -2× version would have returned −87.6%. Its measured daily beta to SPY was -2.00.
What an inverse fund does over time
SDS's annualized volatility was 32.8%. Its beta to SPY was minus 2 and the correlation to SPY was minus 1, so on a day-to-day basis it behaved as designed. The average up day was 1.52% and the average down day was minus 1.48%, but only 45.67% of days were up. Up and down days were about the same size, so the loss came from the count of down days and from compounding while the S&P 500 rose.
The leverage table sets SDS beside SPY and beside a frictionless daily-reset −2× version of SPY. SPY returned 125.29% over the window. The daily-reset ideal would have returned minus 87.63%, and SDS returned minus 79.38%, a smaller loss. By year, in 2021 SPY gained 30.46% and SDS lost 44.46%, against minus 44.12% for the ideal. In 2023 SPY gained 26.18% and SDS lost 31.53%, against minus 40.33%. In 2022 SPY lost 18.16% and SDS gained 30.5%, against 25.5% for the ideal. The ideal leaves out fees, borrowing costs and interest on cash, so the table describes the window and does not say what moved SDS above the line.
The fund's own record shows how one-sided the window was. Its total return was minus 79.38% and its deepest drawdown was 79.94%, running from 2021-01-04 to 2026-08-13 with no recovery. The longest drawdown lasted 1,443 sessions, which is nearly every session in the data of 1,444. The calendar years were minus 44.46% in 2021, 30.5% in 2022, minus 31.53% in 2023, minus 29.49% in 2024, minus 26.76% in 2025 and minus 19.55% in 2026 so far. A single up year inside a long slide is why the templates that were out of the fund in 2021, 2023, 2024 and 2025 look better than the ones that stayed in.
Overnight returns were minus 62.82% in log terms and intraday returns were minus 91.64%, so about 40.67% of the loss came overnight and 59.33% came inside the session.
The best day was 2025-04-04 at 12.05% and the worst was 2025-04-09 at minus 18.63%. These two fall within a week of each other. The other good days, 2022-09-13 at 8.71% and 2022-05-18 at 8.03%, are in 2022. The fund fell 10% or more from a 20-day high 44 times across 295 days. RSI(14) fell under 30 on 81 sessions, with a median forward 5-day return of 0.16% against minus 0.72% for the baseline, and a 20-day return of minus 1.04% against minus 2.97%. RSI(2) fell under 10 on 252 sessions, with forward 5-day returns of minus 0.68% and 20-day returns of minus 3.07%. The baseline for this fund is negative. RSI(14) readings beat it, RSI(2) readings did not, and both left a median loss over 20 days.
The fund spent 18.07% of sessions above its 200-day average and crossed it 34 times. That fits the low exposure of the 200-day filter and golden cross. The lag-1 autocorrelation was minus 0.01.
Seasonally September averaged 5.49% and February 0.22%, while November averaged minus 7.15%, May minus 5.58%, July minus 5.28% and October minus 5%. With five or six observations per month these are a description of this window. The fund was most correlated with SH and SPDN at 1.00, and least correlated with SSO, SPY, VOO and VV at minus 1.00.
Day-of-week figures and the peer inverse funds
Monday averaged negative 0.25% and Wednesday negative 0.13%, against 0% on Thursday, negative 0.07% on Friday and negative 0.01% on Tuesday. The lag-1 autocorrelation was negative 0.01, so a day's return gave no information about the next day's. The fund's volatility of 32.8% sits between PSQ, the unlevered Nasdaq inverse, and QID, the 2x one, and its holding return of negative 21.57% a year sits between them too: PSQ held for negative 13.76% and QID for negative 29.46%. SQQQ held for negative 42.33%, and REW for negative 36.12%.
In the peer table the RSI(2) snapback was the best template on QID, PSQ, SH and SPDN, and the golden cross on SQQQ, SDS and REW. On SH, the unlevered S&P 500 inverse, holding lost 10% a year and the best template lost 0.53%. SDS is the 2x version of that fund, and the best template on it lost 2.33% a year, with a correlation of 1 to SH. The 2x fund lost more than the 1x fund on every holding figure on these pages.
Build it from blocks (or type it in English), backtest it on 5.7 years of minute data in seconds, tweak any parameter, then paper trade it on live data. No card, no broker needed to start.
Frequently asked questions
What was the best strategy for SDS?
Of the 12 templates tested on SDS over 2021-01-04 to 2026-10-02, the strongest by CAGR was golden cross at −2.3% (max drawdown 33.4%), versus −21.6% for buy-and-hold. The best result in hindsight is not a forecast. Check drawdowns and trade counts before drawing conclusions.
Did any strategy beat buying and holding SDS?
12 of 12 templates beat SDS buy-and-hold (−21.6% CAGR) on this window; 12 of 12 had a shallower maximum drawdown than holding (76.5%).
What was the best strategy for SDS from 2021 to 2026?
Golden cross at minus 2.33% a year, with 2 round trips and a 33.37% drawdown. Buy-and-hold lost 21.57% a year and ended at $2,479. All 12 templates beat holding on CAGR, but none had a positive CAGR.
Why did every template beat buy-and-hold on SDS?
The fund fell in five of the six calendar years, so templates that spent time in cash avoided part of the loss. Exposure ranged from 15.2% for momentum breakout to 97.6% for the dip buyer. The lowest-exposure templates lost the least.
How did SDS do in 2022?
It gained 30.5% while the S&P 500 fell 18.16%. It was the only up year. Buy-and-hold gained 28.2% in the test and RSI(2) snapback gained 49.3%, the best of the templates after weekly 7% target at 60.6%.
Does SDS return exactly minus 2 times the S&P 500?
Only daily. Over 5.7 years SPY returned 125.29% and SDS returned minus 79.38%. An ideal daily minus 2x fund would have returned minus 87.63%, so the fund finished slightly above that figure.
How much do trading costs matter on SDS?
RSI(2) snapback fell from minus 3.58% to minus 8.99% at 10 basis points per trade, and weekly 7% target fell from minus 15.53% to minus 21.42%. Golden cross barely moved because it made 2 round trips.
How did the trend rules do on SDS?
In this window the trend rules lost between 2.33% and 11.61% a year. The 200-day filter had 17 round trips and none won. The fund was above its 200-day average on 18.07% of sessions and crossed it 34 times, so signals were short-lived.
Other inverse etfs
Backtests are hypothetical, computed by DeployQuant's engine on minute-resolution consolidated US market data (2021-01-04 to 2026-10-02, $10,000 starting capital, no margin, no fees or slippage in the headline run; buy-and-hold puts 98% of the account in at the first open, as the templates do) and do not guarantee future results. Nothing on this page is investment advice. Live trading involves risk of loss.