Learn › ETFs › SOXS

SOXS trading strategies, backtested

Direxion Daily Semiconductor Bear 3X Shares: -3x daily semiconductors, with heavy decay outside sharp selloffs. Every DeployQuant template run on SOXS over 5.7 years of minute data, same engine, same window, sorted by return.

Quick answer: the best-performing template on SOXS (2021-01-04 → 2026-10-02) was trend + trailing stop at −4.9% CAGR vs −48.3% for buy-and-hold. 7 of 12 templates beat holding; 7 cut the max drawdown.

SOXS is the Direxion Daily Semiconductor Bear 3X fund. It aims to move three times the daily move of a semiconductor index in the opposite direction. From 2021-01-04 to 2026-10-02 the fund's own price history shows a total return of -99.99%, a CAGR of -80.02% and annualized volatility of 114%. The maximum drawdown starts on the first day of the window and ends on the last, with no recovery. Twelve templates were run on it, all with the same engine, the same window and a $10,000 starting balance.

No template finished with a positive CAGR. The best was trend plus trailing stop at -4.93%, and the worst was the monthly cycle at -77.46%, which ended with $1.93. Buy-and-hold in the engine returned -48.27% a year, ended at $227.31 and had a maximum drawdown of 97.81%. Seven of the 12 templates beat that CAGR and seven had a shallower drawdown. Every one of them lost money, so beating holding here means losing less.

The engine's holding line and the fund's own price series do not agree. The engine's holding return was -97.73% over the window against -99.99% for the fund. By calendar year the engine's holding shows -79.4% in 2021, 13.8% in 2022, -76.7% in 2023, -35.7% in 2024, -32.4% in 2025 and -7.7% in 2026. The fund's own years were -81.13%, 15.42%, -84.6%, -59.44%, -85.53% and -95.1%. The two are close in 2021 and 2022 and far apart in the last three years. The facts do not say why. The monthly cycle sat in the fund on 95.2% of sessions, and its yearly returns of -72.8%, 54.1%, -80.6%, -60.8%, -86.6% and -95.6% follow the fund's own years more closely than the holding line does. This page compares templates with the engine's holding figure because every row comes from the same engine, and it reads the fund's own series for the behaviour tables.

StrategyCAGRmax DDSharpetradeswin ratefinal value
SMA-200 Trend + 15% Trailing Stop −4.9% −64.5% 0.0829 34% $7,481
20-Day Momentum + Trailing Stop −13.6% −77.5% -0.01129 31% $4,320
200-Day SMA Regime Filter −19.7% −72.5% -0.2214 0% $2,830
Golden Cross (SMA 50/200) −24.4% −85.5% -0.303 0% $2,000
RSI(2) Dip Snapback −29.9% −94.2% -0.12170 56% (+1 open) $1,302
3-Month Momentum Switch −41.1% −96.0% -0.6127 22% $478
Drawdown Dip Buyer + 8% Target −46.2% −97.9% -0.444 100% (+1 open) $286
EMA 12/26 Trend −48.4% −97.8% -0.6528 7% $224
SMA 10/50 Trend −49.0% −97.9% -0.6519 11% $211
RSI(14) Mean Reversion −53.7% −99.5% -0.5215 40% (+1 open) $120
Weekly Entry + 7% Target −57.8% −99.5% -0.75287 51% $70
First-to-Last Day of Month −77.5% −100.0% -0.8169 29% (+1 open) $2
Buy & hold SOXS −48.3%−97.8% -0.54–– $227

Trend plus trailing stop, the best result

The best template made 29 round trips and won 10 of them, a 34% win rate. The average win was 20.73% and the average loss was 9.92%, which gives a profit factor of 0.83. It was invested 8.1% of the time with an average hold of 5.7 days. Its CAGR was -4.93%, its Sharpe ratio was 0.08 and its maximum drawdown was 64.55%.

The yearly returns show where it came from. The rule returned 0% in 2021, 2023, 2024 and 2026, which means it held no position in those years. In 2022 it lost 54% on 22 round trips, of which 6 won. In 2025 it made 62.6% on 7 round trips, of which 4 won. Against the engine's holding line it came out ahead in five of the six years and behind only in 2022, when holding returned 13.8% and the rule lost 54%.

The best trade, 87.73%, ran from 2025-03-17 to 2025-04-07, 21 days. The next best was 33.51% from 2022-10-04 to 2022-10-13, then 27.64% from 2022-06-27 to 2022-07-06. The worst trade was -16.87% from 2025-04-11 to 2025-04-14, and the stop kept every single loss near its 15% trail: the next worst were -15.83%, -14.39% and -14.12%. The median trade lost 7.5% and the median hold was 4 days. A rule with a one-third win rate and a median loser needs a few very large winners, and in this window the 87.73% trade in spring 2025 was the one that did most of the work.

The maximum drawdown of 64.55% began on 2022-10-14 and reached its trough on 2025-03-05 without recovering by the end of the test. That is the stretch where the fund kept falling and the rule kept buying signals that failed. The rule's own return in 2025 is real, and it still left the account well below where it started: the final value was $7,481.40. Costs changed little. At 5 basis points per trade the CAGR was -5.41% and at 10 basis points it was -5.88%.

Rules that traded rarely

Four templates made a handful of trades, and none of them won often. The 200-day regime filter made 14 round trips and lost all 14, with an average loss of 8.55% and a CAGR of -19.74%. Twelve of those trips ended in 2022 and two in 2025. Its best trade lost 0.23%, held from 2022-09-14 to 2022-11-09. Its worst, -25.3%, ran 14 days from 2025-04-11. SOXS closed above its 200-day average on 10.2% of sessions and crossed it 28 times, so the filter had few chances and each one was a short rise inside a long fall.

Golden cross made 3 round trips and lost all 3: -9.99% over 93 days from 2022-06-14, -48.61% over 66 days from 2022-10-10, and -58.85% over 58 days from 2025-04-02. The CAGR was -24.44% and the maximum drawdown 85.51%. The last trade spans the two biggest sessions of April 2025. It entered the day before the 30.41% jump of 2025-04-03 and was still in when the fund fell 56.32% on 2025-04-09. The averages needed weeks to confirm the first rise, and the fund gave it back in the next few sessions.

The 3-month momentum switch made 27 round trips and won 6, a 22% win rate. The average win was 9.03% and the average loss was 14.69%, giving a profit factor of 0.27. Its best trade was 46.28% from 2022-03-31 to 2022-04-29 and its worst was -53.09% over 36 days from 2022-10-11. The CAGR was -41.12% and the drawdown 96.01%.

The two moving-average rules landed close to holding. SMA 10/50 made 19 round trips and won 2, with an average win of 4.16% and an average loss of 20.69%. EMA 12/26 made 28 round trips and won 2, with a longest losing streak of 16. Their CAGRs were -48.95% and -48.38%, against -48.27% for holding, and their drawdowns were 97.93% and 97.76%. They were invested on 27.3% and 25.5% of sessions, so they sat out part of the fall and still ended in the same place.

Rules that wait for a rise to confirm buy late on a fund like this. SOXS has a lag-1 daily autocorrelation of -0.09, slightly negative, and the median 5-day forward return across all sessions was -2.9%. A rise is more likely to reverse than to continue, and the confirmation rules pay for that on each signal.

Rules that traded often

Momentum breakout made 129 round trips with a 31% win rate. The average win was 12.24% and the average loss was 5.78%, and the profit factor was 0.79. Average hold was 3 days and exposure was 18.3%. The CAGR was -13.6%, the second best in the group, and the drawdown 77.45%. It lost in 2021, 2022, 2023 and 2024, with -31.9%, -6.1%, -43.2% and -34.5%, then made 48% in 2025 and 22.8% in 2026. Its best trade was 40.76% from 2026-07-23 to 2026-07-29. At 10 basis points the CAGR fell to -17.35%.

The RSI(2) snapback made 170 round trips, won 95 of them and had a 56% win rate, the highest win rate among the rules that made more than a handful of trades. The average win was 7.37% and the average loss was 10.04%, and the profit factor was 0.91. The yearly returns are split. It made 5.6% in 2021 and 56.7% in 2022, then lost 14%, 11.6%, 28.7% and 85.5% in 2023 to 2026. The worst trade, -62.49%, ran 27 days from 2026-04-02, and April 2026 was its worst month at -61.24%. A snapback rule buys an extreme low and expects a bounce. On a fund that keeps falling, the occasional trade that does not bounce is larger than the many small ones that do. A position opened on 2026-09-28 was still open at the end, down 10.94%.

RSI(14) mean reversion made 15 round trips and won 6, with an average win of 22.56% and an average loss of 38.02%. Its CAGR was -53.71% and its drawdown 99.45%. It held for 93.1 days on average. It made 124.1% in 2022 and lost 58.5%, 63.8%, 39.1%, 54.2% and 87.2% in the other five years. The two worst trades were -89.55% over 232 days from 2025-12-10 and -73.09% over 195 days from 2025-05-08. Mean reversion on a fund that decays is a bet that the next rise arrives soon, and these two holds show what happens when it does not.

The weekly 7% target made 287 round trips with a 51% win rate. The average win was 7.52% and the average loss was 10.07%, and the profit factor was 0.8. Average hold was 2.4 days and exposure was 45.4%. The CAGR was -57.82% and the drawdown 99.54%, and it ended with $70.38. The target fills often because the fund's average intraday range is 7.83%, so a 7% limit order sits inside a normal day. The median trade made 6.99%. The losers are not capped by price, and on a fund whose average down day is 5.31% they were larger than the winners. The worst trade was -48.95% from 2025-04-07 to 2025-04-10, followed by -31.57% from 2025-04-21 to 2025-04-24. Calendar years were -37.5%, 6.8%, -59.5%, -54.7%, -85.1% and -62.8%.

The monthly cycle made 69 round trips and won 20, a 29% win rate, with an average win of 24.68% and an average loss of 21.68%. It ended with $1.93 and a drawdown of 99.98%. Its best trades were 66.37% in June 2022 and 57.74% in January 2022. Its worst were -63.41% in April 2026, -55.29% in May 2026 and -43.9% in June 2026, three months in a row. A position opened on 2026-10-01 was open at the end, down 10.11%.

The dip buyer shows a different failure. It completed 4 round trips and won all 4 at the 8% target, in February and March 2021. It then bought on 2021-03-09 at an adjusted price of $234,653.22 and was still holding at the end, down 99.99% on that position. Its CAGR was -46.17%, slightly better than holding's -48.27%, with a drawdown of 97.89%. The win rate of 100% describes four small trades before one very large loss. A target that books 8% does nothing for a position that never rises 8% again.

How each strategy traded SOXS

StrategyTime in marketAvg hold (days)Best tradeWorst tradeProfit factorWith 10 bps slippage
trend + trailing stop8.1%687.7%−16.9%0.83−5.9%
momentum breakout18.3%340.8%−12.7%0.79−17.4%
200-day regime filter8.8%13−0.2%−25.3%–−20.1%
golden cross10.5%72−10.0%−58.9%–−24.6%
RSI(2) snapback47.4%625.0%−62.5%0.91−33.9%
3-month momentum17.9%1446.3%−53.1%0.27−41.7%
dip buyer98.3%78.0%8.0%–−46.5%
EMA 12/26 trend25.5%1916.7%−25.3%0.07−48.9%
SMA 10/50 trend27.3%308.1%−36.0%0.02−49.3%
RSI mean reversion67.4%9351.7%−89.5%0.32−54.0%
weekly 7% target45.4%221.4%−49.0%0.80−62.2%
monthly cycle95.2%2866.4%−63.4%0.52−78.0%

Time in market, hold lengths and costs

Exposure ran from 8.1% for trend plus trailing stop and 8.8% for the 200-day filter to 95.2% for the monthly cycle and 98.3% for the dip buyer. The monthly cycle, the rule with the highest exposure among those that traded, had the lowest CAGR of the 12, and the two trend rules with the least exposure, trend plus trailing stop and the 200-day filter, finished first and third. That fits the fund's behaviour: it rose on only 45.81% of sessions and its average down day was -5.31%, so each extra day in it cost money on average. The exceptions are the high-turnover rules. The weekly target was invested 45.4% of the time and the RSI(2) snapback 47.4%.

Average hold ran from 2.4 days for the weekly target and 3 for momentum breakout to 93.1 for RSI(14) mean reversion and 72.3 for golden cross. Profit factors were below 1 for every template that had losses. They were 0.91 for the RSI(2) snapback, 0.83 for trend plus trailing stop, 0.8 for the weekly target, 0.79 for momentum breakout, 0.52 for the monthly cycle, 0.32 for RSI(14) mean reversion, 0.27 for the 3-month switch, 0.07 for EMA 12/26 and 0.02 for SMA 10/50. The 200-day filter and golden cross had a profit factor of 0 because neither won a trade.

Costs of 5 and 10 basis points per trade moved the results by less than the strategies' differences. The weekly target went from -57.82% to -60.75% and -62.23%, the most of any template, since it made 287 round trips. The RSI(2) snapback went from -29.89% to -31.92% and -33.9%. The monthly cycle went from -77.46% to -77.74% and -78.02%. Rules that already lost most of the account have little left for costs to take, and the ranking by CAGR was the same at 10 basis points as at zero.

Liquidity is a separate limit. SOXS traded an average of $775,860,545 a day, with a median minute volume of 128 shares. The daily figure is large, and the minute figure is small. A $10,000 order is small against either.

How SOXS behaved

MeasureSOXS
Data in this test2021-01-04 to 2026-10-02 (1444 sessions)
Total return, buy and hold−100.0%
Annualized volatility114.0%
Deepest drawdown−100.0% (2021-01-04 to 2026-10-02)
Up days45.8%
Average daily range7.83%
Average overnight gap3.19%
Correlation to SPY-0.79
Correlation to QQQ-0.87
Correlation to TLT-0.05
Sessions above the 200-day average10.2%
Crossings of the 200-day average28
Falls of 10% or more from a 20-day high86

Calendar years

YearReturn
2021−81.1%
202215.4%
2023−84.6%
2024−59.4%
2025−85.5%
2026 (part)−95.1%

Biggest single days

Best dayMove
2026-06-0531.4%
2025-04-0330.4%
2025-04-1023.9%
2026-06-2323.8%
2025-01-2723.4%
Worst dayMove
2025-04-09−56.3%
2022-11-10−31.4%
2026-07-30−26.4%
2026-06-11−24.8%
2025-05-12−21.6%

Average return by calendar month

JanFebMarAprMayJunJulAugSepOctNovDec
−12.1%−10.7%−4.9%−2.5%−31.3%−12.6%1.3%0.7%2.8%−7.0%−22.6%−4.6%

Most and least correlated funds

Most correlatedLeast correlated
TECS0.90SOXX-1.00
REW0.89SOXL-1.00
QID0.87XLK-0.90
SQQQ0.87TECL-0.90
PSQ0.87ROM-0.90

Leverage and decay against SOXX

YearSOXSSOXX-3× SOXX, reset daily
2021−81.1%44.7%−81.1%
202215.4%−35.1%12.5%
2023−84.6%67.0%−87.0%
2024−59.4%13.0%−66.4%
2025−85.5%40.7%−87.5%
2026−95.1%95.9%−95.7%

The last column compounds -3× SOXX'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 -3× the underlying's return over a year.

Over 2021-01-04 to 2026-10-02, SOXS returned −100.0% while SOXX returned 388.4% and a perfect daily-reset -3× version would have returned −100.0%. Its measured daily beta to SOXX was -2.99.

The structure of a -3x fund

SOXS tracks SOXX with a daily multiple of -3. Measured against SOXX over the full window, its realized beta was -2.99 with an R-squared of 1. SOXX returned 388.35% over the window, and the fund returned -99.99%. The table above compares the fund each year with a frictionless fund that resets daily to -3 times SOXX. Over the whole window the daily-reset ideal was also -99.99%, so the fund did what daily rebalancing predicts.

The yearly lines are close to the ideal and differ by a few points. In 2021 the fund returned -81.13% against an ideal of -81.06%. In 2022 SOXX fell 35.07% and SOXS rose 15.42%, where the daily-reset ideal gives 12.52%. In 2023 SOXX rose 66.96% and SOXS fell 84.6% against an ideal of -87.01%. In 2024 the fund returned -59.44% against -66.36%, the widest difference of the six years. In 2025 SOXX rose 40.68% and SOXS fell 85.53%, against -87.48%. In 2026 so far SOXX has risen 95.9% and SOXS has fallen 95.1%, against -95.66%. The table's caption notes that fees, borrowing costs and interest on cash put a real fund below or above the ideal line, which fits the fund finishing above it in five of the six years.

In every year in which SOXX rose, SOXS lost most of its value, and in the one year in which SOXX fell, SOXS gained 15.42%. That is the structure the fund description refers to. A daily -3x fund pays for volatility in both directions. The 7.83% average intraday range and the 114% annualized volatility are the inputs. For a trader the practical point is that the fund's return over a long hold depends on the path of the semiconductor index, and the path in this window was up with sharp reversals.

The correlation to SOXX and SOXL was -1, and to XLK, TECL and ROM it was about -0.9. The most correlated funds were TECS at 0.9, REW at 0.89, and QID, SQQQ and PSQ at 0.87. Beta to SPY was -5.52 and to QQQ -4.45, and the correlation to TLT was -0.05.

Daily behaviour, seasonality and RSI

SOXS had up days on 45.81% of sessions. The average up day was 5.4% and the average down day was -5.31%. The average intraday range was 7.83% and the average overnight gap was 3.19%. The overnight log return was -476.61% and the intraday log return was -445.31%, so the loss was spread across both halves of the day.

The worst day was 2025-04-09 at -56.32%, followed by 2022-11-10 at -31.39%, 2026-07-30 at -26.4%, 2026-06-11 at -24.8% and 2025-05-12 at -21.61%. The best days were 2026-06-05 at 31.42%, 2025-04-03 at 30.41%, 2025-04-10 at 23.92%, 2026-06-23 at 23.8% and 2025-01-27 at 23.41%. Two of the best days and the worst day fall within one week of April 2025. A fund that adds 30.41% in one session and loses 56.32% a few sessions later punishes any rule that holds through the reversal, and it rewards a rule that happens to be flat.

By calendar month, May averaged -31.31% and November -22.58%, with June at -12.63%, January -12.12% and February -10.73%. July, August and September were positive at 1.25%, 0.68% and 2.8%. Each month has five or six observations, so these describe this window and do not predict another. The best single month for holding was 2022-06 at 56.32% and the worst was 2022-11 at -46.52%.

The fund closed above its 200-day average on 10.2% of sessions. RSI(14) was below 30 on 119 sessions and above 70 on 8. After an RSI(14) reading below 30, the median forward 5-day return was -2.05% against a baseline of -2.9%, and the 20-day figure was -14.98% against -9.87%. After RSI(2) fell below 10, on 230 sessions, the median 5-day return was -1.74% and the 20-day return -9.44%. The 5-day numbers are slightly better than the baseline and the 20-day RSI(14) figure is worse. A low RSI reading on SOXS does not mark a bottom, and the baseline itself is negative at every horizon.

Falls of 10% or more from a 20-day high occurred in 86 episodes covering 1,049 days, so the fund spent most of the window in or near such a fall. The 200-day filter's 14 losing trades and golden cross's 3 follow from that: a rule that needs the fund above a long average has few chances on a fund that is below it for roughly nine sessions in ten.

Other inverse funds

Among the other inverse funds in the study, the best templates differ. On SQQQ golden cross led with -7.41% against -42.33% for holding. On QID the RSI(2) snapback returned -1.85% against -29.46%, and on PSQ it returned 0.13% against -13.76%. On SDS golden cross returned -2.33% against -21.57%, on TECS trend plus trailing stop returned -11.04% against -46.67%, and on REW golden cross returned -6.71% against -36.12%. EEV had the one clearly positive result, RSI(14) mean reversion at 5% against -16.07%. TBF, which gained 12.21% buy-and-hold, was led by the monthly cycle at 11.93%.

On SOXS the best template was trend plus trailing stop at -4.93%, a different winner from SQQQ, QID, SDS and REW, where golden cross or the RSI(2) snapback led. The test is one window with daily decisions, minute-bar fills, no margin and no fees in the headline run. Prices in the trade lists are adjusted for splits and dividends, so entry prices such as $234,653.22 are adjusted figures and are higher than the quotes printed at the time. A fund that lost 99.99% of its value in the window has no meaningful long-run average, and a different start date would give a different ranking.

Run a SOXS strategy yourself, free →

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 SOXS?

Of the 12 templates tested on SOXS over 2021-01-04 to 2026-10-02, the strongest by CAGR was trend + trailing stop at −4.9% (max drawdown 64.5%), versus −48.3% 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 SOXS?

7 of 12 templates beat SOXS buy-and-hold (−48.3% CAGR) on this window; 7 of 12 had a shallower maximum drawdown than holding (97.8%).

What was the best strategy for SOXS?

Trend plus trailing stop, at -4.93% a year against -48.27% for holding, with a maximum drawdown of 64.55%. It made 29 round trips and won 10. All 12 templates lost money, so the best result is the smallest loss, and nearly all of it came from 2025.

Did any strategy make money on SOXS?

No. The best template, trend plus trailing stop, had a CAGR of -4.93%. It made 62.6% in 2025 and lost 54% in 2022. The other 11 templates ranged from -13.6% for momentum breakout to -77.46% for the monthly cycle.

Why does SOXS lose money over long periods?

It resets to -3 times a semiconductor index every day. The fund's total return was -99.99% from 2021-01-04 to 2026-10-02 while SOXX returned 388.35%. SOXS fell in every calendar year except 2022, including 2023, 2025 and 2026 when it lost 84.6%, 85.53% and 95.1%.

Does the weekly 7% target work on SOXS?

It returned -57.82% a year with a 99.54% drawdown from 287 round trips. It won 51% of them, but the average win was 7.52% and the average loss was 10.07%. The profit factor was 0.8, and the account ended with $70.38.

What was the worst day for SOXS?

2025-04-09, when it fell 56.32%. The best day was 2026-06-05 at 31.42%, and 2025-04-03 added 30.41%. Golden cross entered on 2025-04-02, spanned both days and lost 58.85%.

Is RSI a useful signal for SOXS?

Not in this window. After RSI(14) fell below 30, the median 20-day forward return was -14.98%, against a baseline of -9.87%. The RSI(2) snapback returned -29.89% and RSI(14) mean reversion returned -53.71%, though the RSI(2) rule did make 5.6% in 2021 and 56.7% in 2022.

Does holding SOXS match the fund's own return?

Not in the engine. Holding in the engine returned -97.73% over the window and the fund's own price series returned -99.99%. The two are close in 2021 and 2022 and differ from 2023, and the facts do not explain the gap.

Other inverse etfs

SQQQProShares UltraPro Short QQQQIDProShares UltraShort QQQPSQProShares Short QQQSHProShares Short S&P500SDSProShares UltraShort S&P500SPDNDirexion Daily S&P 500 Bear 1X SharesTECSDirexion Daily Technology Bear 3X SharesREWProShares UltraShort Technology

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.