20-Day Momentum + Trailing Stop on SOXL

Direxion Daily Semiconductor Bull 3X Shares: 3x daily leveraged semiconductors, one of the more volatile ETFs listed. Backtest 2021-01-04 to 2026-10-02, $10,000 starting capital, computed by the same engine that runs live DeployQuant strategies.

The 20-Day Momentum + Trailing Stop template bought SOXL 138 times between 2021-01-04 and 2026-10-02, with one more position still open at the end. It made 19.1% a year and turned $10,000 into $27,225. Holding SOXL over the same window made 33.3% a year and ended at $52,126, so the strategy finished well behind on return. It took a maximum drawdown of 67.5% where holding took 89.6%. The Sharpe ratio was 0.59 against 0.82.

SOXL is the 3x semiconductor fund, with annualized volatility of 113.3%. On a fund that moves like that, a rule that buys after a gain of more than 10% in 20 sessions and sells on a 10% trailing stop makes many short trades. The win rate was 37% and the average hold was 5.6 days. The average win was 14.55% and the average loss was 6.23%, which gives a profit factor of 1.20. The strategy was invested 37.0% of the time.

The test is a backtest only. Trailing-stop exits are not deployable live yet, at any broker, so this template can be studied and backtested but not switched to real money for now. Results here come from the same engine that runs live strategies, over 5.74 years, with no fees or slippage in the headline run. The template ranked seventh of the 12 on SOXL, and SOXL was the second-best of the 59 funds for this template. Both rankings come from the same window and describe it, not a later period. Other results on the same fund are on the SOXL page and on the momentum breakout strategy page.

Year by year

Year by year against holding SOXL

The calendar-year table shows how differently the two lines behaved. Holding SOXL made 119.0% in 2021, lost 84.8% in 2022, made 211.2% in 2023, lost 12.1% in 2024, made 53.6% in 2025 and made 285.2% in the partial year 2026. The strategy lost 37.2% in 2021, lost 14.6% in 2022, made 28.6% in 2023, made 12.0% in 2024, made 68.9% in 2025 and made 108.8% in 2026.

It beat holding in three of the six years: 2022, 2024 and 2025. In 2022 the gap was 70.2 points in its favour. The strategy lost 14.6% in a year that cost holders 84.8%, because the 10% trailing stop got it out of the early part of every decline and the entry condition, a 20-day gain above 10%, kept it out of the market during most of the slide. In 2024 it made 12.0% while the fund lost 12.1%. In 2025 it made 68.9% against 53.6%.

It lost badly in the years when SOXL rose hardest. In 2023 the fund made 211.2% and the strategy 28.6%. In 2026 the fund made 285.2% against 108.8%. The strategy was in cash through most of those rallies, and it was shaken out of the positions it did take by 10% pullbacks that are routine in a fund with a 7.77% average daily range.

2021 was the worst year for the rule. It lost 37.2% while the fund made 119.0%. The rule bought bursts that reversed. In 2021, 6 of its 20 closed round trips won, and the first trade of the window, from 2021-02-03 to 2021-02-04, lost 8.94% in a single day. That is the failure mode the template warns about: a burst that reverses at once, so the trail exits about 10% below the entry.

The tables for related rules on this fund show different shapes. The golden cross on SOXL made 45.9% a year and the RSI(2) snapback on SOXL made 39.2%, both above the holding return. The momentum breakout made less than both and less than holding, with a drawdown close to theirs.

Month by month

The months that carried the result

The monthly table is very uneven. The best month was April 2026 at 112.66%, and the worst was June 2026 at negative 24.19%. Holding made 162.61% in April 2026 and lost 56.9% in July 2026. The strategy's other large months were June 2025 at 52.54%, May 2026 at 44.61%, October 2025 at 40.92%, December 2023 at 27.22% and July 2022 at 26.62%.

April 2026 alone is a large part of the total. The trade behind it ran from 2026-04-07 to 2026-04-28 for a gain of 100.74%, entered at an adjusted price of 54.14 and exited at 108.68 after 21 days. It is the best trade in the test by a wide margin. Without it the strategy would have trailed holding by more, and the 19.1% would be lower.

Several months are exactly zero. The strategy was in cash through September and October 2022, September and October 2023, November 2024, and February to April 2025. The strategy held nothing in those months, so it avoided the fund's moves in both directions.

The sequence of losing months in 2021 is the clearest case of the whipsaw. February, March and April 2021 lost 5.22%, 5.28% and 15.37%, followed by July, August and September at negative 13.29%, negative 14.01% and negative 12.29%. In 2022 the strategy lost in January, March, April, June and August. Late in the year, November 2022 made 18.96%.

The stretch from May 2023 to March 2024 shows the opposite, with pauses in between. May, June and July 2023 made 20.92%, 17.97% and 14.13%, December 2023 made 27.22%, and January to March 2024 made 12.77%, 16.14% and 18.58%. Momentum rules work when the fund trends for weeks, and this stretch was one. The months do not distribute evenly, so a different start date would give a different total.

Every trade

momentum breakout on SOXL made 138 closed round trips and one position still open at the end of the test, an average hold of 6 days, an average winner of 14.55%, an average loser of −6.23%, a profit factor of 1.20, a longest losing streak of 10. It held a position at the close on 37.0% of trading days.

Best 10 round trips

Worst 10 round trips

Prices are adjusted for splits and dividends, so they sit below the quotes printed at the time. An open position is marked at the last close.

What the 139 trades looked like

The strategy made 138 completed round trips and one more position was open at the end, entered on 2026-09-29 at an adjusted price of 148.50 and showing a gain of 10.27%. The median trade lost 2.79%. The median hold was 4 days, the longest was 33 days and the shortest was under a day. Of the 138 closed trips, 51 won.

The median trade losing while the average trade wins is the signature of a trend rule. Most trades stop out for a small loss and a few run far enough to cover them. The average win of 14.55% is more than twice the average loss of 6.23%, and the profit factor is 1.20. The longest winning streak was 4 trades and the longest losing streak was 10.

The five best trades were 2026-04-07 to 2026-04-28 at 100.74%, 2023-05-24 to 2023-05-31 at 37.76% in 7 days, 2026-05-20 to 2026-05-27 at 34.44% in 7 days, 2025-05-05 to 2025-05-19 at 32.30% in 14 days, and 2024-01-03 to 2024-01-26 at 32.09% in 23 days. All five are under a month. The largest gains came over days, because a 3x fund can double in three weeks.

The worst trades were all between negative 10.0% and negative 11.81%, which is what the trailing stop does: it caps the loss near the 10% trail, and gaps carry it a little further. The worst was 2022-11-16 to 2022-11-17 at negative 11.81%, followed by 2025-08-29 to 2025-09-02 at negative 11.41%, 2026-06-18 to 2026-06-23 at negative 11.05%, 2025-05-22 to 2025-05-23 at negative 10.36% and 2024-10-17 to 2024-10-23 at negative 10.00%. The stop worked as designed. No trade lost more than 11.81%.

The winners by exit year show the rhythm. In 2021, 6 of 20 closed trips won. In 2022, 7 of 18 won. In 2023, 8 of 23. In 2024, 10 of 23, in 2025, 9 of 21, and in 2026, 11 of 33. The 33 trips in 2026 is the busiest year, with a strong market giving many entry signals. The yearly win rate was lowest in 2021 and highest in 2024 and 2025.

Largest drawdowns

Buy-and-hold's deepest drawdown ran from 2021-12-27 to 2022-10-14 and reached −89.6%.

How deep the drawdowns went

The strategy's deepest drawdown was 67.5%, from 2021-02-16 to 2023-05-24, and it recovered on 2024-05-28. That is 827 days down and 370 days to recover. The second was 57.8%, from 2026-06-03 to 2026-09-29, and it had not recovered by the end of the test. The third was 44.5%, from 2024-07-10 to 2025-05-06, recovering on 2025-06-24.

Holding SOXL had a 89.6% drawdown from 2021-12-27 to 2022-10-14, and it did not recover until 2026-02-25, which is 1,230 days after the trough. A second drawdown of 69.3% began on 2026-06-22 and reached its trough on 2026-07-29 without recovering. A third was 43.0% in early 2026, recovered in 10 days.

The strategy's drawdowns are shallower than holding's, but they are large in absolute terms. A fall of 67.5% is a large loss to sit through, and the strategy was inside that first drawdown for more than two years. The 10% trailing stop limits each trade to about 10%, but it does not limit the sequence. A run of losing trades can compound, and the strategy lost 10 trades in a row at one point.

The drawdown figure does not show the whole risk of the live fund. SOXL's own worst days, negative 30.71% on 2026-06-05 and negative 29.79% on 2025-04-03, are bigger than the trailing stop. A stop at 10% below the high-water mark cannot cap a single-day move of that size, so a day like that inside a position would cost more than the stop level.

The drawdown also grew under costs, from 67.5% to 69.1% at 5 basis points and 70.5% at 10 basis points.

With trading costs

The headline run fills at the bar price. These runs charge slippage on every fill.

What costs take from a strategy that trades every few days

At 5 basis points per trade the strategy made 16.2% a year and ended at $23,683, with a Sharpe ratio of 0.55. At 10 basis points it made 13.6% and ended at $20,806, with a Sharpe ratio of 0.51. The headline 19.1% is the figure with no cost.

With 277 fills the effect is visible but less severe than on a rule whose average trade is small. The average win is 14.55%, so a cost of 10 basis points on each fill is a small share of a winning trade. It is a larger share of the median trade, which lost 2.79%. SOXL is also very liquid for a leveraged fund. It trades about $2,243,524,454 a day and the median minute bar has 101,543.5 shares, so the cost figure is a fair assumption for a small order. These runs assume a flat cost and ignore the gap risk at a stop: a trailing stop that fires in a fast market can fill below its level.

After costs the strategy remains well behind holding at 33.3%. The gap to holding is wider after costs than before them. It is not a cost problem, since the strategy trails by a large margin with no costs at all.

Changing the parameters

What the three parameter variants show

The published rule buys after a gain of 10% in 20 sessions and trails the stop 10% below the high-water mark. The page runs three variants.

A 5% trailing stop made 22.59% with a 45.1% drawdown and 310 trades, of which 111 won, a win rate of 36%. It had a Sharpe ratio of 0.69. A tighter stop exits sooner, takes more small losses and cuts the drawdown sharply. It made more than the published rule because it avoided part of the deepest decline.

A 15% trailing stop made 7.64% with an 80.8% drawdown and 81 trades, of which 34 won. A wider stop gives back more of each gain before it fires and it exposes the strategy to more of each decline. This is the weakest of the variants, and the drawdown is close to holding's 89.6%.

A 20% return entry made 40.66% with a 50.7% drawdown and 98 trades, of which 42 won, a win rate of 43%. It ended at $70,903, above the $52,126 of buy-and-hold, with a Sharpe ratio of 0.90. It is the only variant that beat holding on return. Requiring a gain of 20% in 20 sessions before buying filtered out weaker bursts and left the strategy with stronger signals.

The four versions range from 7.64% to 40.66%, a wide spread. That spread is the main finding. On this fund, the entry threshold mattered more than the stop width. It would be a mistake to take the 40.66% as a forecast, since the best variant of three, chosen after seeing the results on a single window, tends to look better than it would on new data. The comparison does show that the published rule's 10% entry was too loose for SOXL in this window.

How SOXL behaved

How SOXL behaved and why a 10% stop was tight

SOXL's annualized volatility was 113.3%. The average daily range was 7.77%, the average up day was 5.31% and the average down day was negative 5.29%. Up days were 53.4% of sessions. A 10% trailing stop on a fund that routinely moves 5% in a day is only a couple of average days wide. That is why the strategy averaged 5.6 days in a position and the median trade was a loss.

The fund's calendar years were 121.7% in 2021, negative 85.6% in 2022, 226.5% in 2023, negative 12.3% in 2024, 55.0% in 2025 and 289.9% for the partial year 2026. The best days were 2025-04-09 at 55.98%, 2022-11-10 at 31.49% and 2026-07-30 at 25.07%. The worst were 2026-06-05 at negative 30.71%, 2025-04-03 at negative 29.79% and 2025-04-10 at negative 25.07%. The fund fell 10% or more from a 20-day high on 102 occasions, covering 769 days.

The overnight share of the log return was 156.3% and the intraday share was negative 56.3%. The fund gained 261.62% in log terms overnight and lost 94.21% during sessions. The average overnight gap was 3.17%. A rule that buys at the open and exits on a trail that triggers during the session collects less of the overnight gain than a rule that holds across closes, and the strategy's average hold of 5.6 days limits how much it captures.

The fund spent 58.5% of its sessions above the 200-day average and crossed it 44 times. The lag-1 autocorrelation was negative 0.08, so there was a slight tendency to reverse. That helps a mean reversion rule and hurts a momentum rule. After RSI(2) closed below 10, on 140 sessions, the median 5-day return was 1.70% against 1.22% on any day, and the median 20-day return was 6.65% against 2.96%. After the 14-day RSI closed below 30, on 7 sessions, the 5-day return was 11.92% and the 20-day return was 16.81%. The rebound after a washout was large, and the RSI(2) snapback made 39.2% on SOXL where the momentum breakout made less.

SOXL tracks three times the daily return of SOXX. Its realized beta to SOXX was 2.97. Over the window SOXX returned 388.35% and SOXL returned 450.98%. A daily-rebalanced model of 3x gives 854.10%. By calendar month, May averaged 33.67% and November 24.84%, while August averaged negative 8.44% and March negative 4.43%. Each month has 5 or 6 observations.

Among the leveraged funds, the same rule made 22.5% on QLD, 17.3% on SPUU, 15.4% on SSO, 10.2% on TECL and 0.6% on TQQQ. SOXL at 19.1% was second among the 59 funds. The median for the leveraged group was 13.3% and the median across all 59 funds was 0%.

The rules

Buy after a +10% four-week burst and let a 10% trailing stop manage the exit.

Momentum entry, trailing-stop exit. A +10% gain over 20 sessions marks a burst of momentum. The template buys the strength and exits on a 10% trailing stop that ratchets up beneath the highest close and never moves down. There is no profit target, so winners run until the trail is hit.

Good for: high-momentum assets where trends extend, such as semiconductor and leveraged tech ETFs.
Watch out: buying strength means buying high; when a burst immediately reverses, the trail exits about 10% below the entry.

How the two rules interact with SOXL

The entry buys 98% of the sleeve at the open when the 20-day return is above 10% and the strategy is flat. The exit is a managed trailing stop that follows 10% below the position's high-water mark and ratchets up beneath the highest close. It never moves down. There is no profit target.

On SOXL the 10% entry fires often, because a fund with a 7.77% average daily range gains 10% in 20 sessions in many weeks. That is why there were 138 trades. The 10% trail is then too close to the price for the fund's daily swings. The result is a median trade of negative 2.79% and a median hold of 4 days, with a few large winners carrying the total.

The variants show where the interaction breaks. A 5% stop made more by cutting the drawdown, and a 15% stop made less by giving back more. A 20% entry made the most by firing less often and entering only after stronger bursts. The result depends more on how selective the entry is than on the stop distance.

For comparison, the weekly 7% target on SOXL made 29.5% and the SMA 10/50 trend made 18.2%. The momentum breakout's 19.1% sits near the lower middle of the 12 templates on this fund.

This template is backtest-only for now. Trailing-stop exits are not deployable live yet, so the live behaviour of the stop is not part of this test. The test stops at one fund and one window of 5.74 years, with the largest single trade in April 2026. None of the figures is a forecast.

Related

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.