RSI(14) Mean Reversion on TECL

Direxion Daily Technology Bull 3X Shares: 3x daily leveraged technology sector. Backtest 2021-01-04 to 2026-10-02, $10,000 starting capital, computed by the same engine that runs live DeployQuant strategies.

The RSI(14) mean reversion template buys when the 14-day RSI falls under 30 and sells the whole position when RSI climbs back over 70. It has no profit target and no stop. On TECL, the Direxion Daily Technology Bull 3X fund, it turned $10,000 into $34,212 between 2021-01-04 and 2026-10-02. That is a CAGR of 23.89%, a total return of 242.12% and a maximum drawdown of 64.11%, with a Sharpe ratio of 0.67.

Buying and holding TECL did more in raw terms. It ended at $63,981 with a CAGR of 38.16% and a Sharpe ratio of 0.82. The rule gave up 14.27 points of CAGR to holding. In return it was in the fund only 32.1% of the time and its drawdown was shallower than the 77.2% that holding went through. The rule beat holding in two calendar years, 2022 and 2024.

The template closed 18 round trips and won 13 of them, a win rate of 72%. The average win was 20.27% and the average loss was -18.72%, which gives a profit factor of 2.82. The average hold was 37.3 days. On this fund, RSI(14) under 30 was a rare condition: it appeared on 17 sessions in 1,444, and the median 20-day return after those sessions was 9.59% against a baseline of 3.74%.

TECL ranks first of the 59 funds for this template by CAGR. Among the 12 templates on TECL, RSI mean reversion ranked fifth. The test is one fund over one window of 5.74 years, with no fees or slippage in the headline run.

Year by year

How each year compared with holding TECL

The yearly table shows the profile of a rule that buys weakness and sells strength. The account returned 41.4% in 2021, -51.5% in 2022, 57.6% in 2023, 100.6% in 2024, -14.4% in 2025 and 84.5% in 2026 to date. Holding TECL returned 121.5%, -73.5%, 194.9%, 35.9%, 38% and 109.2% in the same years.

In the two years the rule won, the reasons differ. In 2022 it lost 51.5% against 73.5% for holding, an edge of 22 points, simply because it was out of the fund for part of the decline. In 2024 it returned 100.6% against 35.9%, an edge of 64.7 points. That is the one year in which the template made more than the fund by a wide margin, and it came from four trades that all won: 2024-01-05 to 2024-01-24 at +32.74%, 2024-04-19 to 2024-05-10 at +13.98%, 2024-07-26 to 2024-08-26 at +6.95% and 2024-09-09 to 2024-09-25 at +25.82%.

The years the rule trailed are the strong trend years. In 2023 TECL gained 194.9% and the rule gained 57.6%, a gap of 137.3 points. In 2021 the gap was 80.1 points and in 2025 it was 52.4 points. A rule that sells when RSI passes 70 leaves the position early in a sustained climb, and it can only re-enter after RSI has fallen under 30, which on this fund took long stretches of time.

The 2025 result needs one more detail. The rule entered on 2025-03-04 at an adjusted price of $65.33 and was still holding at the April 2025 low. It exited on 2025-04-30 at $47.74 for a loss of 26.92%. After that exit the account stayed in cash until 2025-11-19. TECL itself returned 38% for 2025, so the one losing trade of the year decided the comparison.

The 2026 figure of 84.5% against 109.2% for holding covers a partial year. The two 2026 trades both won: 2026-03-30 to 2026-04-16 at +49.58% and 2026-07-30 to 2026-08-18 at +24.74%. The first of these was the best trade of the whole test.

Month by month

The months with a position and the months without

Most months in the table are zeros. The template held a position in only part of the test, and the exposure figure of 32.1% shows how much. Long runs of cash include 2025-05 to 2025-10, 2026-01 and 2026-02, and 2024-10 through 2025-02.

The best month was April 2026 at +43.15%, produced by the 2026-03-30 entry at $83.2 and the 2026-04-16 exit at $124.45. Holding TECL made +68.31% in the same month, which was its best too. The worst month was September 2022 at -32.84%, almost the same as holding's -32.73%. In that month the account was in the August 2022 trade, which was bought on 2022-08-31 at $31.11 and sold on 2022-11-25 at $26.14.

The 2022 monthly path shows the main weakness of the template. It lost 3.51% in January, 15.63% in February, 12.3% in April, 6.47% in May, 27.27% in June and 32.84% in September. The only gains were March at +13.69%, July at +9.7%, October at +19.05% and November at +9.67%. Three buys in 2022, on 2022-01-18, 2022-04-19 and 2022-08-31, each found RSI under 30 and each held for 70 to 86 days while the fund kept falling.

The large positive months came in short, successful trades. January 2024 returned 31.96%, September 2024 returned 25.22% and November 2023 returned 21.98%. January 2023 returned 20.6%, mostly the tail of the 2022-12-21 entry at $21.66, which was sold on 2023-01-24 at $24.85.

The fund's own calendar-month averages are in the ETF profile below. May averaged 22.39% and November 13.4% over six and five observations, and September averaged -4.71%. The template does not use the calendar, and with five or six observations per month those averages say little.

Every trade

RSI mean reversion on TECL made 18 closed round trips, an average hold of 37 days, an average winner of 20.27%, an average loser of −18.72%, a profit factor of 2.82, a longest losing streak of 3. It held a position at the close on 32.1% of trading days.

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.

The 18 round trips

The first trade ran from 2021-03-05 to 2021-04-09, 35 days, from $34.47 to $44.38, a gain of 28.75%. The last ran from 2026-07-30 to 2026-08-18, 19 days, from $165.46 to $206.39, a gain of 24.74%. All trades were closed by the end of the data. Prices are adjusted for splits and dividends, so they are lower than the quotes printed at the time.

The median trade returned 13.98% and the median hold was 30 days. The shortest hold was 14 days, the longest was 86 days. The best five trades were 2026-03-30 (+49.58%), 2024-01-05 (+32.74%), 2021-03-05 (+28.75%), 2023-10-27 (+28.69%) and 2024-09-09 (+25.82%). The five worst were 2022-04-19 (-35.35%), 2025-03-04 (-26.92%), 2022-08-31 (-15.98%), 2023-08-08 (-7.74%) and 2022-01-18 (-7.63%).

Hold length separates winners from losers. The winning trades tended to last 14 to 35 days. The losing trades lasted 57 to 86 days. A buy at RSI under 30 that works snaps back within a few weeks. One that does not work keeps the position open while the fund slides, and with no stop in the template the loss runs until RSI recovers above 70. The 2022-04-19 entry is the example: it bought at $44.27, was still in on 2022-07-11 and sold at $28.62.

The record by exit year shows the same split: 3 of 3 wins in 2021, 0 of 3 in 2022, 3 of 4 in 2023, 4 of 4 in 2024, 1 of 2 in 2025 and 2 of 2 in 2026. All 5 losing trades fall in 2022, 2023 and 2025, which includes the three 2022 losers.

The longest winning streak was 5 trades and the longest losing streak was 3, the three 2022 trades. A win rate of 72% together with an average loss nearly as large as the average win means the edge rests on the win rate. The ratio of average win to average loss is close to one, so a run of losers of the 2022 kind does serious damage.

Largest drawdowns

Buy-and-hold's deepest drawdown ran from 2021-12-27 to 2022-10-12 and reached −77.2%.

Three drawdowns and the 2022 bear market

The deepest drawdown was 64.11%. It began on 2021-10-19, reached its low on 2022-10-12, 358 days later, and recovered on 2024-04-23 after 559 more days. It contains the three 2022 losing trades and the small 2023-08-08 loser. Holding TECL had a deeper one, 77.2%, from 2021-12-27 to 2022-10-12, which recovered on 2024-06-11.

The second drawdown was 48.72%, from 2025-03-05 to 2025-04-08, only 34 days. It was the entry on 2025-03-04 meeting the April 2025 sell-off. Recovery took 365 days, to 2026-04-08, and came through the 2026-03-30 trade. The third was 27.04% from 2024-07-31 to 2024-08-07 and recovered in 8 days, inside the 2024-07-26 trade.

Holding had its own second and third drawdowns of 66.26% and 46.41%. The rule's second drawdown was shallower than holding's second. The rule's maximum drawdown was 13.09 points shallower than holding's in the gap table, a result of time spent in cash. It remains a 64.11% fall in an account that was only 32.1% invested on average.

A drawdown of this size comes from the fund's volatility more than from the rule. TECL had annualized volatility of 75.11% and its own maximum drawdown was -77.96%. Its worst day was -20.4% on 2025-04-03, followed by -19.69% on 2025-04-04. The rule had no way to leave during a drop of that kind, because it checks RSI at the open and has no stop.

A reader comparing the two series should keep one fact in mind: the rule's drawdown figure is measured on the account, which sat in cash for long stretches, while the 77.2% figure for holding is measured on a position that was always in the fund.

With trading costs

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

What 5 and 10 basis points of cost did

The headline run has no fees or slippage. Adding 5 basis points per trade moved the CAGR from 23.89% to 23.49% and the end equity from $34,212 to $33,576. At 10 basis points the CAGR was 23.1% and end equity was $32,978. The maximum drawdown went from 64.11% to 64.31% and then 64.36%, and the Sharpe ratio from 0.67 to 0.666 and 0.66.

The cost sensitivity is small because the template trades 18 times in 5.74 years, with 36 fills in all. Each fill pays the cost once, and the account is in cash between trades. A weekly or daily template would pay far more. The numbers say nothing about market impact. TECL's average daily dollar volume was $122,978,604 and the median minute volume was 2,800 shares, so a $10,000 account is a very small order against that volume. A larger account would see more slippage than the 10 basis point run assumes.

Changing the parameters

What the threshold variants show

Four variants change one threshold each. Lowering the buy level to 25 cut the CAGR to 11.29%, with 12 trades and 8 wins and a 60.08% drawdown. Raising it to 35 gave 25.84% with 22 trades and 16 wins, a 64.76% drawdown and a Sharpe ratio of 0.691. A stricter entry waits for deeper washouts and trades less often, and on TECL that left the rule out of several rebounds.

The sell level mattered more. Selling above 65 gave a CAGR of 30.29%, a drawdown of 53.4% and a Sharpe ratio of 0.794, with 20 trades and 16 wins. Selling above 75 gave a CAGR of 42.61%, a drawdown of 59.91% and a Sharpe ratio of 0.936, with 17 trades and 13 wins and an end equity of $76,731. The base rule, selling above 70, sits between them at 23.89%.

The spread between the three exit levels is wide, and it is not ordered the way the thresholds are. A 65 exit and a 75 exit both beat 70 by a large margin. A smooth response would put 70 between the other two. The pattern points to the result depending on a few specific trades: a small change in the exit level shifts which days the position is closed, and on a fund moving as fast as TECL a few days can change a trade by tens of percentage points.

The variant that beat holding on CAGR was the 75 exit at 42.61% against 38.16%. It did so in a window where one sell threshold was tested after the others. A single variant that happens to win in one window is a weak basis for choosing a parameter, and nothing in this test checks the 75 exit on other data.

How TECL behaved

Why TECL suited an RSI rule

TECL is a 3x daily fund on a technology sector index. Its realized beta to XLK was 2.99 with an R-squared of 1. Its beta to SPY was 4.16 and to QQQ 3.24, and its correlation was 0.91 with SPY and 0.97 with QQQ. Against TLT the correlation was 0.07.

It returned 591.96% over the window, a CAGR of 40.06%, with annualized volatility of 75.11%. The calendar years were 124.31%, -74.28%, 202.78%, 36.37%, 38.33% and 109.96% to date. A fund that swings that much produces RSI readings under 30 and over 70, and TECL had 17 sessions under 30 and 129 over 70. The rule needs both to cross, and the asymmetry shows in the trade record: entries are scarce, exits come quickly in rallies.

The fund bounced after oversold readings. The median 5-day return after RSI(14) under 30 was 11.44% against a baseline of 1.09%, and the median 20-day return was 9.59% against 3.74%. That is the evidence a mean reversion rule needs, though 17 observations is a small sample. The 144 sessions with RSI(2) under 10 gave a median 5-day return of 1.77% and a median 20-day return of 6.58%. The faster indicator fires more often and gave a smaller bounce per signal, and the RSI(2) snapback template reached 30.2% on this fund.

The lag-1 autocorrelation of daily returns was -0.03, close to zero, so yesterday's move gave little information about today's. The price sat above its 200-day average on 68.43% of sessions and crossed that line 40 times. The daily range averaged 5.28% and the average overnight gap was 2.06%. About 59.53% of the log return came overnight and 40.47% intraday.

Leverage decay was mixed. Over the whole window the fund returned 591.96% while three times XLK's 226.13% would be 678.38%, a gap of -86.42 points. In 2023 the fund beat that comparison by 34.75 points, and in 2025 it trailed by 35.49. A mean reversion rule that holds for weeks pays this drag only while it is invested.

The rules

Buy when the 14-day RSI drops below 30 (oversold), sell when it recovers above 70 (overbought).

A standard mean-reversion setup. The Relative Strength Index measures how stretched recent price action is. Readings under 30 have historically marked short-term washouts in uptrending assets. This template buys at the next session open and holds until RSI crosses back above 70. It has no profit target and no stop.

Good for: assets that trend up over time but overshoot on the way, such as broad index ETFs.
Watch out: in a persistent downtrend, RSI can stay oversold for weeks while the position keeps losing; there is no stop-loss in this template.

How the RSI rule compares with the other templates on TECL

On TECL the ranking by CAGR was monthly cycle at 35.97%, weekly 7% target at 35.49%, RSI(2) snapback at 30.2%, dip buyer at 24.71% and RSI mean reversion at 23.89%. Below it came EMA 12/26 trend at 19.83%, SMA 10/50 trend at 17.97%, trend plus trailing stop at 14.4% and the 200-day regime filter at 11.53%.

The four templates ahead of it share a trait: they buy dips or hold through the month. The trend rules trailed it. With a drawdown of 64.11%, RSI mean reversion was not the lowest-risk option either: the weekly 7% target had a 43.24% drawdown and the RSI(2) rule had 51.05%.

Across the 59 funds, the median CAGR of this template was 2.98%, and TECL was the best of them at 23.89%. Among the leveraged funds the median was 12.08%. Peers in that category were FAS at 22.26%, SOXL at 14.43%, ROM at 12.76%, TQQQ at 12.08% and QLD at 8.78%. TMF lost 25.04% a year, and UST lost 5.04%.

TQQQ and TECL are both 3x technology funds with a correlation of 0.97 between TECL and TQQQ, and the rule's results differ by a wide margin, 23.89% against 12.08%. Entry dates and the path between entries and exits explain this, and this test cannot say which gap is signal and which is luck.

The template description lists its limits: in a persistent downtrend RSI can stay oversold for weeks while the position keeps losing, and there is no stop. The 2022 trades are that case. The page covers one window, daily-decision rules and a headline run with no fees. Past backtest results do not tell what a later period will do.

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.