RSI(14) Mean Reversion on TBF

ProShares Short 20+ Year Treasury: -1x daily long-term Treasuries, a rising-rates position. Backtest 2021-01-04 to 2026-10-02, $10,000 starting capital, computed by the same engine that runs live DeployQuant strategies.

RSI mean reversion enters when the 14-day RSI sinks under 30 and exits when it rises over 70, with no stop and no profit target. TBF is ProShares' fund shorting 20+ year Treasuries, aiming at -1 times the daily move of long-term Treasuries, so it gains when those bonds fall. In this window the fund returned 96.88% in total, which makes it an unusual pairing: a mean-reversion rule written for assets that overshoot downward, applied to a fund that mostly drifted up.

From 2021-01-04 to 2026-10-02 a $10,000 sleeve ended at $10,537 under the rule. That is a total return of 5.37%, a CAGR of 0.92%, a maximum drawdown of 14.2% and a Sharpe ratio of 0.15. Seven of its 11 round trips closed higher (64%), and the profit factor was 1.3. A position was open on 34.4% of sessions.

Holding TBF throughout finished at $19,377, with a CAGR of 12.21%, a 17.58% drawdown and a Sharpe of 0.85. The rule trailed by 11.29 points of CAGR and kept a shallower drawdown by 3.38 points. It ranks 11th of the 12 templates on TBF. Among the 59 funds the template was run on, TBF ranks 35th, and the template's median CAGR across all 59 is 2.98%. Limits: a single 5.74-year window, no fees or slippage in the headline run, and one decision per day.

Year by year

Year by year against holding

Year by year, holding TBF gave 1.7%, 41.2%, 2.3%, 15.2%, 2.1% and 12.8% from 2021 to the 2026 partial year. The rule gave -7.8%, 4.7%, -3.4%, 5.3%, 3.4% and 3.9% in the same years. It beat holding once, in 2025, and the gaps in the other five years ran from 5.7 points in 2023 to 36.5 points in 2022.

2021 is the opening loss. The first entry was on 2021-06-03 at 15.25 and the position stayed open until 2021-10-05, selling at 14.38 for -5.7%. June lost 4.24% and July lost 3.82%. A second trade started on 2021-11-01 at 14.2 and closed on 2022-01-04 at 14.29. The year ended at -7.8% while the fund itself gained 1.7%.

2022 is where the shortfall comes from. TBF rose 42.06% in the calendar year, and the rule made 4.7%. RSI(14) fell below 30 on only 14 sessions in the whole window, and a fund climbing steadily rarely gets that oversold. The rule sat in cash from early January until 2022-11-25, when it bought at 19.18 and sold on 2022-12-28 at 19.49, a gain of 1.62%. November returned -0.1% and December 1.69%. January 2022 added 3.04% from the tail of the previous trade. Most of the year's gain went by while the rule waited for a signal.

2023 produced the best trade and the worst month. The position opened on 2023-01-19 at 18.16 and closed on 2023-02-17 at 19.48 for 7.27%. January made 0.92% and February 6.14%. Then the rule bought again on 2023-11-09 at 22.4, just before holding's own worst month of the sample, November 2023 at -8.68%. The rule lost 2.58% in November and 7.45% in December, the worst month of its run, and exited on 2024-01-18 at 21.16 for -5.54%. The year came out at -3.4% against 2.3%.

2024 returned 5.3% against 15.2%. January 2024 made 4.9% from the end of that losing trade. The long trade started on 2024-05-16 at 21.88 and ended on 2024-10-01 at 20.63 for -5.71% over 138 days, the longest hold in the sample and the worst trade. Each month from June to October was negative, from -1.08% to -2.75%. The short trade from 2024-12-05 to 2024-12-19 returned 6.47% and gave December 2024 its 6.34%, the best month of the run.

2025 is the only year the rule finished ahead, at 3.4% against 2.1%. The position from 2025-02-05 at 22.93 to 2025-05-20 at 24.09 returned 5.06% over 104 days. A second trade from 2025-08-05 at 23.45 to 2025-11-06 at 23.09 lost 1.54%. In 2026 the rule made 3.9% against 12.8%, using two trades, 2.76% from 2026-02-23 and 1.15% from 2026-06-10.

Month by month

The monthly pattern

Most months read 0 because the sleeve was in cash. The positive months were February 2023 at 6.14%, December 2024 at 6.34%, January 2024 at 4.9%, May 2025 at 4.68% and March 2026 at 3.88%. The negative months were December 2023 at -7.45%, June 2021 at -4.24%, November 2021 at -4.07%, July 2021 at -3.82% and September 2025 at -2.99%.

Winners came in runs of up to 3 and losers in runs of up to 2, so the 4 losing trades never ran together for long. The monthly table does show clustering inside single trades: the 2024 position gave five consecutive negative months from June to October, and the 2021 position gave negative months in June and July. A rule with no stop sits through those, and the exit rule, RSI above 70, was rare. TBF had only 75 sessions with RSI(14) above 70.

TBF's own month-of-year averages show April at 3.04%, September at 3.74% and October at 3.11%, with November at -3.94%. Each figure rests on 5 or 6 observations. In the rule's own results, Novembers were -4.07% in 2021, -0.1% in 2022, -2.58% in 2023 and 0.51% in 2025, which gives no usable calendar edge.

Every trade

RSI mean reversion on TBF made 11 closed round trips, an average hold of 66 days, an average winner of 3.57%, an average loser of −4.62%, a profit factor of 1.30, a longest losing streak of 2. It held a position at the close on 34.4% 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 11 trades

There are 11 closed trades and none open at the end. A typical trade made 1.15% over 64 days, and the mean hold was 65.7 days. Holds ran from 14 to 138 days. The average winner made 3.57% and the average loser lost 4.62%. Winners were more frequent and losers were larger, which is how 7 wins in 11 produce a profit factor of only 1.3.

In date order the trades were -5.7% (2021-06-03 to 2021-10-05), 0.63% (2021-11-01 to 2022-01-04), 1.62% (2022-11-25 to 2022-12-28), 7.27% (2023-01-19 to 2023-02-17), -5.54% (2023-11-09 to 2024-01-18), -5.71% (2024-05-16 to 2024-10-01), 6.47% (2024-12-05 to 2024-12-19), 5.06% (2025-02-05 to 2025-05-20), -1.54% (2025-08-05 to 2025-11-06), 2.76% (2026-02-23 to 2026-03-16) and 1.15% (2026-06-10 to 2026-07-13).

The three worst trades each lost between 5.5% and 5.71%: the 2021 entry at 15.25, the 2023 entry at 22.4 and the 2024 entry at 21.88. All three were long holds of 124, 70 and 138 days, and all three entered while RSI was below 30 on a fund that kept sliding. The template's own caveat describes this: in a persistent downtrend RSI can stay oversold for weeks while the position keeps losing. The best trades were shorter: 29 days for the 7.27% gain and 14 days for the 6.47% gain.

Grouped by the year each trade closed: 2021 had one trade and no winner, 2022 two trades and two winners, 2023 one and one, 2024 three trades and a single winner, 2025 two and one, 2026 two and two. The 2024 group is where the rule lost most ground.

Which big days the rule held through

Lining the trade dates up against the fund's largest daily moves shows how much of the profile the rule missed. TBF's two best days of 2022, 3.62% on 2022-03-02 and 3.13% on 2022-06-13, fell inside the long stretch between the exit on 2022-01-04 and the next entry on 2022-11-25, so the sleeve was in cash for both. The worst days of that year, -3.37% on 2022-09-28 and -3.84% on 2022-11-10, also fell in that gap. The same is true of -3.43% on 2023-03-10. The cash position cut both tails of the fund's distribution, and the cost was the 41.2% that buy-and-hold made in 2022.

Two large days landed inside open trades. The -3.24% day on 2022-12-01 came while the 2022-11-25 position was open, and that trade still closed at 1.62%. The best day of 2025, 3.3% on 2025-04-07, came during the 104-day trade from 2025-02-05, which returned 5.06%. A few single days of this size are the main source of variation in a trade that holds for a month or longer on a fund with a 0.93% average daily range.

Only 14 sessions closed with RSI(14) under 30, while 75 closed above 70. The entry signal is therefore the scarce event, and the exit is common once a trade is open. That fits the hold times: the three losing trades of 5.5% or more ran 70 to 138 days, which means the exit waited a long time for RSI to climb, while the winners of 7.27% and 6.47% needed 29 and 14 days.

Largest drawdowns

Buy-and-hold's deepest drawdown ran from 2023-10-19 to 2023-12-27 and reached −17.6%.

Drawdowns

The rule's deepest drawdown, 14.2%, began at the 2023-11-09 peak, bottomed on 2024-09-16 after 312 days and recovered on 2025-08-18 after another 336. It is the same stretch as the 2023 and 2024 losing trades. The second was 11.34% from 2021-06-03 to 2022-12-07, 552 days to the trough, recovering by 2023-02-10 after 65 days. The third was 6.47% from 2025-09-02 to 2025-10-22, 50 days, and it had not recovered at the end of the data.

Holding shows three falls of similar size. In the worst, 17.58%, TBF slid for 69 days from 2023-10-19 and needed until 2025-05-21 to recover, which is 511 days. Another, 15.78%, ran from 2022-10-24 to 2022-12-07 and healed on 2023-08-21 after 257 days. The last, 15.69%, ran 260 days from 2021-03-18 and healed on 2022-04-07 after 125.

The rule's maximum drawdown is only 3.38 points shallower than holding's. A rule that is in the fund a third of the time might be expected to avoid more of the damage. It was in the fund through the November 2023 to January 2024 slide, the same slide that produced holding's deepest drawdown. The rule's Sharpe ratio of 0.15 against 0.85 for holding shows the cost: it carried comparable drawdowns for much less return.

With trading costs

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

Cost runs

Across 5.74 years the rule placed 22 fills. Charged 5 basis points, it earned a CAGR of 0.73% and finished at $10,425, with a 14.29% drawdown and a 0.129 Sharpe. Charged 10 basis points, it earned 0.54% and finished at $10,313, with a 14.38% drawdown and a 0.106 Sharpe. With no cost the figure is $10,537.

Each cost step trims the CAGR a little, and the result stays positive at 10 basis points. TBF is not heavily traded. It averaged about $14,836,750 a day with a median minute volume of 810 shares, much thinner than a broad index fund. A $10,000 sleeve fits inside that, though larger sizes and wider spreads in thin minutes were not modelled. Buy-and-hold makes one fill and pays almost none of this.

Changing the parameters

Changing the thresholds

Each variant changes one threshold. An entry at 25 gave a CAGR of 0.93%, an 11.35% drawdown and 5 wins from 8 trades. An entry at 35 gave 3.77%, a 12.73% drawdown and 12 wins from 17 trades, for $12,370 and a Sharpe of 0.423. An exit at 65 gave -0.07%, a 17.25% drawdown and 7 wins from 12 trades. An exit at 75 gave 3.31%, a 13.38% drawdown and 7 wins from 10 trades, for $12,053.

The exit level moved the result most. Selling at 65 cut the winners short and ended below the starting capital, while waiting for 75 let them run, lifting the CAGR to 3.31% with 10 trades. The entry threshold mattered in the other direction. Buying at 35 admitted more signals, 17 against 11, and 12 of them won. A looser entry took advantage of a fund that dipped but rarely washed out, since RSI(14) fell below 30 on only 14 sessions. Buying at 25 is stricter than the default and made 8 trades with about the same CAGR and a shallower drawdown.

Every variant stayed well below the 12.21% buy-and-hold CAGR. The best two, at 3.77% and 3.31%, are single parameter changes picked after seeing the data, so their results describe this sample and should not be taken as the better setting.

How TBF behaved

How TBF behaved

On its own price TBF gained 96.88%, a CAGR of 12.52% at annualized volatility of 15.32%. The worst fall was 17.79%, from 2023-10-19 to 2023-12-27, and the fund regained its peak on 2025-05-21 after a drawdown stretching 396 sessions. Calendar years came in at 1.76%, 42.06%, 2.34%, 15.38%, 2.15% and, partial, 12.91%. No year was negative.

Its beta to TLT was -1 and its correlation to TLT was -1, while the correlation to SPY was -0.07 and to QQQ -0.09. Up days made 50.17% of sessions, averaging 0.8%, while down days averaged -0.74%. Typical intraday range was 0.93% and the typical overnight gap 0.55%. The overnight leg supplied 120.62% of the log return and the session leg subtracted 20.62%, so all of the gain arrived between close and next open.

The largest up days were 3.62% on 2022-03-02, 3.3% on 2025-04-07 and 3.13% on 2022-06-13. The worst days were -3.84% on 2022-11-10, -3.43% on 2023-03-10 and -3.37% on 2022-09-28. A 10% drop inside 20 days happened twice, over 7 days in all. TBF sat above its 200-day average for 78.8% of sessions and crossed it 55 times.

Oversold readings carried some information. After the 14 sessions with RSI(14) under 30, the median return was 0.85% over 5 days and 3.26% over 20, against baselines of 0.21% and 0.88%. After the 138 sessions with RSI(2) under 10, the medians were 0.59% over 5 days and 0.87% over 20, which equals the 20-day baseline. Fourteen sessions is a small sample.

The leverage table compares TBF with TLT. TLT lost 40.49% over the window while TBF gained 96.88%. Minus one times TLT's total is 40.49%, and the daily-rebalanced ideal is 46.87%, so TBF's result is well above both, with a gap of 56.39 points over the simple multiple. A daily-reset short compounds upward through a long decline, and the yearly rows show it: 2022 had TLT at -31.21% and TBF at 42.06%, against 31.21% for the simple multiple. In 2023 and 2025, when TLT rose, TBF still returned 2.34% and 2.15%.

Weekday and seasonal averages, and the inverse-fund peers

TBF's weekday averages are close to zero. Monday averaged 0.21%, Thursday 0.10%, Friday 0.07%, Tuesday -0.01% and Wednesday -0.09%, and the lag-1 autocorrelation of daily returns was -0.04. A mean-reversion rule that waits for RSI(14) under 30 is looking for a statistical rebound, and the day-to-day series gives little of one: the fund's median 5-day return after RSI(2) under 10 was 0.59% against a baseline of 0.21%, the only sign of a short-lived bounce, and it faded to 0.87% against 0.88% at 20 days.

The month-of-year averages are lopsided. September averaged 3.74%, October 3.11% and April 3.04%, while November averaged -3.94% and June -1.23%. The rule's own Septembers returned 2.79% in 2021, 0% in 2022 and 2023, -1.2% in 2024 and -2.99% in 2025, which does not follow the fund's average. These averages come from 5 or 6 observations per month, so they describe this window only.

Against other inverse funds the picture is mixed. EEV returned 5% a year with the rule and a 56.79% drawdown on 17 round trips, while TBF returned 0.92% with 14.2% and 11 round trips. TBF's drawdown was the smallest in the category on this rule, and SPDN had 28.52%, SH 29.52% and PSQ 40.92%. SPDN and SH, the inverse index funds, finished below zero. The TBF fund page shows the same profile figures for all 12 templates.

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 rule fits a rising-rates fund

The RSI template is built for assets that trend up over time and overshoot on the way, such as broad index funds. TBF is the opposite case: a fund whose price rose through the window and dipped in short episodes. Only 14 sessions were oversold, which is why the rule traded 11 times and stayed invested 34.4% of the window. Inside a long trend the entry condition seldom fires.

It ranks 11th of 12 on TBF. The leader is the monthly cycle at 11.93% with a 17.64% drawdown, followed by trend plus trailing stop at 11.25% and RSI(2) snapback at 9.67% with a 7.67% drawdown, the shallowest in the group. The weekly 7% target made 9.36%, the golden cross 8.47%, the EMA 12/26 trend 7.6% and the SMA 10/50 trend 6.68%. The 200-day regime filter returned 4.27% and the dip buyer 2.69% with an 8.18% drawdown. Only the momentum breakout did worse than RSI mean reversion, at -2.3%. Templates that stayed invested longer did better here, because holding was the winning position.

Among inverse funds, RSI mean reversion returned 5% on EEV with a 56.79% drawdown and 0.92% on TBF, the second best of the group. SPDN returned -3.97%, SH -4.26% and PSQ -5.72%. SDS returned -11.04%, QID -14.18% and REW -18.49%. The leveraged funds sat at the bottom: SQQQ at -23.58%, TECS at -29.14% and SOXS at -53.71%. The category median was -11.04%.

Eleven trades in one window on one fund is the whole sample. The CAGR is positive at 0.92%, and a different window may not repeat that.

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.