TBF trading strategies, backtested

ProShares Short 20+ Year Treasury: -1x daily long-term Treasuries, a rising-rates position. Every DeployQuant template run on TBF over 5.7 years of minute data, same engine, same window, sorted by return.

TBF is the ProShares Short 20+ Year Treasury, a fund that aims for the opposite of the daily return of long-dated US Treasuries. It gains when long bond prices fall. Each of the 12 DeployQuant templates started with $10,000 on 2021-01-04 and ran to 2026-10-02. Buying and holding TBF returned 12.21% a year, ended at $19,377, and had a 17.58% max drawdown. The Sharpe was 0.85.

None of the 12 templates beat holding on CAGR. Eleven had a positive CAGR and 7 had a shallower drawdown than holding. The closest are the monthly cycle at 11.93% and the trend plus trailing stop rule at 11.25%. The RSI(2) snapback made 9.67% with a 7.67% drawdown and the highest Sharpe in the table at 1.02, and it is the template that gives up the least risk.

The fund is unusual in this dataset. It is an inverse fund that went up. Long Treasuries fell through most of the window, so a short position earned more than 12% in each of 2022, 2024 and 2026, and a fund that earns by a persistent trend rewards the rule that stays in. The templates that were flat for long stretches gave up return, and the table shows by how much.

All figures are one backtest with daily-decision rules. The headline run has no fees or slippage, and two cost runs add 5 and 10 basis points. The numbers describe this fund in this window and do not forecast a later one.

Why holding beat every rule on TBF

The fund's calendar years were 1.76% in 2021, 42.06% in 2022, 2.34% in 2023, 15.38% in 2024, 2.15% in 2025 and 12.91% in 2026 to date. Most of the return came from 2022. Any rule that was not in the fund for that year had a hard time catching up, and any rule that was in it throughout earned about what holding earned.

The two templates that stayed invested most of the time are the closest to holding. The monthly cycle was invested 95.2% of the time and made 11.93% a year with a 17.64% drawdown. It returned 42% in 2022. The weekly 7% target was invested 88.3% of the time, and its return of 9.36% includes 29.6% in 2022. The trend plus trailing stop rule was invested 81.8% of the time and made 11.25% with only 2 round trips. Both trades won, with an average win of 19.14% and an average hold of 347 days. It returned 0% in 2021 because it was flat that year, and 34.7% in 2022 against 41.2% for holding.

The golden cross made 8.47% with 4 round trips, 3 of them wins, and a profit factor of 17.62. Its average win was 13.88% against an average loss of 1.83%. It trailed holding in every one of the six calendar years, as the empty list of years beaten shows. It returned 33.6% in 2022, which is below the 41.2% for holding. A trend rule that waits for the 50-day average to cross the 200-day average enters after the move has begun.

The SMA 10/50 trend rule made 6.68% with a 13.88% drawdown, and the EMA 12/26 trend rule made 7.6% with a 17.96% drawdown. The SMA rule lost 8.6% in 2021 and 3.6% in 2025, and beat holding only in 2023 by 14 points, with 16.3% against 2.3%. Both rules show the same pattern, an early start to the 2022 gain that was partly given back by late exits.

The 200-day regime filter made 4.27% with a 23.97% drawdown, the deepest of the 12 templates. It lost 11.6% in 2023. That drawdown began on 2022-10-24 and had not recovered by the end of the window. A filter that needs the fund above its 200-day average to hold it exits during the pullbacks that this fund had in 2022 and 2023, and 55 crossings of that average made it flip often.

The mean reversion and dip rules on a trending fund

TBF closed above its 200-day average on 78.8% of sessions. A fund that spends that much of the time above a slow average is one where a rule that buys weakness has few chances. The 14-day RSI closed under 30 on only 14 sessions, and the 2-day RSI closed under 10 on 138. The median 5-day forward return after the 14-day signal was 0.85% against a baseline of 0.21%, and the median 20-day return was 3.26% against 0.88%. Only 14 observations stand behind those figures.

The RSI mean reversion template made 0.92% a year. It made 11 round trips, won 64%, and was invested 34.4% of the time. The average win of 3.57% was smaller than its average loss of 4.62%, and its max drawdown was 14.2%. The 2021 return was negative 7.8%. A rule that enters at RSI under 30 and waits for RSI above 70 sat out most of the trend, and the exit at 70 on a fund that closed above 70 on 75 sessions came before the move was over.

The RSI(2) snapback works differently. It made 170 round trips, won 59%, and was invested 38.5% of the time. The average win of 1.19% was above the average loss of 0.93%, giving a profit factor of 1.85. It was positive in every calendar year: 7.5% in 2021, 11.1% in 2022, 15.4% in 2023, 4.9% in 2024, 7.5% in 2025 and 9.3% in 2026. Its max drawdown of 7.67% was the shallowest of the templates that traded often. Holding returned 41.2% in 2022 and the snapback returned 11.1%, which is the largest shortfall of any year.

The cost runs matter for this rule. At 5 basis points its CAGR falls to 6.54%, with end equity of $14,387. At 10 basis points it falls to 3.49% and $12,179, with a Sharpe of 0.406. TBF's average daily dollar volume was $14,836,750 and the median minute volume was 810 shares, which is thin next to the broad funds. The 10 basis point run is the more cautious case for a fund with that volume.

The dip buyer made 2.69% with 2 round trips, both wins, and a drawdown of 8.18%. It was invested only 13.8% of the time. The 3-month momentum rule made 1.17% with a 23.14% drawdown and a 33% win rate. The momentum breakout rule is the only one with a negative CAGR, at negative 2.3%, with 4 round trips, a 25% win rate and a profit factor of 0.03. Its average win was 0.42% and its average loss was 4.57%.

How each strategy traded TBF

Trade statistics and cost sensitivity

The trade table groups the templates by how long they hold. The golden cross held its average trade 286.5 days, the trend and trailing-stop rule 347 days, and the dip buyer 148 days. The RSI(2) snapback held 4.9 days, the weekly target 21.4, and the monthly cycle 28.4. Fewer, longer trades mean less cost exposure. The slow rules lose almost nothing to the cost runs: the trend and trailing-stop rule goes from 11.25% to 11.16% at 10 basis points, and the golden cross from 8.47% to 8.31%.

The fast rules lose much more. The weekly target goes from 9.36% to 7.86% at 5 basis points and 4.63% at 10, and its max drawdown deepens from 19.49% to 22.84%. The snapback loses more still, as described above. The monthly cycle goes from 11.93% to 10.62% and 9.32%. For the monthly cycle, the 10 basis point run still leaves it below holding at 12.21%.

The weekly 7% target shows a pattern that is the reverse of what its design suggests. It won only 24% of its 82 round trips and had a longest loss streak of 11 trades. The average win was 7.17% and the average loss was 1.37%, giving a profit factor of 1.61. The 7% target is large for a fund with an average intraday range of 0.93%, so the target was reached rarely, and the Thursday exit cut most of the trades at a small loss. Its average hold of 21.4 days shows that trades that were not losing on Thursday carried on.

The trade-by-trade picture also shows how small the losses were on this fund. The worst trade for the snapback was negative 4.9%, for the weekly target negative 4.48%, and for the trend and trailing-stop rule there was no losing trade. Compared with the leveraged funds such as SQQQ, where single trades lost far more, TBF is a quiet fund with a 15.32% annualized volatility.

How TBF behaved

Calendar years

Biggest single days

Average return by calendar month

Most and least correlated funds

Leverage and decay against TLT

The last column compounds -1× TLT'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 -1× the underlying's return over a year.

Over 2021-01-04 to 2026-10-02, TBF returned 96.9% while TLT returned −40.5% and a perfect daily-reset -1× version would have returned 46.9%. Its measured daily beta to TLT was -1.00.

What TBF did in the window

TBF's price series returned 96.88% over the window, or 12.52% a year, with annualized volatility of 15.32%. Its max drawdown was 17.79%, from the peak on 2023-10-19 to the trough on 2023-12-27, and it recovered on 2025-05-21. The longest time under a prior high was 396 sessions. Two earlier drawdowns in the buy-and-hold record were 15.78% from 2022-10-24 to 2022-12-07 and 15.69% from 2021-03-18 to 2021-12-03.

The daily moves are modest. TBF rose on 50.17% of sessions, with an average up day of 0.8% and an average down day of negative 0.74%. The best days were 3.62% on 2022-03-02, 3.3% on 2025-04-07 and 3.13% on 2022-06-13. The worst were negative 3.84% on 2022-11-10, negative 3.43% on 2023-03-10 and negative 3.37% on 2022-09-28. There were 2 drawdown events of 10% or more inside 20 days, spread over 7 sessions, compared with 85 on SQQQ.

The overnight and intraday split is lopsided. Overnight log return was 80.91% and intraday was negative 13.83%, which gives an overnight share of 120.62% and an intraday share of negative 20.62%. The fund earned all of its return between the close and the next open and lost a little during the trading day. The average overnight gap was 0.55% and the average intraday range was 0.93%. A rule that buys at the open and sells at the close would have lost on this fund over the window, and the templates here hold across nights.

First-order autocorrelation was negative 0.04. TBF closed above its 200-day average on 78.8% of sessions and crossed it 55 times, more crossings than TQQQ or SQQQ had. That count helps explain why the slower trend filters flipped so often on a fund with such low volatility.

Inverse exposure to long Treasuries

TBF's measured daily beta to TLT was negative 1, with an r-squared of 0.99 and a correlation of negative 1. TLT returned negative 40.49% over the window. A frictionless daily-reset −1× version of TLT would have returned 46.87%, and TBF returned 96.88%, well above that line. The leverage table leaves out fees, borrowing costs and interest on cash, which put a real fund below or above the ideal, and the facts do not split the difference into those parts.

The year-by-year table shows where. In 2022 TLT fell 31.21% and TBF rose 42.06%, against 39.41% for the ideal. In 2024 TLT fell 8.07% and TBF rose 15.38%, against 6.65%. In 2026 TLT fell 8.01% and TBF rose 12.91%, against 7.95%. In 2023 TLT rose 2.79% and TBF still rose 2.34%, against negative 5.9% for the ideal, and in 2025 TLT rose 4.23% and TBF rose 2.15%, against negative 5.41%. In years where long bonds rose modestly, the fund did not lose. In 2021 TBF returned 1.76% against 2.8% for the ideal, the one year it finished below it.

Some of this comes from the path. A daily-resetting inverse fund compounds gains when the underlying falls steadily, and TLT fell for most of the window. The facts do not separate path effects from other fund-level effects, so the table is a description of the window. It is not a model of what TBF should return.

The practical consequence is that TBF behaves like a rates trade rather than an equity hedge. Its correlation to SPY was negative 0.07 and to QQQ negative 0.09, and its beta to SPY was negative 0.07. The funds most correlated with it are RINF at 0.41, KMLM at 0.4 and CTA at 0.28. It does not move with the Nasdaq or the S&P 500, so the templates that work on SQQQ or QID for equity reasons say nothing about TBF.

Calendar months and the other inverse funds

The average return by calendar month was highest in September at 3.74%, October at 3.11% and April at 3.04%. It was lowest in November at negative 3.94% and June at negative 1.23%. Each month has 5 or 6 observations, so a single year such as 2022 drives each average. The data describes the window and does not give a calendar rule.

Among the other inverse funds, TBF is the one where holding made money. On QID the best template, the RSI(2) snapback, made negative 1.85% against negative 29.46% for holding. On PSQ it made 0.13% against negative 13.76%. On SH it made negative 0.53% against negative 10%, on SPDN negative 1.45% against negative 9.68%, and on SDS the golden cross made negative 2.33% against negative 21.57%. On SOXS the trend plus trailing stop rule made negative 4.93% against negative 48.27%, and on REW the golden cross made negative 6.71% against negative 36.12%.

On those funds, the template choice was about losing less. On TBF the template choice was about how much of the buy-and-hold return to keep, and the best of them kept most of it. The EEV fund, where the RSI mean reversion template made 5% against negative 16.07%, is the one other case with a positive best result.

The test stops at one window in which long bonds fell hard. A window with rising bond prices would reverse the fund's returns, and the templates would face a decaying fund of the kind shown on the SQQQ page. The results here describe how the rules handled a persistent trend in the fund's favour.

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.