DeployQuant3-Month Momentum Switch on TLT
iShares 20+ Year Treasury Bond ETF: long-duration Treasuries, one of the most rate-sensitive bond ETFs. Backtest 2021-01-04 to 2026-10-02, $10,000 starting capital, computed by the same engine that runs live DeployQuant strategies.
TLT holds long-dated US Treasuries, and over this window it lost money. The 3-month momentum switch bought it only when the 63-day return was above +5% and sold when that return dropped below 0%. Between 2021-01-04 and 2026-10-02 the rule turned $10,000 into $8,760.14, a CAGR of -2.28% with a Sharpe ratio of -0.33 and a max drawdown of 18.58%.
Buy-and-hold on TLT did worse in every summary measure: $6,106.38 at the end, a CAGR of -8.23%, a Sharpe of -0.52 and a max drawdown of 42.77%. The rule lost money too, so the comparison is between a smaller loss and a larger one. It ended $2,653.76 ahead of holding and beat it in four of the six calendar years, 2021, 2022, 2024 and 2026.
Trading produced 7 round trips and 14 fills, of which 2 won (29%), for a profit factor of 0.16. The average winner was 1.24% and the average loser was -3.13%. The account was invested 23.1% of the time and in cash the rest.
The rule acted as a loss limiter on this fund and produced no gain. TLT fell 40.49% over the window, and the rule's only job was to be somewhere else for most of it. Among the 12 templates it placed 5th on TLT, and TLT placed 41st of 59 funds for the rule.
Year by year
| Year | 3-month momentum | buy & hold |
|---|---|---|
| 2021 | 2.2% | −4.4% |
| 2022 | 0.0% | −30.2% |
| 2023 | −3.8% | 2.7% |
| 2024 | −5.1% | −7.7% |
| 2025 | −6.1% | 4.0% |
| 2026 | 0.0% | −7.7% |
Year by year
The yearly table shows the rule ahead of holding in 2021, 2022, 2024 and 2026, and behind in 2023 and 2025.
In 2021 the rule returned 2.2% and holding returned -4.4%. January to May show 0.0%: TLT was falling over the first half of the year, so the 63-day return never cleared +5%. The rule bought on 2021-06-14 at an adjusted $118.18, after the fund had risen. June added 1.53% and July added 3.65%, the best month of the run. August lost 0.29% and September lost 2.88%, and the rule sold on 2021-10-05 at $120.92 for a gain of 2.32%. It was the best trade of the whole run and came in the first year.
2022 is the year the rule did its work. It returned 0% because it never held a position, while TLT lost 30.2%. Buy-and-hold fell 9.19% in April alone. The 63-day return was negative or below +5% for the whole year, so every monthly figure is 0.0%. The gap of 30.2 points is the largest in the table and explains most of the final gap between the rule and holding.
2023 is where the rule lost ground. It returned -3.8% against 2.7% for holding. The rule bought on 2023-01-10 at $89.85 and sold on 2023-02-28 at $86.56, a loss of 3.66%. January added 2.46% and February lost 5.89%, the worst month of the run. It bought again on 2023-03-31 at $90.95 and sold on 2023-04-17 at $90.41, losing 0.59% over 17 days. TLT then went on to a strong autumn, and holding's best month was November 2023 at 9.49%. The rule sat in cash from May to November and bought on 2023-12-15 at $87.60, after most of that rally.
2024 was a losing year for both, with the rule at -5.1% and holding at -7.7%. The December 2023 entry ran until 2024-03-15 and lost 5.09% over 91 days, with January at -2.21%, February at -2.2% and March at -0.96%. The next trade ran from 2024-07-15 at $84.31 to 2024-10-28 at $84.45, a gain of 0.17%. July, August and September added 1.9%, 2.09% and 1.86%, then October gave back 5.48%.
2025 went against the rule. It returned -6.1% and holding returned 4%, a gap of 10.1 points. The trade that did the damage began on 2025-04-02 at $86.31 and ended on 2025-04-22 at $81.15, a loss of 5.98% over 20 days. It was the worst trade of the run. April lost 5.81%. The rule bought a second time on 2025-09-09 at $85.16 and sold on 2025-12-08 at $84.87 for a small loss of 0.34%.
2026 so far shows 0.0% for every month, because the rule has had no position since December 2025. Holding lost 7.7% in 2026 to date, so the rule is 7.7 points ahead for the year.
Month by month
| Year | Jan | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep | Oct | Nov | Dec |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2021 | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 1.5% | 3.6% | −0.3% | −2.9% | 0.3% | 0.0% | 0.0% |
| 2022 | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% |
| 2023 | 2.5% | −5.9% | 0.8% | −1.4% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.3% |
| 2024 | −2.2% | −2.2% | −1.0% | 0.0% | 0.0% | 0.0% | 1.9% | 2.1% | 1.9% | −5.5% | 0.0% | 0.0% |
| 2025 | 0.0% | 0.0% | 0.0% | −5.8% | 0.0% | 0.0% | 0.0% | 0.0% | −0.1% | 1.4% | 0.2% | −1.8% |
| 2026 | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | – | – |
Months in and out
Most months are 0.0% because the rule was in cash, which fits the 23.1% exposure. Of the months with exposure the best was July 2021 at 3.65% and the worst was February 2023 at -5.89%. The other large losers were October 2024 at -5.48% and April 2025 at -5.81%. Each of those was a single month in which TLT dropped sharply right after the rule had been buying.
Buy-and-hold's worst month was April 2022 at -9.19%, a month the rule was not in, and its best was November 2023 at 9.49%, which the rule missed. By calendar month across the fund's history the average was 4.43% in November and -3.16% in September, -2.63% in October and -2.55% in April, each from five or six years. Those are small samples and the rule does not use the calendar.
Every trade
3-month momentum on TLT made 7 closed round trips, an average hold of 69 days, an average winner of 1.24%, an average loser of −3.13%, a profit factor of 0.16, a longest losing streak of 3. It held a position at the close on 23.1% of trading days.
| Entry | Entry price | Exit | Exit price | Return | Days held |
|---|---|---|---|---|---|
| 2021-06-14 | $118.18 | 2021-10-05 | $120.92 | 2.3% | 113 |
| 2023-01-10 | $89.85 | 2023-02-28 | $86.56 | −3.7% | 49 |
| 2023-03-31 | $90.95 | 2023-04-17 | $90.41 | −0.6% | 17 |
| 2023-12-15 | $87.60 | 2024-03-15 | $83.14 | −5.1% | 91 |
| 2024-07-15 | $84.31 | 2024-10-28 | $84.45 | 0.2% | 105 |
| 2025-04-02 | $86.31 | 2025-04-22 | $81.15 | −6.0% | 20 |
| 2025-09-09 | $85.16 | 2025-12-08 | $84.87 | −0.3% | 90 |
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.
All seven trades
In order, with adjusted prices:
- 2021-06-14 to 2021-10-05, $118.18 to $120.92, +2.32% in 113 days.
- 2023-01-10 to 2023-02-28, $89.85 to $86.56, -3.66% in 49 days.
- 2023-03-31 to 2023-04-17, $90.95 to $90.41, -0.59% in 17 days.
- 2023-12-15 to 2024-03-15, $87.60 to $83.14, -5.09% in 91 days.
- 2024-07-15 to 2024-10-28, $84.31 to $84.45, +0.17% in 105 days.
- 2025-04-02 to 2025-04-22, $86.31 to $81.15, -5.98% in 20 days.
- 2025-09-09 to 2025-12-08, $85.16 to $84.87, -0.34% in 90 days.
Only the 2.32% trade and the 0.17% trade closed above their entry price. The other five lost, and the median trade lost 0.59%. Holds averaged 69.3 days, from 17 at the shortest to 113 at the longest. The longest losing streak was 3, which ran from the first 2023 trade through the March 2023 trade and into December 2023, and the longest winning streak was 1.
The entries show what the +5% threshold does on a bond fund in a rate-driven decline. TLT made lower highs and lower lows through most of the window, and the three-month return crossed +5% in a few short rallies. The rule bought each of them, and five of the seven did not continue.
The exit trigger of 0% worked slowly. The 2023-12-15 position lost 5.09% over 91 days because the 63-day return stayed above 0% for most of that time while the price drifted lower. The 2025-04-02 position was the reverse, a fast decline of 5.98% in 20 days, and it exited close to the low. The drawdown table shows that low on 2025-04-21.
The April 2025 trade next to TLT's worst days
The 20-day trade of April 2025 is the one trade where the daily data of the fund explains the loss. The rule bought on 2025-04-02 at an adjusted 86.31. TLT's worst days of the window included -3.05% on 2025-04-07 and -2.72% on 2025-04-10, and both fell inside that holding period. The rule sold on 2025-04-22 at 81.15 and the drawdown table puts the account's low one day earlier, on 2025-04-21. The signal that sold the position, a 63-day return under 0%, only turned negative after most of the fall had happened.
The March 2023 entry shows the opposite problem. The fund's best day of 2023 was 3.42% on 2023-03-10, three weeks before the rule bought again on 2023-03-31 at 90.95. By then the move was in the price, and the position gave back 0.59% by 2023-04-17. The same lag shows up in autumn 2022. TLT's best days of the whole window, 3.9% on 2022-11-10 and 3.24% on 2022-12-01, came while the rule held nothing, and the 2022 row of the monthly table is zero in every column.
Only 3 episodes of a 10% drop from a 20-day high occurred in the fund, covering 5 days in total. The fund's losses came as a slow slide with small daily steps, which a 63-day lookback has no trouble seeing, and the problem was the opposite one: the rallies inside the slide were frequent and short, so the +5% entry kept buying them.
Largest drawdowns
| Peak | Low point | Depth | Days to low | Recovered | Days to recover |
|---|---|---|---|---|---|
| 2021-09-22 | 2025-04-21 | −18.6% | 1307 | not yet | – |
| 2021-07-19 | 2021-08-12 | −3.2% | 24 | 2021-09-14 | 33 |
| 2021-06-18 | 2021-06-25 | −2.4% | 7 | 2021-07-06 | 11 |
Buy-and-hold's deepest drawdown ran from 2021-01-04 to 2023-10-19 and reached −42.8%.
Drawdowns
The rule's largest drawdown was 18.58%. The peak was 2021-09-22 and the trough was 2025-04-21, 1,307 days later, and the account had not recovered by 2026-10-02. This is a drawdown measured against the account's highest value, and the account made its high in 2021 and never got back to it. Most of those 1,307 days were spent in cash, so the depth came from a handful of losing trades and not from time in the market.
Holding's only listed drawdown was 42.77%, from 2021-01-04 to 2023-10-19, 1,018 days, and it also had not recovered. The fund's own maximum drawdown was 44.06% over the same dates, and its longest time under water was 1,443 sessions, which is essentially the whole window.
The rule's second drawdown was 3.19%, from 2021-07-19 to 2021-08-12, recovering on 2021-09-14. The third was 2.44%, from 2021-06-18 to 2021-06-25, recovering on 2021-07-06. Both come from the first trade. The rule's deep drawdown is the 18.58% figure, and everything after it was a continuation of the same slide, with each new trade losing a bit more.
The comparison with holding is therefore lopsided in depth and neutral in direction. The rule's max drawdown was 18.58% against 42.77%, and both are losses of capital over the window.
With trading costs
The headline run fills at the bar price. These runs charge slippage on every fill.
| Slippage per fill | CAGR | Max drawdown | Final value | Sharpe |
|---|---|---|---|---|
| None (headline) | −2.3% | −18.6% | $8,760 | -0.33 |
| 5 basis points | −2.4% | −19.0% | $8,695 | -0.35 |
| 10 basis points | −2.5% | −19.5% | $8,637 | -0.36 |
Cost runs
With 14 fills the rule paid for slippage twice per round trip. At 5 basis points the CAGR was -2.41%, the max drawdown 19.03%, the final value $8,695.46 and the Sharpe -0.347. At 10 basis points the CAGR was -2.52%, the drawdown 19.46%, the final value $8,637.13 and the Sharpe -0.365. The headline run has no fees or slippage.
Liquidity is no concern for TLT, which traded $2,159,787,473 on an average day, with 35,636 shares in the median minute. A $10,000 order is tiny next to that. Each cost step trimmed the final value a little, and the loss on the account comes from price, with costs a minor addition.
Changing the parameters
| Version | CAGR | Max drawdown | Round trips | Win rate | Final value |
|---|---|---|---|---|---|
| Published rules | −2.3% | −18.6% | 7 | 29% | $8,760 |
| Enter above 0% | −5.8% | −35.0% | 52 | 25% | $7,081 |
| Enter above 10% | −4.0% | −22.0% | 4 | 0% | $7,901 |
| Enter above 15% | −2.2% | −13.9% | 2 | 0% | $8,804 |
Changing the entry threshold
Three variants moved the entry level and left the 0% exit alone.
- Entering above 0% returned -5.83% with a max drawdown of 35.01% and a Sharpe of -0.664. It made 52 trades and 13 won.
- Entering above 10% returned -4.02% with a max drawdown of 21.97% and a Sharpe of -0.874. It made 4 trades and none won.
- Entering above 15% returned -2.19% with a max drawdown of 13.92% and a Sharpe of -0.675. It made 2 trades and none won.
The base rule with a +5% entry returned -2.28%. The 0% entry is the clearest result. With no buffer between entry and exit, the rule flipped 52 times, returned -5.83% and drew down 35.01%. That shows what the +5% entry gap in the base rule buys: it filters out most of the whipsaw. The higher entries did not help much. The 15% entry returned -2.19%, a hair above the base, with a shallower drawdown of 13.92% because it only made 2 trades. The 10% entry sat between them at -4.02%, and all 4 of its trades lost money.
None of the four entry levels produced a positive return. The result of the rule on TLT is therefore not a question of tuning. Every setting tested lost money, and the settings differ in how much of the decline they avoided. The sample of trades is small for the higher entries, and a result built on 2 or 4 trades has little statistical weight.
How TLT behaved
| Measure | TLT |
|---|---|
| Data in this test | 2021-01-04 to 2026-10-02 (1444 sessions) |
| Total return, buy and hold | −40.5% |
| Annualized volatility | 15.3% |
| Deepest drawdown | −44.1% (2021-01-04 to 2023-10-19) |
| Up days | 49.0% |
| Average daily range | 0.94% |
| Average overnight gap | 0.54% |
| Correlation to SPY | 0.08 |
| Correlation to QQQ | 0.09 |
| Sessions above the 200-day average | 35.2% |
| Crossings of the 200-day average | 95 |
| Falls of 10% or more from a 20-day high | 3 |
How TLT behaved
Long-dated Treasuries, held through the iShares 20+ Year Treasury Bond ETF, are among the most rate-sensitive bond exposures. TLT lost 40.49% from 2021-01-04 to 2026-10-02, a CAGR of -8.64% at 15.33% annualized volatility. Its worst drawdown was 44.06% and ran from the opening day to 2023-10-19, with no recovery since. Calendar returns were -4.53%, -31.21%, 2.79%, -8.07% and 4.23% for 2021 to 2025, then -8.01% in 2026 to date.
Equities explain almost none of its moves. Beta was 0.07 to SPY and 0.06 to QQQ, at correlations of 0.08 and 0.09. TMF, the leveraged version, was perfectly correlated at 1.0. BND and AGG followed at 0.92, IEF at 0.91 and UST at 0.86. The inverse fund TBF showed -1.0, and the lowest others were RINF at -0.4, KMLM at -0.39, CTA at -0.28 and USDU at -0.24.
Trend persistence is what a momentum rule needs, and TLT offered little. It closed above its 200-day average on only 35.18% of sessions, and it flipped across that line 95 times. For a bond fund that is a lot of failed recoveries during a decline. A 63-day return above +5% in this data usually marks a short rally inside the slide.
The fund rose on 49% of sessions. Average up and down days were 0.74% and -0.77%, the daily range averaged 0.94%, and lag-1 autocorrelation was -0.05. The largest gain was 3.9% on 2022-11-10 and the largest loss was -3.47% on 2022-03-02, both in a year when the rule held nothing.
Most of the damage arrived overnight. Overnight log return was -70.11% against 18.78% intraday. The rule buys at the open, so each entry takes whatever gap the fund printed, and on balance that gap has pointed down.
Oversold readings gave no bounce. After RSI(14) closed under 30 on 61 sessions, the median 5-day return was 0.02% against -0.14% for all days, and the 20-day median was -1.63% against -0.6%. RSI(2) under 10 on 203 sessions was followed by medians of 0.03% over 5 days and -0.65% over 20. That fits the RSI mean reversion rule returning -5.4% a year here.
Calendar and weekday averages for TLT
The month-of-year table has a clear shape for this window, and it is a small sample. November averaged 4.43% and June 1.51%, while September averaged -3.16%, October -2.63%, April -2.55% and March -1.74%. The rule bought in April 2025 and September 2025, two of the weaker calendar months, and in October 2024, which lost 5.48% for the account. Its best month, July 2021 at 3.65%, was a calendar month with a 0.27% average for the fund, so the rule's results do not line up with the seasonal pattern.
Weekday averages were small. Monday averaged -0.19%, Wednesday 0.12%, Thursday -0.08%, Friday -0.05% and Tuesday 0.02%. None is large enough to matter against a daily range of 0.94%.
Among the other bond funds, IEF had a maximum drawdown of 9.07% under this rule, BND 6.55% and AGG 5.72%, against 18.58% for TLT. Duration is the difference between those funds and TLT, and the rule's drawdown follows it. IGIB made 4 round trips and TLT made 7, so the longer-duration fund also generated more false signals. Under the same rule, SGOV made no trades and showed 0% with no drawdown.
The rules
Hold while the trailing 3-month return is positive (above +5% to enter, below 0% to exit).
- WHEN the market opens · IF not invested AND the 63-day return > +5% · THEN buy with 98% of the sleeve
- WHEN the market opens · IF invested AND the 63-day return < 0% · THEN sell the whole position
Time-series momentum on a quarterly lookback, the horizon much of the academic momentum literature uses. The template enters after a +5% three-month run and exits when the same measure turns negative. The gap between entry (+5%) and exit (0%) is a buffer against flip-flopping around a single threshold.
Good for: assets with long, persistent cycles, such as index, sector and managed-futures ETFs.
Watch out: a three-month lookback is slow; V-shaped crashes and recoveries can see it exit near the bottom and re-enter well off the low.
How 3-month momentum compares with the other templates on TLT
Template results on TLT, ordered by CAGR with the max drawdown after it:
- RSI(2) snapback: 0.96%, 22.81%
- EMA 12/26 trend: -0.67%, 15.92%
- SMA 10/50 trend: -0.85%, 14.24%
- Momentum breakout: -1.2%, 10.32%
- 3-month momentum: -2.28%, 18.58%
- Golden cross: -2.46%, 20.58%
- Weekly 7% target: -3.78%, 30.96%
- Dip buyer: -4.32%, 27.81%
- RSI mean reversion: -5.4%, 35.95%
- Trend plus trailing stop: -7.11%, 38.76%
- 200-day regime filter: -7.58%, 40.21%
- Monthly cycle: -8.08%, 46.24%
Only the snapback made money. The 3-month rule sits in the middle at 5th, and the bottom four also carried the deepest drawdowns. It stayed in cash through most of the window, and that explains its place better than any skill in picking entries.
Other bond funds fared better under the same rule. IGIB earned 0.25%. SGOV earned 0% with no trades. IEI lost 0.59%, AGG 0.74%, BND 0.98% and IEF 1.35%. TLT finished last of the group, with the deepest drawdown and the most round trips. Over all 59 funds the middle result for this rule was 0%, and across the bond funds it was -0.74%. Shorter-duration funds declined less, which gave the rule fewer false starts. Every template tested on this fund is listed on the TLT fund page.
Where the test stops
Seven trades on one bond fund over 5.74 years form a small sample. The headline run charges nothing per fill. All of it is hypothetical and none of it is investment advice.
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
Did 3-month momentum beat buy-and-hold on TLT?
Over 2021-01-04 to 2026-10-02, 3-month momentum on TLT returned −2.3% annualized vs −8.2% for buy-and-hold: it beat buy-and-hold by 6.0% per year, with a maximum drawdown 24.2 points shallower than holding (18.6% vs 42.8%).
How many trades did it make?
7 completed round trips over 5.7 years (14 fills), with 29% of round trips closing profitably.
Why 63 days?
63 trading days is about one quarter, a common momentum lookback. You can sweep it in DeployQuant to see how the horizon changes results.
Did 3-month momentum beat buy-and-hold on TLT?
Both lost money, and the rule lost less. It returned -2.28% a year to holding's -8.23%, ending at $8,760.14 against $6,106.38. The rule's max drawdown was 18.58% against 42.77% for holding.
Why did the rule make nothing in 2022?
TLT's 63-day return never cleared the +5% entry level that year, so no order went in. Holding lost 30.2% in 2022, and that gap is the widest in the yearly table.
How many trades did the rule make and how many won?
Seven round trips and 14 fills produced 2 winners, a 29% hit rate. The best trade gained 2.32% over 113 days and the worst lost 5.98% over 20 days.
What happens with a different entry level?
Entering above 0% returned -5.83% over 52 trades. Entering above 10% returned -4.02% and above 15% returned -2.19%. None of the entry levels tested made a profit on TLT.
How much do costs change the result?
A charge of 5 basis points a fill gave -2.41% and 10 basis points gave -2.52%. Costs added a small loss to a result that falling prices had already set.
Is the 18.58% drawdown still open?
Yes. The drawdown peak was 2021-09-22, the trough was 2025-04-21, and the account had not regained that peak by the end of the data. Most of the time was spent in cash.
Which template did best on TLT?
The [RSI(2) snapback](/learn/strategies/rsi2-dip-snapback/tlt/) was the only one with a positive CAGR, at 0.96%. The 3-month momentum switch was 5th at -2.28%, and TLT ranked 41st of 59 funds for it.
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.