3-Month Momentum Switch on XLF

Financial Select Sector SPDR Fund: S&P 500 banks, insurers and asset managers. Backtest 2021-01-04 to 2026-10-02, $10,000 starting capital, computed by the same engine that runs live DeployQuant strategies.

The 3-month momentum switch buys XLF, the Financial Select Sector SPDR Fund, after its 63-day return rises above 5%, and sells when that return falls below 0%. From 2021-01-04 to 2026-10-02, $10,000 became $11,682, a CAGR of 2.74% with a maximum drawdown of 23.03% and a Sharpe ratio of 0.3. Holding XLF over the same window gave $19,742, a CAGR of 12.58%, a drawdown of 25.47% and a Sharpe of 0.76.

The rule trailed holding in every calendar year, and the shortfall was largest in 2021 and 2025. Its drawdown was 2.44 points shallower than the fund's, which is a small benefit against a large gap in return. XLF ranks 22nd of 59 funds for this rule, and the rule ranks tenth of 12 strategies on XLF.

The account was invested 55.5% of the time. It made 15 closed round trips, of which 9 won, a win rate of 60%. Winners averaged 5.39% and losers averaged 4.84%, so the profit factor was 1.53. Two long trades did most of the work: 17.02% from 2024-07-16 to 2025-03-05 and 14.17% from 2023-12-05 to 2024-06-13. The other winners were small.

XLF holds banks, insurers and asset managers from the S&P 500. In this window it returned 102.88% with annualized volatility of 18.43%, and it fell from 2022-01-12 to 2022-10-12. A momentum filter with a three-month lookback is slow by design, and the test shows what that costs on a sector that rallies hard and drops in brief spells. More on the rule is on the 3-month momentum strategy hub, and the fund is on the XLF page. These figures are hypothetical, from one window, with no fees or slippage in the headline run.

Year by year

How each year went

2021 was the biggest miss. XLF returned 35.8% and the rule returned 7.4%. The account stayed in cash from January to the first entry on 2021-04-07 at an adjusted $31.64. That trade ended on 2021-07-20 at $32.02 for 1.2%. A short trade on 2021-08-12 lost 1.47%, and a third, from 2021-09-20 at $33.47 to 2021-11-29 at $36.06, gained 7.74%. October was the best month of the year at 7.12% and July lost 4.41%. A rule that needs a 5% three-month gain before it buys will buy after much of the move has happened, and 2021 was a year that rewarded early buying.

2022 was a down year, with negative 10.4% for the fund and negative 16.5% for the rule. It made four round trips and won one. The trade from 2022-03-23 lost 1.65% in 9 days. The August trade, from $32.71 on 2022-08-12 to $31.17 on 2022-08-29, lost 4.71%. The next trade, from $31.59 on 2022-09-13 to $28.96 on 2022-09-26, lost 8.33%, the worst of all 15. September 2022 lost 8.15% in the strategy's series. The signal had flipped to buy after a rebound in August, and the fund dropped again. A threshold rule gets caught by this kind of bounce inside a decline. The last trade of the year, from 2022-11-23, held into 2023 and made 0.27%. December 2022 lost 5.07%.

2023 returned 10.4% for the rule against 11.9% for the fund, which was nearly a match. January gained 6.67% and July gained 4.7%, while August lost 2.59% and September lost 3.02%. The rule won both of its closed trades that year, including the long one from 2023-06-13 to 2023-10-02 for 0.8%. The account was in cash in March, April, May and November.

2024 was the best year for the strategy at 22.5%, against 29.9% for holding. The December 2023 entry at $34.37 ran to 2024-06-13 at $39.24 for 14.17%, and the next trade started on 2024-07-16 at $41.60. November 2024 gained 10.17%, the strategy's best month. December 2024 lost 5.33%. The rule still trailed holding by a wide margin, because it exited on 2024-06-13 and did not buy again until 2024-07-16.

2025 returned negative 1.3% while XLF returned 14.7%. The long trade ended on 2025-03-05 at $48.68 with a gain of 17.02%, and the strategy bought again on 2025-03-25 at $49.25. That position lost 8.02% by 2025-04-04, in the week XLF's worst day of negative 7.36% fell on 2025-04-04 and the one before it lost 4.97%. April 2025 lost 6.89% for the strategy. A re-entry on 2025-06-10 gained 3.77% by 2025-10-13. 2026 is partial, ending 2026-10-02, with negative 2.3% for the rule against negative 1.1% for the fund. A January trade lost 4.88% and a trade from 2026-06-10 gained 2.6%.

Month by month

Reading the month table

The table is half empty. The zeros mark months in cash: January to March 2021, May to July 2022, October 2022, March to May 2023, November 2023, May 2025, November and December 2025, and February to May 2026. Exposure of 55.5% is the sum of all those blank months.

The best strategy months were November 2024 at 10.17%, October 2021 at 7.12% and January 2023 at 6.67%, with January 2025 at 6.34% and July 2026 at 6.05% close behind. The worst were September 2022 at negative 8.15%, September 2026 at negative 7.03% and April 2025 at negative 6.89%. Two of those three are Septembers, and September is the fund's weakest calendar month at an average of negative 3.37% over six observations. November is its strongest at 4.85% over five. With so few observations these averages describe the window, not a calendar pattern.

The fund's worst month, June 2022 at negative 10.64%, was a month the strategy sat out with a 0.0% reading. Its best month, October 2022 at 11.62%, was also missed. Cash in October 2022 was the cost of the filter: the signal needs a 5% rise over three months, and at that point the fund had only just started to turn. The rule's response to a V-shaped market is to leave near the bottom and re-enter well above it, which is what the rule's own caveat says.

Every trade

3-month momentum on XLF made 15 closed round trips, an average hold of 78 days, an average winner of 5.39%, an average loser of −4.84%, a profit factor of 1.53, a longest losing streak of 3. It held a position at the close on 55.5% 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 15 round trips

Median return was 0.8% and the median hold was 70 days. The longest hold was 232 days and the shortest was 9. The average hold of 77.9 days sits above the median because a few long trades pull it up. The longest winning streak was 4 and the longest losing streak was 3.

The five best trades are 17.02% from 2024-07-16, 14.17% from 2023-12-05, 7.74% from 2021-09-20, 3.77% from 2025-06-10 and 2.6% from 2026-06-10. The five worst are negative 8.33% from 2022-09-13, negative 8.02% from 2025-03-25, negative 4.88% from 2026-01-07, negative 4.71% from 2022-08-12 and negative 1.65% from 2022-03-23. Three of the five losers lasted 13 days or fewer. That is fast for a three-month signal. The rule had just turned positive on a rebound, bought at the open, and the rebound reversed within days.

By exit year, 2021 had 3 round trips and 2 wins, 2022 had 4 and 1, 2023 had 2 and 2, 2024 had 1 and 1, 2025 had 3 and 2, and 2026 had 2 and 1. The record in 2022 stands out. The winners came in the years when the fund trended, and the losers came where it reversed.

The final trade in the list, entered on 2026-06-10 at $51.98, closed on 2026-10-01 at $53.33 for 2.6%. The test shows no open position at the end. Prices are adjusted for splits and dividends, so they sit below the quotes printed at the time.

Exits and re-entries that cost or saved money

Several pairs of trades show what a 0% exit and a 5% entry do in practice. On 2022-08-29 the rule sold at $31.17. It bought again on 2022-09-13 at $31.59, which is a higher price, and that trade fell to $28.96 by 2022-09-26. The same sequence appeared in 2025: a sale on 2025-03-05 at $48.68, a purchase on 2025-03-25 at $49.25 and a sale on 2025-04-04 at $45.30. In both cases the exit was early enough to be right and the re-entry was late enough to be wrong.

The January 2026 trade went the other way after it closed. The rule bought on 2026-01-07 at $55.58 and sold on 2026-01-29 at $52.87 for -4.88%. It then stayed in cash from February through May and bought on 2026-06-10 at $51.98, below the January entry. The cash months of February to May 2026 show as zeros in the table. The 2021 sequence also ended with a lower re-entry: a sale at $34.94 on 2021-09-02 and a purchase at $33.47 on 2021-09-20, which returned 7.74%.

The 2023-06-13 entry at $31.32 and the 2023-10-02 exit at $31.57 held for 111 days and returned 0.8%, which is also the median return of all 15 trades. A long hold with a tiny result is common in this list: the 2022-11-23 trade ran 97 days for 0.27%, and the 2021-12-17 trade ran 34 days for 0.97%. Two trades above 14% supplied most of the profit.

Largest drawdowns

Buy-and-hold's deepest drawdown ran from 2022-01-12 to 2022-10-12 and reached −25.5%.

Drawdowns

The deepest drawdown was 23.03%, from the peak on 2022-01-12 to the low on 2022-12-19, which took 341 days, and recovery came on 2024-08-30, 620 days after the low. Holding XLF had the same peak date and a deeper drop of 25.47%, to 2022-10-12, and recovered on 2024-02-22. The strategy's equity bottomed two months later and recovered six months later. It was long into the September 2022 low and then missed part of the rebound, so a filter with the stated aim of avoiding bear markets spent a longer time underwater than holding did.

The second drawdown ran from 2025-02-18 to 2026-01-28 and reached 13.68%. It had not recovered by the end of the test. Holding had a 15.33% drop from the same peak to 2025-04-08 that recovered on 2025-06-27, so the rule's drawdown was shallower than holding's but never ended. The January 2026 trade lost 4.88% in 22 days and the low came on 2026-01-28. The third drawdown was 11.45% from 2021-06-04 to 2021-08-19, and it recovered on 2021-10-22.

The fund's own deepest fall was 25.87% in its profile and 25.47% in the buy-and-hold run, both from 2022-01-12. The fund's worst day was negative 7.36% on 2025-04-04, with negative 4.97% on 2025-04-03 and negative 4.07% on 2023-03-09 close behind. Falls of 10% or more from a 20-day high happened 10 times over 45 days. The March 2023 drops sit in a month the strategy spent in cash, which is one example where the filter did what it was designed to do.

With trading costs

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

Slippage

The rule made 30 fills. At 5 basis points per fill, CAGR fell from 2.74% to 2.5%, the drawdown deepened from 23.03% to 23.32%, the final value was $11,522 and the Sharpe was 0.276. At 10 basis points, CAGR was 2.24%, the drawdown was 23.64%, the final value was $11,355 and the Sharpe was 0.253.

The effect is modest in absolute terms because the rule makes few fills. XLF has an average daily dollar volume of $1,682,204,673 and a median minute volume of 62,901 shares. The strategy does not turn a profit that costs wipe out. The shortfall to holding comes from time in cash and late entries, and the cost runs add only a little to that.

Changing the parameters

The entry threshold

The three variants change the entry level and keep the exit at 0%. Entering above 0% returned 0.96% with a drawdown of 29.88% and a Sharpe of 0.139, over 45 trades with 22 wins. Entering above 10% returned 2.88% with a drawdown of 15.55% and a Sharpe of 0.351, over 8 trades with 4 wins. Entering above 15% returned 2.14% with a drawdown of 10.3% and a Sharpe of 0.343, over 5 trades, all winners.

The published 5% entry sits between them. A low threshold produced many trades and the worst result, because the rule flipped on small moves. Raising the bar to 10% cut the trade count and lifted the Sharpe, and at 15% the account had all winners and the shallowest drawdown but made less than at 10%. A strict filter leaves the account in cash longer and gives up the first leg of a rally. The all-winners result at 15% is five trades, a sample too small to rely on.

None of the variants approach the 12.58% for holding. The result on XLF is better described by the shape of the fund's year than by the choice of threshold. The 200-day regime filter on XLF is another slow filter and ranks last of the 12 strategies at 2.06%, so the weak result is shared by slow rules on this fund.

How XLF behaved

What XLF's data show

XLF returned 102.88% in the data, a CAGR of 13.11%, with annualized volatility of 18.43%. Calendar returns were 36.61% in 2021, negative 10.55% in 2022, 12.11% in 2023, 30.43% in 2024, 14.86% in 2025 and negative 1.15% in 2026 to date. Three of those years were over 14% and two were near 30% or above. A rule that waits for a three-month gain misses a large share of such years.

The fund spent 71.16% of sessions above its 200-day average and crossed it 38 times. It rose on 52.88% of days, and the average up day and the average down day were both 0.86%. The lag-1 autocorrelation was 0.02. The average overnight gap was 0.49% and the overnight share of return was 62.56%, with the intraday share at 37.44%. The rule buys at the open, so it collects the intraday part and takes the gap on entry.

The fund's beta to SPY was 0.87 and its correlation to SPY was 0.77. Its correlation to FAS was 1, to VTV and VOOV 0.89, and its least similar funds were SH at negative 0.78 and SDS at negative 0.77. Oversold signals worked on this fund: RSI(14) fell below 30 on 30 sessions, and over 26 of them with a forward window the median five-day return was 0.75% against a baseline of 0.42%, and the 20-day median was 3.18% against 1.5%. RSI(2) below 10 on 169 sessions gave 0.38% over five days and 1.89% over 20. Dips on XLF were bought in this window, which suits mean reversion more than momentum. The RSI mean reversion rule on XLF made 11.82% and the monthly cycle made 11.63%, both far above 2.74%.

Calendar tendencies and peers in the sector group

XLF's average return by month shows November at 4.85%, July at 4.03%, October at 3.3% and January at 2.77%, against September at -3.37% and March at -1.11%. The rule's own cash months fall in different places, so it did not line up with that calendar. Its weekday averages for the fund were 0.08% on Monday, 0.04% on Tuesday, 0.07% on Wednesday, 0.03% on Thursday and 0.07% on Friday, and none is large enough to matter for a rule that acts on a 63-day return.

Within the six sector funds, the rule made 18 round trips on XLK with a CAGR of 6.06% and a drawdown of 30.88%, 17 on XLY with -3.94% and a drawdown of 34.61%, and 13 on XLP with -1.67%. XLF's 15 trades gave 2.74% with 23.03%. XLP, the consumer staples fund, also lost money under the rule, so a sector with smaller swings did not help a momentum filter here.

The parameter runs add one more point on Sharpe. The published 5% entry scored 0.3. Entering above 10% scored 0.351 and entering above 15% scored 0.343, both a little higher, while entering above 0% scored 0.139. Holding scored 0.76. Every version of the rule on XLF stayed below the fund's own ratio, so the gap to holding is a property of the filter on this fund and not of any one threshold.

The rules

Hold while the trailing 3-month return is positive (above +5% to enter, below 0% to exit).

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 the rule and the fund fit together

The entry needs the 63-day return above 5% and the exit needs it below 0%. The gap between the two is a buffer against flip-flopping. On XLF the buffer was too narrow in 2022, when three losing trades came within a few weeks of one another, and too wide in 2021 and 2024, when it kept the account out of the early part of strong runs.

Among sector funds, SOXX returned 14.3% with a drawdown of 35.43%, VOX returned 8.92%, XLK returned 6.06%, XLF returned 2.74%, XLP returned negative 1.67% and XLY returned negative 3.94%. The median for this rule across the sector category is 6.06%, and across all 59 funds it is 0%. XLF sits below the category median and above the all-fund median. The rule worked best on the sector with the most persistent trend, and the SOXX version is the one to compare. The VOX version trades only 11 times.

On XLF the best strategies were RSI mean reversion at 11.82%, the monthly cycle at 11.63% and the RSI(2) snapback at 9.72%. The ranking puts the 3-month momentum switch tenth, ahead of only the trend plus trailing stop at 2.62% and the 200-day filter at 2.06%. The slow trend rules are at the bottom, as they are on most funds in this window.

The limits are those of a single test. The window is 5.74 years, with one clear bear market in 2022 and a sharp drop in April 2025. There are 15 trades. The headline run has no costs, and the rule decides once a day at the open. Results are hypothetical and do not forecast future behaviour.

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.