RSI(14) Mean Reversion on SPY

SPDR S&P 500 ETF Trust: the most-traded S&P 500 ETF and a common benchmark for US large caps. Backtest 2021-01-04 to 2026-10-02, $10,000 starting capital, computed by the same engine that runs live DeployQuant strategies.

This is the standard 30/70 RSI(14) rule run on SPY from 2021-01-04 to 2026-10-02. It turned $10,000 into $12,780, a 4.37% CAGR, against $21,820 and 14.56% for holding the fund. The win rate was 77% on 13 round trips, the profit factor was 2.62, and the max drawdown was 17.09% against 23.99% for holding. It was invested on 26.8% of trading days.

The result is a good hit rate and a poor return. Ten of the 13 trades made money, and the rule still ranks 11th of 12 templates on SPY and 26th of 59 funds for this strategy. The reason is in the size of the trades against the time out of the market. The rule sat in cash for nearly three quarters of the window while SPY made 125.29% on its price series.

The last two sections cover what SPY's own statistics say about an oversold signal and what the rules do when they meet this fund. The test is one window, with daily-decision rules, one trade at a time, and no fees in the headline run.

Year by year

Four of six years behind a rising fund

Held, SPY returned 29.6% in 2021, -17.8% in 2022, 25.5% in 2023, 24.3% in 2024, 17.4% in 2025 and 13.6% in 2026 to date. The RSI rule returned 3.5%, -7.1%, 15.2%, 12.2%, -3.3% and 6.2%. It was ahead in one year, 2022, by 10.7 points, and behind in the other five.

2021 shows the cost of waiting. The first signal did not come until 2021-09-21, so the first months of the year were spent in cash and the whole 3.5% came from one trade, 3.61% from 2021-09-21 to 2021-10-21 at adjusted prices of 407.73 and 422.43. SPY was up 29.6% in the year, which is why the gap to holding is the widest in the table.

2022 is the year the template looks best against the fund and still lost money. Three trades closed in 2022 and only one won. The first, from 2022-01-18, lost 1.23% over 69 days. The second, from 2022-04-25 to 2022-07-11, lost 8.29% over 77 days. The third, from 2022-09-01 to 2022-11-25, made 2.65%. A fund that fell 17.8% and a rule that lost 7.1% is a better outcome for the rule, but both trades that lost money were buys of a fall that kept going.

2023 and 2024 are the best years for the rule. Four of four trades closing in 2023 were winners, at 3.22%, 3.87%, 3.51% and 4.18%, and two of two in 2024, at 3.85% and 8.47%. The rule made 15.2% and 12.2%, which is well below what the fund made in each year because the position was held for only part of it.

2025 is the second poor year. The position bought on 2025-02-28 at 573.80 was sold on 2025-04-30 at 536.90 for -6.43%, which is the second-worst trade in the table. The one later trade, from 2025-11-19, made 3.32%. The year total was -3.3% against 17.4% held. 2026 to date is 6.2%, mostly the 6.33% trade from 2026-03-19 to 2026-04-15. Against the fund's 13.6%, the gap is smaller than in other years because SPY had a sharp drawdown early in 2026 and a quick recovery.

Month by month

Most months are zero

The month table is mostly zeros because the sleeve is flat most of the time. Most months from January 2021 to September 2026 show no return at all. A zero means the position was not open for any day of that month, which for this rule is the usual state.

The worst months are June 2022 at -8.01% and September 2022 at -8.35%. The June figure belongs to the 2022-04-25 trade, which was open through the June low. September 2022 is the month after the 2022-09-01 entry, which was bought into a falling market and eventually closed positive on 2022-11-25. Buy-and-hold's worst month was also September 2022, at -9.00%. The best months are August 2024 at 8.20% and October 2022 at 7.67%. August 2024 is the 2024-08-06 trade, the best one in the table, and October 2022 is the same September 2022 position as it started to recover. Each of these months is a single trade.

March 2025 at -5.42% and April 2025 at -2.30% are the two halves of the 2025-02-28 trade. The strategy made 1.48% in February 2025 on the first day of the trade and lost it over the next two months.

The 2023 months show the pattern of a working mean-reversion rule: January 3.28%, March 3.77%, August 3.73% and November 7.48%, with small losses of -0.10%, -1.38% and -2.10% in between, and nothing in the other months. A typical winning month is 3% to 4% and comes from a trade of a month or two.

Because the strategy is invested so rarely, the monthly view says little about the signal and a lot about time in the market. Buy-and-hold had a best month of 10.33% in April 2026, a month in which the rule made 6.76% because it was holding the position from 2026-03-19 until 2026-04-15, and then sold before the rest of the move.

Every trade

RSI mean reversion on SPY made 13 closed round trips, an average hold of 43 days, an average winner of 4.30%, an average loser of −5.31%, a profit factor of 2.62, a longest losing streak of 2. It held a position at the close on 26.8% 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.

What the 13 trades look like

The 13 round trips held for an average of 43.2 days, with a median of 33 and a range from 20 to 85 days. The average winner returned 4.30% and the average loser lost 5.31%, so the rule wins often with moderate gains and loses rarely with larger losses. The profit factor of 2.62 comes from 10 winners against 3 losers. The median trade returned 3.51%.

The three losers are the 2022-01-18 trade at -1.23%, the 2022-04-25 trade at -8.29% and the 2025-02-28 trade at -6.43%. The first two were entered in the 2022 bear market and held for 69 and 77 days. The rule has no stop, so each was held until RSI(14) recovered above 70, which in a downtrend takes a long time. The worst trade, -8.29% at an adjusted entry of 397.90, was larger than the best winner's 8.47% in the other direction, and it took the -7.1% for the 2022 year on its own.

The best trades are the 8.47% from 2024-08-06 to 2024-08-26 at 506.10 and 548.96, the 6.33% from 2026-03-19 to 2026-04-15 at 650.73 and 691.92, and the 4.18% from 2023-09-22 to 2023-11-15 at 416.74 and 434.18. Each came after a short sharp fall: the August 2024 and March 2026 entries are in the weeks of the shallower drawdowns. The 2025-11-19 trade made 3.32% in 21 days, the shortest hold besides the 20-day August 2024 trade.

The entries cluster. Two trades in 2023, on 2023-08-18 and 2023-09-22, were entered a few weeks apart at nearly the same price, 416.71 and 416.74. The first exited on 2023-09-15 and the second entered a week later, which shows RSI(14) dropping below 30 twice in a short window with a recovery above 70 between them. It made 3.51% and 4.18%.

The longest hold, 85 days, was the 2.65% trade from 2022-09-01. At the time of entry SPY was close to its 2022 low, and the position outlasted the low on 2022-10-12. The strategy ended with no open position, and its last trade closed on 2026-04-15. For the rest of the window the rule was in cash while RSI(14) did not fall below 30, which is consistent with the fund's 18 oversold sessions in the whole window.

Largest drawdowns

Buy-and-hold's deepest drawdown ran from 2022-01-03 to 2022-10-12 and reached −24.0%.

Drawdowns and why the rule has them

The max drawdown was 17.09%, from 2022-05-04 to 2022-10-12, which took 161 days to reach and 313 days to recover, ending on 2023-08-21. It sits inside the 2022 bear market and is the effect of the 2022-04-25 trade. That trade entered at 397.90 and exited at 364.93, and the fund kept falling until 2022-10-12.

The second drawdown was 15.63%, from 2025-02-28 to 2025-04-08, in 39 days, and recovery took until 2026-04-14, which is 371 days. It is the one trade of 2025 that lost 6.43%, followed by a long wait for the equity curve to come back, since the rule made only a 3.32% trade in November 2025 and a 6.33% trade in March 2026. The third, 9.10% from 2021-10-20 to 2022-03-08, recovered on 2022-05-04.

Buy-and-hold's deepest drawdown was 23.99%, from 2022-01-03 to 2022-10-12, and it took until 2023-12-13 to recover. Its second was 18.41% from 2025-02-19 to 2025-04-08, recovered on 2025-06-26. The rule's drawdown is shallower than the fund's by 6.90 points in the first case, and it did not recover faster. The fund recovered from its 2025 low in 79 days and the rule needed 371.

The pattern is that the drawdowns come from holding an oversold position, not from being oversold at the right time. The rule buys at RSI(14) below 30 and the next leg down happens while the position is open. A stop would have limited the two large losing trades, and the template does not have one. The description of this template says so: no stop and no profit target.

The drawdown figure also flatters the rule. A 17.09% drawdown was produced by 3 losing trades, and for much of the time the rule was in cash with equity flat, so the rule spent a lot of the window recovering from losses while the fund was making new highs. The fund's own 23.99% drawdown was a 24.51% decline on the price series, 2022-01-03 to 2022-10-12, with a recovery date of 2023-12-13.

With trading costs

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

What 5 and 10 basis points did

The cost runs cost little here because the strategy trades rarely. With 26 fills, the CAGR goes from 4.37% to 4.09% at 5 basis points per fill and 3.84% at 10. The final value goes from $12,780 to $12,587 and then $12,417. The max drawdown rises from 17.09% to 17.28% and 17.57%, and the Sharpe ratio falls from 0.42 to 0.40 and then 0.38.

The cost of 10 basis points is small next to the gap to holding, which is 10.19 points of CAGR. Costs do not decide the result for this template on this fund. SPY is also the most liquid fund in the test, with an average daily dollar volume of $30,453,859,411 and a median minute volume of 107,870 shares, so the 5 and 10 basis point assumptions are conservative for it. The test applies the same assumption to every fund and does not measure real execution.

The point is that a rule that makes 13 trades is insensitive to cost and sensitive to everything else, and that is also why its ranking is low. The templates above it on SPY, such as the RSI(2) snapback, trade far more often and carry the cost burden that the page for that template covers.

Changing the parameters

The sell threshold matters more than the buy threshold

Four variants change one number at a time. Moving the buy threshold barely changes the CAGR. Buying below 25 gives 4.37% with 11 trades and 8 wins, the same CAGR as the published rules with a max drawdown of 18.26%. Buying below 35 gives 4.40% with 16 trades and 13 wins, and the drawdown deepens to 23.77%. A looser entry adds trades and win rate and adds drawdown without adding return.

The sell threshold is where the result moves. Selling above 65 gives 5.29% with 15 trades, 11 wins, a max drawdown of 15.36% and a final value of $13,443. Selling above 75 gives 7.53% with 12 trades, 10 wins, a max drawdown of 15.65% and a final value of $15,174. The published 70 sits between them at 4.37%.

That is a large change from a small edit, and it should be read with care. Both sell variants have a shallower drawdown than the published rules and a higher CAGR. The 75 version holds longer and captures more of each recovery. The test has only 13 trades, and moving one exit can change which trades are open in a given drawdown. The difference between 4.37% and 7.53% comes from a handful of trades whose exits moved, and the ordering of the variants may not hold on a different window. The 75 variant at 7.53% is still half of the 14.56% from holding.

The published parameters are the conventional 30 and 70 and were not tuned. The variants are not a search for the best setting. They show that the rule's weakness is the early exit rather than the entry: the oversold signal found trades that made money, and selling at 70 left part of each recovery on the table.

How SPY behaved

What SPY's own numbers say about an oversold signal

SPY returned 125.29% on its price series over 1444 sessions, 15.20% a year, with annualized volatility of 16.41%. Up days were 54.40% of sessions. The deepest drawdown was 24.51% from 2022-01-03 to 2022-10-12, recovered by 2023-12-13, and the longest drawdown lasted 488 sessions. The fund was above its 200-day average on 77.99% of sessions and crossed it 30 times.

The oversold signal is rare. RSI(14) fell below 30 on 18 sessions in the whole window and was above 70 on 134. After those 18 sessions the median 5-day forward return was 3.74% and the median 20-day return was 2.86%, against baselines of 0.42% and 1.73% for all sessions. That is a real edge in the data, though it rests on 18 observations, and they are clustered in a few episodes, so the effective sample is smaller. The rule trades that edge for 13 round trips because each episode can trigger a position that lasts weeks.

Because RSI(14) stays above 70 on 134 sessions, the exit is also a slow signal. A rule that sells at 70 waits for the fund to be stretched, while SPY spends much of its time in a rising trend. That is the reason the sell-above-75 variant earned more: the fund's rises are long enough to keep RSI(14) elevated.

SPY had 11 falls of 10% or more from a 20-day high, over 23 days. Those are the sell-offs the rule is meant to buy, and the fund recovered from each. The RSI(2) below 10 signal fired on 146 sessions with a median 5-day return of 0.89% and 20-day return of 2.29%, so a faster signal has more occurrences and a smaller edge per occurrence.

SPY earned 62.46% of its log return overnight and 37.54% during the session, and the average overnight gap was 0.44%. The correlation to QQQ was 0.94 and to TLT 0.08, with a beta of 0.69 to QQQ. Calendar returns were 30.46% in 2021, -18.16% in 2022, 26.18% in 2023, 24.86% in 2024, 17.72% in 2025 and 13.77% in 2026 to date.

Among broad index funds, this template did better on EEM at 9.51%, VOOV at 8.01% and QQQE at 7.22%. It did worse on IWM at 0.73%. On VOO, the closest fund to SPY, it made 4.22% from 13 round trips, nearly the same as here. The median CAGR across all 59 funds was 2.98% and the median among the broad index group was 6.46%, so SPY sits below its group median.

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 rules meet this fund

The rules are two lines. Buy with 98% of the sleeve when RSI(14) is below 30 and the sleeve is flat, sell everything when RSI(14) is above 70. Orders go in at the open. There is no stop and no target.

On SPY the rules produce a strategy that is invested about a quarter of the time. The fund's strong drift means most of its sessions are above 30 and many are above 70, so the rule waits for rare dips and then holds for a month or two. When it is in, it does well: 10 winners of 13. When it is out, SPY keeps rising. The strategy is a bet on a specific event, and the event is infrequent.

The weakest part is the absence of a stop in a bear market. The two largest losses came in 2022 and early 2025, when the fund kept falling after the signal. The strongest part is the entry. After RSI(14) fell below 30 on SPY, forward returns were well above the baseline.

The weekly 7% target made 13.88% on SPY and the RSI(2) snapback 13.17%, while the dip buyer made 2.61%. This template sits at 4.37% between the high-frequency rules and the dip buyer. The full ranking is on the SPY page, and the strategy page lists all 59 funds.

The limits of this page are the window, the 13 trades and the headline run without fees. A different window with more oversold readings would produce a different ranking.

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.