The Test Bench › Mean reversion › Study 015
Which RSI makes the better dip-buying entry? We backtested RSI(2), RSI(14), Stochastic RSI and ConnorsRSI under the same rules on 483 S&P 500 stocks, 2006 to 2026.
STUDY 015We ran one dip-buying rule on 483 S&P 500 stocks, 2006 to 2026: after an oversold close in a stock above its 200-day average, buy at the next day’s open; after the first close above the 5-day average, sell at the next open. Only the RSI signal used to buy changed.
RSI(2) came first, Stochastic RSI a close second, ConnorsRSI third and RSI(14) last, and the order held at a higher trading cost and with fills at the signal close. With no trading cost, Stochastic RSI finished ahead of RSI(2), so the top place depends on costs.
Buying and holding finished higher on most stocks, but it was invested every day; the rules held a stock only 0.5% to 14% of the time.
Against a position of the same average size held every day, every rule finished higher on most stocks (a comparison we added after seeing the main results).
Each entry explained, with worked examples: RSI and RSI(2), ConnorsRSI and Stochastic RSI. How mean reversion works, with its entries, exits and risks: mean reversion trading explained.
A historical backtest under stated assumptions: long positions only, one stock at a time, trades at the next day’s open, a fixed cost on each buy and sale. It describes what happened in the past.
| Part | Rule |
|---|---|
| Trend filter | The close is above its 200-day simple moving average |
| Entry | RSI(2) below 10, RSI(14) below 30, StochRSI %K below 20 and ConnorsRSI below 10, one per strategy, read at the close |
| Exit | The first close above its 5-day simple moving average |
| Execution | Buys and sales at the next day’s open after the signal close; fills at the signal close itself, as in Connors’ published rules, as a check |
| Stop | None, as in Connors’ published rules |
| Position | $10,000, all cash on each buy, whole shares, long only; idle cash earns nothing |
| Cost | 0.02% on each buy and each sale; 0% and 0.04% as checks |
The typical account for each rule, from $10,000 in 2006: the median across the 368 stocks with data from 2006, so every year has the same stocks behind it. RSI(2)’s typical account ended at $15,251; buying and holding ended at $46,813. At the low of the 2008–09 sell-off, buying and holding was down to $5,498, while RSI(2) stood at $10,189: with most stocks below their 200-day average, the rules held almost nothing.
The rules wait for a dip in a stock above its 200-day average, so most of the time they hold nothing. Stochastic RSI, the busiest, held a position in 13% of the stocks in a typical week. In the 2008 sell-off, when most stocks fell below their 200-day average, the share dropped to 0.5%.
Compared stock by stock, the order held in every pair at both costs and with signal-close fills: RSI(2), then Stochastic RSI, then ConnorsRSI, then RSI(14). The closest pair was RSI(2) against Stochastic RSI: RSI(2) finished ahead of Stochastic RSI on 243 of 483 stocks, with a median gap of +0.5 points of total return per stock. With no trading cost, Stochastic RSI was ahead of RSI(2) on 253 of 483 stocks.
Each rule finished ahead of buying and holding on 42 to 61 of 483 stocks. Its falls were smaller, and so was the time it was invested.
The entries that found more dips made more: Stochastic RSI traded a median 209 times per stock and RSI(2) traded a median 132 times per stock, against 35 for ConnorsRSI and 6 for RSI(14). One possible explanation is that they caught more short rebounds; the test does not isolate that. The rules sat in cash 86% to 99% of the time, earning nothing here, while buying and holding was invested all the time.
The rules, the cost, the stocks and what would count as one entry finishing ahead of another were set on 6 October 2026 in our internal study record, before any backtest was run; the record is internal, not a public preregistration. Each pair of entries is compared stock by stock: one entry is ahead when it finished higher on most stocks with a positive median gap at 0.02%, at 0.04% and with signal-close fills.
One change after the results: the fill. The record filled trades at the close of the signal day and checked the next day’s open. The signal needs that day’s closing price, so we lead with the next-open fills. The change lowered every figure and left the ranking unchanged.
The 483 stocks are today’s S&P 500 companies that passed the data screens of our RSI event studies, on the same daily bars, 2006 to 2026. Buying and holding the same stock with the same cost is the reference.
Survivorship bias. This is not the historical index. It is today’s S&P 500 members projected backwards, so companies that failed or left the index after a deep fall are missing. That can materially favour a rule that buys falls.
In the strategy builder, add RSI as a custom signal with period 2 or 14, and SMA with periods 5 and 200. Buy when the close is above SMA200 and RSI is below its level; sell when the close is above the 5-day SMA, with orders at the next open. For the other two entries, use Stochastic RSI %K below 20 or ConnorsRSI (Momentum group) below 10 in place of RSI. Set the cost to 0.02% on each buy and sale and run the backtest on daily bars.
Added after the results. Buying and holding is invested every day and the rules are not, so we compared each rule with a position of the same average size held all the time (a yardstick that ignores the cost of adjusting it daily), and with yearly return divided by the deepest fall (a simple measure of our own, not a standard ratio such as Sharpe or Calmar). Against the same-size position, every rule finished higher on most stocks; measured against the deepest fall, buying and holding still came out ahead (0.14 against 0.04 to 0.08).
On the 461 stocks whose data covers the 2008 and 2020 sell-offs, RSI(2) was ahead of Stochastic RSI on only 228: a change of direction against all 483 stocks.
| Entry | Median total return per stock, 2006 to 2026 | Time in the market | Deepest fall | Beat buy and hold on (stocks) | Beat the same-size position on (stocks) |
|---|---|---|---|---|---|
| RSI(2) below 10 | +45% | 11% | −28% | 57 of 483 | 298 of 483 |
| RSI(14) below 30 | +5% | 0.5% | −6% | 42 of 483 | 325 of 483 |
| StochRSI %K below 20 | +44% | 14% | −30% | 61 of 483 | 254 of 483 |
| ConnorsRSI below 10 | +13% | 3.2% | −18% | 47 of 483 | 292 of 483 |
| Buy and hold (reference) | +361% | 100% | −66% | — | — |
| Entry | Trades (median) | Win rate | Average win | Average loss | At 0% | At 0.02% | At 0.04% | Signal-close fills, 0.02% |
|---|---|---|---|---|---|---|---|---|
| RSI(2) below 10 | 132 | 67% | +2.0% | −2.8% | +53% | +45% | +38% | +52% |
| RSI(14) below 30 | 6 | 67% | +2.3% | −2.1% | +5% | +5% | +4% | +6% |
| StochRSI %K below 20 | 209 | 65% | +1.6% | −2.2% | +56% | +44% | +33% | +46% |
| ConnorsRSI below 10 | 35 | 67% | +2.3% | −2.9% | +15% | +13% | +11% | +17% |
| Comparison | At 0.02% | At 0.04% | Signal-close fills, 0.02% |
|---|---|---|---|
| RSI(2) ahead of RSI(14) | 381 of 483 · +41.4 points | 364 of 483 · +34.6 points | 391 of 483 · +46.0 points |
| RSI(2) ahead of Stochastic RSI | 243 of 483 · +0.5 points | 256 of 483 · +3.5 points | 251 of 483 · +1.1 points |
| RSI(2) ahead of ConnorsRSI | 365 of 483 · +29.6 points | 356 of 483 · +24.3 points | 386 of 483 · +32.3 points |
| RSI(14) ahead of Stochastic RSI | 129 of 483 · −36.9 points | 148 of 483 · −26.3 points | 112 of 483 · −40.5 points |
| RSI(14) ahead of ConnorsRSI | 166 of 483 · −8.5 points | 177 of 483 · −7.3 points | 154 of 483 · −11.1 points |
| Stochastic RSI ahead of ConnorsRSI | 339 of 483 · +27.1 points | 308 of 483 · +17.8 points | 338 of 483 · +28.7 points |
Comparison cells: the stocks on which the first entry finished higher, and the median of the per-stock differences in percentage points. RSI(14)’s win rate rests on a median of 6 trades per stock, so it is rough.
Using this research. The analysis, tables and figures on this page are GU Analyser Ltd’s own work, published under a Creative Commons BY-NC 4.0 licence. You are welcome to quote it, cite it and build on it, with credit and a link. Suggested credit: GU Analyser, The Test Bench, Study 015, linking to guanalyser.com/test-bench/rsi-mean-reversion-backtest/. For commercial use, ask us at hello@guanalyser.com.
Rebuild the test
Build the rule on the strategy canvas: a close above the 200-day SMA and an RSI entry, a sale on the first close above the 5-day SMA, and backtest it on daily bars at your own cost.
Create a free account and run your own tests → or read how RSI is calculated →The method: how a backtest is built