How to trade the RSI(2) dip strategy
Published October 3, 2026 · Closing Note research
The rules for trading the RSI(2) dip scan as plain steps, and what happened in 2026 when one rule was changed at a time. The results are for a whole account, not single trades, and before trading costs.
How to trade it
- Get the ranked list shortly before the close. The daily alert arrives at 15:45 ET (21:45 CET).
- Buy in the last 15 minutes of the session, at or near the closing price. Market or market-on-close orders on these liquid stocks fill close to it.
- Split the account into 15 equal positions, about 6.7% each. If that is too small to buy whole shares, use fractional shares rather than fewer positions.
- Buy only the top-ranked candidates, not every stock from the scan.
- If a stock you hold signals again, buy it again as a new position.
- Each day near the close, sell every position whose price is above its 5-day average. Sell anything still open on day 5. No stop-loss.
How these numbers were made
Every comparison below starts from the steps above: the top-ranked candidates from the daily scan between January 2 and September 30, 2026, 15 positions of 6.7% each, bought at the close and sold at the first close above the 5-day average or on day 5. Each one changes one rule and runs the whole account again.
All results are before trading costs, because those depend on your broker. A round trip of 0.10% would cut the model's +71.4% to +63.5%, and 0.20% to +55.9%. The track record deducts 0.10%.
2026 was a good year for dip buying, and the model's settings were chosen on these same months. That is part of why most changes below do worse than the model: it fits this year. The differences between the rows say more than the size of the returns.
A better average trade can make less money
Deeper dips, fewer positions and longer holds each raised the average trade, and each left the account with less. Deeper dips are rarer, so most of the money sat in cash. Fewer positions did the same. Holding up to 10 days let slow trades recover, but every slow trade kept its position blocked, so the account took 277 trades instead of 713 and missed most of the later signals.
| Change | Account | Worst drop | Trades | Avg trade |
|---|---|---|---|---|
| The model | +71.4% | −3.7% | 713 | +1.15% |
| Only the deepest dips (RSI(2) below 5) | +39.5% | −1.8% | 403 | +1.26% |
| 10 positions instead of 15 | +56.5% | −3.2% | 524 | +1.30% |
| Hold up to 10 days instead of 5 | +27.7% | −7.4% | 277 | +1.30% |
| Hold up to 3 days | +67.7% | −4.8% | 840 | +0.95% |
Holding only up to 3 days went the other way: more trades, a lower average, and a deeper worst drop, because many were sold before the bounce was over.
Taking every signal
Taking every stock from the daily scan, with the deepest RSI(2) first when positions are full, gave more trades and much less return. The extra trades were weaker, and they fell together on bad days.
| Candidates | Account | Worst drop | Trades | Avg trade |
|---|---|---|---|---|
| Top-ranked, 15 positions | +71.4% | −3.7% | 713 | +1.15% |
| All candidates, 15 positions | +12.1% | −12.4% | 832 | +0.23% |
| All candidates, 30 positions, borrowed money | +8.1% | −24.1% | 1,582 | +0.15% |
Doubling to 30 positions with borrowed money, to fit more of them in, lowered the return again and doubled the worst drop.
When to buy
A stock bought on a dip often keeps falling for a day. Of all 1,241 top-ranked signals in 2026, 54% traded at least 1% below the buy price the next day, 33% at least 2%, and 20% at least 3%. Still, 70% of all signals ended in profit within the five days. Connors and Alvarez saw the same in their tests and left stops out of their rules for that reason.
Waiting to buy lower did not pay. Bought at the close, the average signal made +1.25%. Bought at the next morning's open it made +1.05%, because much of the bounce happens overnight. An order 1% below the close filled on 54% of the signals. Those trades made +1.47%, but the ones it missed made +2.67%, so over all signals it came to +0.80%.
These averages are per signal and include the signals a 15-position account had no room for, which is why they differ from the account's average trade above.
Repeat signals
A stock you bought often keeps falling and signals again a day or two later. You can buy it again as a new position, or skip it because you already hold it. KLA in February:
| Day | Close $ | 5-day avg $ | |
|---|---|---|---|
| Tue Feb 3 | 135.21 | 149.73 | Buy at the close |
| Wed Feb 4 | 130.39 | 143.35 | Lower and signals again: buy a second position |
| Thu Feb 5 | 132.76 | 136.29 | Below the 5-day average: hold both |
| Fri Feb 6 | 143.92 | 136.59 | Sell both: first close above the 5-day average |
With $1,000 per position, the first buy made $64 and the second $104, on the same bounce bought at a lower price. Over the year, buying again came out ahead:
| Repeat signal on a held stock | Account | Worst drop | Trades |
|---|---|---|---|
| Buy again | +71.4% | −3.7% | 713 |
| Skip | +53.3% | −4.0% | 593 |
| Skip, positions 17% bigger | +65.0% | −4.7% | 593 |
Skipping left money in cash. The last row puts that money into bigger positions instead; buying again still made more, with a smaller worst drop. It goes wrong when a stock keeps falling. Micron signaled five days in a row in March:
| Bought | Price $ | Sold | Price $ | Result |
|---|---|---|---|---|
| Mon Mar 23 | 404.12 | Mar 30, day 5 | 321.75 | −20.4% |
| Tue Mar 24 | 395.30 | Mar 31, day 5 | 337.79 | −14.5% |
| Wed Mar 25 | 381.87 | Apr 1, above avg | 367.79 | −3.7% |
| Thu Mar 26 | 355.26 | Apr 1, above avg | 367.79 | +3.5% |
| Fri Mar 27 | 357.02 | Apr 1, above avg | 367.79 | +3.0% |
With $1,000 per position, the first buy alone lost $204; all five together lost $321. For a few days a third of a 15-position account was in one stock. Each repeat buy uses one of the 15 positions, so the account never holds more than it has, but more of it can end up in one stock. The position sizing sandbox lets you switch repeat buys off and compare.
Selling a gap up
Sometimes a stock opens far above the buy price the next morning, and taking the profit right away is tempting. Southwest Airlines (LUV) was bought at $40.51 on January 28, opened at $43.01 the next day (+6.2%) and closed at $48.08, above its 5-day average, for +18.7%. Carvana (CVNA), bought the same day at $82.01, opened at $86.25 (+5.2%), drifted lower and was sold on day 5 at $78.61, for −4.1%.
37 top-ranked trades opened 4% or more higher in 2026. Selling them at the open made +6.12% on average; holding them made +7.32%, but did worse than selling in 41% of them. For the account, the answer depended on whether the money had somewhere else to go:
| Sell at a gap up of | 15 positions | Worst drop | 30 positions | Worst drop |
|---|---|---|---|---|
| Never | +71.4% | −3.7% | +129.0% | −4.4% |
| 2% or more | +64.7% | −3.9% | +120.5% | −5.2% |
| 4% or more | +69.8% | −3.7% | +135.6% | −4.4% |
With 15 positions there was usually cash to spare, and selling early gave up the rest of the bounce. With 30 positions the money was often fully used, and freeing it a few days early for the next signal helped at 4%. Selling at every 2% gap did worse in both. With 37 large gaps in nine months, the difference is small and could turn the other way next year.
More on sizing
How much to put into each trade, how many positions to hold, how far borrowed money deepens the worst drop, and what to do when dozens of stocks signal on the same day are on the position sizing page, with a sandbox to try them on the signals. Every top-ranked trade, losers included, is in the track record.
Get the daily alert
Every trading day at 15:45 ET (21:45 CET), before the U.S. close: the full candidate list, ranked, with the top-ranked names marked and exits flagged for open trades.
Free during the open beta. One click to unsubscribe.
Questions
Is a higher average trade always better?
No. A rule that makes each trade better but takes fewer of them, or holds them longer, can leave money idle and earn less in total. Judge a change by what the whole account made and how far it fell.
Should I buy at the open instead of the close?
No. In 2026 a trade bought at the next morning's open made 0.2 percentage points less on average than one bought at the close, because much of the bounce happens overnight.
Should I buy again when a stock I hold signals again?
In 2026 it paid: the second buy gets the same bounce at a lower price. It also puts more money into a stock that may keep falling, so the losses on one bad stock get bigger.
Should I sell when a stock gaps up the next morning?
It depends on whether you have other trades waiting for the money. Holding made more on average, but selling at a 4% gap freed the cash for the next signal, which helped only when the account was fully invested.
References
- Connors, L. and Alvarez, C. (2008). Short Term Trading Strategies That Work. TradingMarkets Publishing. ISBN 978-0-9819239-0-1. openlibrary.org
Origin of the RSI(2) rules; tests on 1995–2007 data.