The RSI 2 strategy is one of the best-known short-term trading systems ever published, and on Nasdaq-100 futures it held up better than most ideas we test: every oversold threshold from 5 to 45 was profitable after costs, and in both the 2010s and the 2020s. It also has a weakness that the headline win rate hides.
The system is most closely associated with Larry Connors, who popularized RSI(2) as a mean-reversion tool; the book most traders point to is Short Term Trading Strategies That Work (2008), co-written with Cesar Alvarez. The idea is simple: in a long-term uptrend, buy sharp short-term pullbacks and sell into the first bounce.
Connors RSI 2 rules we tested
- Market: NQ continuous front-month futures, daily bars, June 7, 2010 to April 16, 2026. Long only.
- 200-day moving average filter: the daily close must be above its 200-day simple moving average.
- Entry: buy at the close when the 2-period RSI (Wilder smoothing) closes below the threshold. We tested thresholds of 5, 10, 15, 20, 25, 30, 35, 40 and 45.
- Exit: sell at the close on the first day the close is above its 5-day simple moving average.
- Stop: none. One position at a time.
- Costs: 1.0 NQ point per round trip.
All results are in NQ index points for one contract. MNQ is $2 per point and NQ is $20 per point, so MNQ dollars below are simply points × 2. Every trade has 1.0 point deducted for costs (slippage plus commission). This is the version of the system most commonly attributed to Connors; we did not check every detail against the book, and his published versions include variations.
RSI 2 strategy results on NQ futures
| Entry | Trades | Win rate | Avg pts | Total pts | Profit factor | Closed-trade DD (pts) | 2010–2019 | 2020–2026 | PF 2020–2026 |
|---|---|---|---|---|---|---|---|---|---|
| RSI(2) < 5 | 82 | 73.2% | +49.93 | +4,094.2 | 1.86 | 1,993.5 | +544.8 | +3,549.5 | 2.06 |
| RSI(2) < 10 | 166 | 71.7% | +60.48 | +10,039.2 | 2.6 | 1,993.5 | +1,415 | +8,624.2 | 3.0 |
| RSI(2) < 15 | 225 | 72.4% | +52.42 | +11,793.8 | 2.55 | 1,993.5 | +2,017.5 | +9,776.2 | 2.9 |
| RSI(2) < 20 | 284 | 73.6% | +43.11 | +12,242.2 | 1.99 | 2,495 | +1,724.2 | +10,518 | 2.21 |
| RSI(2) < 25 | 328 | 73.2% | +34.68 | +11,374 | 1.82 | 2,495 | +1,777.2 | +9,596.8 | 2.03 |
| RSI(2) < 30 | 372 | 71.2% | +27.16 | +10,103.5 | 1.6 | 2,495 | +1,799 | +8,304.5 | 1.71 |
| RSI(2) < 35 | 412 | 72.6% | +28.57 | +11,771.8 | 1.64 | 2,495 | +1,979.2 | +9,792.5 | 1.75 |
| RSI(2) < 40 | 455 | 73.4% | +28.27 | +12,862.5 | 1.65 | 2,495 | +2,205.5 | +10,657 | 1.76 |
| RSI(2) < 45 | 492 | 74.2% | +25.95 | +12,767.5 | 1.6 | 2,495 | +2,439.8 | +10,327.8 | 1.67 |
The first thing to notice is consistency. Win rates sit between 71.2% and 74.2% at every threshold, and every row is positive in both 2010–2019 and 2020–2026. That matters: NQ went from roughly 1,800 to roughly 26,000 over the test, so the 2020s dominate the point totals, and a strategy that only worked in the 2020s would be a red flag. This one did not have that problem.
The second thing is where the sweet spot lies. The classic RSI(2) < 5 setting is actually the weakest row: 82 trades, 4,094.2 points, profit factor 1.86. Waiting for extreme readings meant fewer trades and bigger average losses (-215.17 points). Moving to < 10 doubled the trade count to 166 and produced the best profit factor in the table, 2.6 (3.0 in 2020–2026), with 10,039.2 points. RSI(2) < 15 was close behind: 11,793.8 points, profit factor 2.55.
Above 20, total points kept roughly flat — RSI(2) < 40 made the most, 12,862.5 points — but by < 30 the profit factor had fallen to around 1.6, and the closed-trade drawdown grew from 1,993.5 to 2,495 points. Looser thresholds add trades of lower quality.
The catch: average loss bigger than average win
At most thresholds the average losing trade was larger than the average winner. At < 5 the average win was 147.13 points and the average loss -215.17; at < 40 it was 97.81 versus -165.06. Only the < 10 and < 15 settings had wins and losses of similar size. With no stop loss, the RSI 2 strategy wins most of the time and occasionally holds a falling market for days. The closed-trade drawdown figure understates this, because it only counts losses once a trade is closed, not the open loss while you are waiting for price to get back above the 5-day average.
Year-by-year observations
| Year | RSI(2) < 5 | < 10 | < 15 | < 25 |
|---|---|---|---|---|
| 2016 | +40 | -188.8 | -216.8 | -49.8 |
| 2018 | -269 | -89.8 | +262 | -121.5 |
| 2020 | -455.2 | +481.8 | +738 | +1,131 |
| 2021 | +1,270.5 | +2,801 | +2,967.5 | +2,977 |
| 2022 | no trades | +35 | +35 | -1,481.5 |
| 2024 | +2,683.2 | +2,385.8 | +3,135.5 | +4,595.8 |
| 2025 | -996.2 | +1,012.8 | +380.5 | +205.2 |
- There are no trades in 2010 at any threshold, because the 200-day moving average needs 200 days of data first.
- RSI(2) < 10 was positive in 13 of 16 years with trades; < 15 in 14 of 16. No losing year for < 10 was worse than -188.8 points.
- 2022 shows the value of the 200-day filter. The stricter thresholds (< 5 took no trades that year; < 10 and < 15 made 35 points) stayed mostly out of the bear market. From < 20 upward, the same -1,481.5-point loss appears in 2022, which suggests one or two bad trades that the looser settings could not avoid.
- 2025 hit the extreme setting hardest: RSI(2) < 5 lost 996.2 points that year while < 10 made 1,012.8.
- 2021 and 2024 were the big years across the board.
Our take: The RSI 2 strategy is one of the rare published systems that still looks sensible on a modern market years after it was popularized. On NQ, we would favour a threshold around 10–15 over the famous < 5: more trades, the best profit factors and better-balanced wins and losses. What we would not do is trade it without thinking about the lack of a stop. The rules as tested can sit in a losing position for a long time, and a leveraged futures account may not survive that. If you want to compare RSI periods, see RSI 2 vs 3 vs 6 vs 14.
Versus buy and hold
Holding one NQ contract through the whole period gained 24,677.2 points, with a 5,878.2-point drawdown. The best RSI(2) rows made roughly half of that. The RSI 2 strategy is not a way to beat buy and hold in a strong bull market; it is a way to capture part of it while holding only after pullbacks.
Limitations
- This is a backtest; past results do not guarantee future results.
- Continuous-contract data has roll gaps; fills are assumed at the exact daily close. Taxes are not modelled.
- No stop is used, and closed-trade drawdown understates the real drawdown while trades are open.
- Only one market (NQ) was tested, and it spent most of this period in a strong uptrend.
Key takeaways
- Every RSI(2) threshold from 5 to 45 was profitable in both 2010–2019 and 2020–2026 on NQ futures.
- RSI(2) < 10: 166 trades, 71.7% wins, 10,039.2 pts, profit factor 2.6 (best in the table).
- The classic < 5 setting was the weakest row: 82 trades, profit factor 1.86.
- Average losses exceeded average wins at most thresholds; there is no stop.
- 1.0 point per round trip deducted; MNQ = $2/pt, NQ = $20/pt.
Put the research to work on your Desk
Explore the tools behind the research, compare the available access, or open your Desk if you already have an account.
FAQ
What is the RSI 2 strategy?
A mean-reversion system popularized by Larry Connors: when price is above its 200-day moving average, buy at the close when the 2-period RSI is very low, then sell when price closes above its 5-day moving average.
Does the RSI 2 strategy work on Nasdaq futures?
In our 2010-2026 NQ backtest, every threshold from 5 to 45 was profitable after 1 point of costs per trade, in both 2010-2019 and 2020-2026. RSI(2) below 10 made 10,039.2 points with a profit factor of 2.6. Past results do not guarantee future results.
What is the best RSI(2) threshold?
On NQ, thresholds of 10-15 had the best profit factors (2.6 and 2.55). The classic below-5 setting had the lowest profit factor (1.86) and fewest trades (82).
Does the RSI 2 strategy use a stop loss?
Not in the version we tested. That is why average losses were larger than average wins at most thresholds, and why closed-trade drawdown understates the real risk.
Loomi.Ai provides educational market research, not financial advice. Trading futures and options involves substantial risk of loss, and backtested results are historical and do not guarantee future returns.