RSI-2 Mean Reversion on Forex: Does It Still Work After Costs?
Search for “RSI 2 strategy” and you will find two kinds of results. The first kind shows a smooth equity curve and a high win rate, usually on one stock index, usually without costs. The second kind says it stopped working years ago. Neither tells you what a part-time trader on MetaTrader 5 actually wants to know: does it work on forex and index CFDs today, after spread, commission and swap, on data it was never tuned on?
We tested exactly that before putting RSI-2 on the PriceEdge Scanner. This article explains the strategy, how it was tested, what the results were, and which popular tweaks made it worse. The short answer: it works, the edge is small, and it only shows up when the rules are followed every time across many instruments.
The Idea Behind RSI-2
RSI-2 comes from Larry Connors, who described it in Short Term Trading Strategies That Work (2008). It is a mean reversion strategy: it bets that a short, sharp move against the trend tends to snap back.
The tool is the Relative Strength Index with a period of 2 instead of the usual 14. With such a short period, the RSI swings to extremes after just one or two strong days. A reading below 10 means the last two days were an unusually hard sell-off. A reading above 90 means an unusually strong rally.
The strategy only takes those signals in the direction of the bigger trend. In an uptrend, a two-day sell-off is often profit-taking or noise, not the start of a reversal, so it buys the dip. In a downtrend it does the mirror image and sells the rally.
The Rules as We Tested Them
Connors’ original rules were written for stocks. We kept the core and made the entry and exits suitable for forex and index CFDs on daily candles that close at 5 PM New York.
Trend filter. Longs only when the daily close is above the 200-day moving average. Shorts only when it is below.
Signal. For a long, RSI(2) closes below 10. For a short, RSI(2) closes above 90.
Entry. A limit order 0.5 × ATR below the day’s close for a long (above it for a short), valid for the next session only. If price does not dip that far, there is no trade. This is important: the limit entry buys a little lower than the signal close, and it is the reason costs stay low.
Stop. 2 × ATR from the entry. That is wide on purpose. Mean reversion needs room, and a tight stop turns many eventual winners into losses.
Exit. At the first daily close back above the 5-day moving average for a long (below it for a short). If that has not happened after 5 daily closes, the trade is closed anyway. That second rule is the time stop.
There is no take profit in the usual sense. The exit is a rule on the daily close, not a fixed price.
How It Was Tested
A backtest is only as honest as the process around it. It is easy to try a hundred settings and report the best one; that result says more about luck than about the strategy. So the process was fixed before any test ran:
- The rules and the pass mark were written down first. To go live, a strategy needs at least 30 closed trades, a positive average result per trade, and a profit factor of at least 1.2.
- Variants were compared on one period only: 2023 to 2024, the selection period.
- The chosen settings had to pass again on 2021 to 2022, years they were never tuned on.
- The held-out period, 2025 onwards, was run once, with the settings fixed. No adjustments afterwards.
Every result includes spread, commission and swap. The test ran on 20 instruments: major and cross FX pairs and stock indices.
Profit factor is the total of all winning trades divided by the total of all losing trades. Above 1.0, the strategy made money. 1.2 is the bar we set for “worth trading after costs, with a margin for error”.
The Results
| Period | Role | Profit factor | Average per trade |
|---|---|---|---|
| 2021 to 2022 | Unseen | 1.44 | about +0.1R |
| 2023 to 2024 | Selection | 1.36 | about +0.1R |
| 2025 to Oct 2026 | Held-out, run once | 1.48 | about +0.1R |
Backtested results, including spread, commission and swap. Past results are not a promise of future results.
R is the amount risked on one trade. An average of +0.1R means that over many trades, each one made about a tenth of what it risked. At 1% risk per trade, that is about 0.1% of the account per trade, before the ups and downs.
The strategy produces about two trades a week across the 20 instruments. Most trades win, about six in ten, and most winners are small. The losers that hit the full stop are larger. That shape is typical of mean reversion: many small wins, occasional full losses, and losing weeks and months along the way.
Why One Pair Is Not Enough
The edge per trade is small, and the result of a single trade swings far more than the edge itself. On any one instrument, a year of RSI-2 can easily be negative. Some public tests of RSI-2 on a single currency pair show a loss, and that fits what we found: the edge only becomes visible when it is spread across many instruments and many trades.
This has a practical consequence. Trading RSI-2 on EURUSD alone, or skipping the setups you do not like the look of, changes the strategy into something that was never tested.
What Did Not Improve It
Part of testing is trying the obvious improvements. Several looked promising and did not hold up.
Looser settings for more trades. Looser RSI thresholds and limit distances gave two to three times as many trades, but roughly the same total result, with deeper drawdowns. The best three variants failed on 2021 to 2022.
Taking profit early. Closing a trade once it was in profit and price stopped making new highs for a few hours won more often (up to 73% of trades, against 63%), but every one of the nine variants made less in total on 2023 to 2024. The trades closed early would have done better if left alone.
More markets. The same rules on oil and gas, precious and industrial metals, Bitcoin and Ether, and exotic currencies all failed. Wider spreads and swap eat the small edge on those markets.
Different exits and stops. Removing the time stop, making it longer, exiting on an RSI(2) level, or using a wider stop all did worse or were indistinguishable from luck.
Every one of these is published in the PriceEdge research library, with its numbers.
Should You Trade It?
RSI-2 is not exciting. It will not double an account. What it offers is a written, tested method with a small edge after costs, that held up on years it was never tuned on.
It suits a trader who can follow rules without exceptions, accept that most of the result comes from many trades rather than any single one, and live with losing months. It does not suit someone looking for a high return from a few trades, or someone who will skip the next setup after a loss.
Where PriceEdge Fits
RSI-2 is one of the strategies on the PriceEdge Scanner. After every New York close, the Scanner checks your instruments against the rules and lists the setups with the entry, the stop, the exit rule in plain words and lot sizes for the risk you choose. For every instrument without a setup, it shows the rule it stopped at, so you can see the method working rather than take it on trust.
You can try it for 60 days on a demo account, with no credit card.