RSI(2) Mean Reversion Backtest: Why We Retired It for Stocks
We don't build strategies to sell them. We build them to test them — and when the data says no, we retire them. Today, two of them: RSI(2) Mean Reversion (Larry Connors' classic) and the Capitulation Finder. Both are mean-reversion approaches: they buy weakness and sell strength instead of following a trend.
We ran both with default parameters on daily candles from 2018 to 2026 — across 50 crypto pairs, 30 stocks, 10 ETFs and 10 commodities. The benchmark: average Buy & Hold return (Avg B&H) across all entry points. Here are the numbers, unvarnished.
The data
| Strategy · Asset class | Beats Avg B&H | Avg strategy CAGR | Avg B&H CAGR | Avg ΔCAGR | Avg trades |
|---|---|---|---|---|---|
| RSI(2) · Crypto Top 50 | 31 / 50 | +5.4% | −4.3% | +9.7% | 64 |
| RSI(2) · Stocks (30) | 2 / 30 | +3.3% | +18.1% | −14.7% | 117 |
| RSI(2) · ETFs (10) | 1 / 10 | +2.8% | +12.5% | −9.7% | 117 |
| RSI(2) · Commodities (10) | 3 / 10 | +0.8% | +2.9% | −2.2% | 76 |
| Capitulation · Crypto Top 50 | 29 / 50 | −5.3% | −4.3% | −1.0% | 16 |
| Capitulation · Stocks (30) | 3 / 30 | +7.8% | +18.1% | −10.3% | 13 |
| Capitulation · ETFs (10) | 2 / 10 | +6.3% | +12.5% | −6.2% | 11 |
| Capitulation · Commodities (10) | 3 / 10 | +2.4% | +2.9% | −0.5% | 11 |
Does RSI(2) work on stocks? Mean reversion loses to buy & hold
On stocks, RSI(2) beats Buy & Hold in 2 of 30 cases. On ETFs, 1 of 10. That's not bad luck, it's structure: U.S. equities have been in a secular bull market since 2009. A strategy that regularly steps into cash to re-enter "cheaper" misses exactly the upside that drives most of the return. It delivers +3.3% CAGR while simply holding returns +18.1%.
The honest conclusion is uncomfortable for anyone selling "stock signals": for broad stocks and index ETFs, regularly buying and holding (DCA) is the edge. Not the timing. That isn't a defeat — it's one of the most valuable things an honest backtest can tell you. Accept it and you save on fees, taxes and stress.
That's why we keep these runs visible — in the Strategy Library and in Insights. We just no longer offer the two strategies for stocks/ETFs/commodities/forex in the backtester.
RSI(2) on crypto: it works — for an uncomfortable reason
On crypto the picture flips: RSI(2) beats Avg B&H on 31 of 50 pairs, by +9.7% on average. That sounds like an edge — and it is one, but you have to understand why.
The average Buy & Hold of the crypto Top 50 over this period is negative (−4.3% CAGR). Many altcoins that sat in the Top 50 in 2021 have lost 80–95%. A strategy that spends most of its time in cash and only catches short rebounds will almost inevitably beat holding a dying coin. Put differently: on crypto, RSI(2) wins mainly by not holding bags. With 64 trades on average and a 63% hit rate, that's a statistically solid, defensive property — not noise. So RSI(2) stays active for crypto.
Capitulation Finder backtest: too rare to be evidence
With the Capitulation Finder the case is clearer — and worse. On stocks (3/30) and ETFs (2/10) it loses to Buy & Hold as expected. But even on crypto, where mean reversion should have a chance, it only reaches a tie with Buy & Hold (−1.0% ΔCAGR) — and that at an average maximum drawdown of −62%. A breakeven result for which you risk nearly two-thirds of your capital along the way is not a usable profile.
On top of that, our hard rule: under 30 trades, a backtest is an anecdote, not evidence. The Capitulation Finder fires only 11 to 16 trades per asset over eight years. Even where the numbers happen to look good, the sample is too small to prove anything. Rare signals sound exclusive — statistically they're worthless.
So we're retiring the Capitulation Finder entirely, across all asset classes.
What stays visible
Retiring isn't deleting. Both strategies remain visible in the Strategy Library and in Insights — marked "no longer active" and showing exactly these aggregate numbers. If you want to understand why something didn't work, you should be able to. Vanishing strategies are precisely the kind of selective memory we criticize in others.
Platform bridge
The Backtesting Arena is built to answer questions like these yourself — across hundreds of assets, not one well-chosen example. You can run RSI(2) on crypto against Buy & Hold, see the drawdowns and check the trade count. That transparency is the point: a strategy is only as good as its most honest backtest — and sometimes the most honest result is "don't."
FAQ
What is the RSI(2) strategy? The RSI(2) strategy, popularized by Larry Connors, uses the 2-period RSI as a short-term mean-reversion filter: it buys when RSI(2) is very low (oversold) and sells back into strength. On daily data it's a classic "buy the dip" approach — it bets that prices revert to their mean after a short drop.
Why build strategies just to retire them? Because that's the process. Building a strategy is cheap; testing it honestly is the real work. A retired candidate with documented numbers is worth more than ten untested ones sitting in the catalog.
Does this mean mean reversion just doesn't work? No. It works where prices actually revert to a mean — in sideways and bear phases, like many altcoins. In a strong, long uptrend (stocks since 2009) it loses to simply holding. Context decides, not the strategy itself.
Why is Buy & Hold the "fair" benchmark and not another strategy? Because Buy & Hold is the zero-effort alternative. If an active strategy can't even beat passive holding, there's no reason for the effort, the risk and the taxes. It's the toughest, most honest yardstick.
Isn't a +9.7% edge on crypto just survivorship bias itself? Partly — and we say so openly. RSI(2) wins mainly by not holding dying alts. That's a real but defensive edge: loss avoidance, not gain maximization. On an asset in a clear uptrend (e.g. BTC alone) the strategy lags just like it does on stocks.
Why aren't 11–16 trades enough? With so few trades, chance drives the result. A single lucky or unlucky trade flips the whole statistic. Only from around 30 trades does a result become interpretable at all — below that it's a story, not proof.
What should I do with stocks now? This isn't investment advice, just what the data says: for broad stocks and index ETFs, regularly buying and holding beat every timing strategy we tested. Spend the energy on savings rate and discipline, not on signals.