V I S O R

Strategies · intermediate · 8 min

Mean-Reversion Basics

Where trend-following bets that a move continues, mean reversion bets the opposite: that price stretched too far from its recent average will snap back. The everyday version is "buy the dip" — when something has fallen hard, buy it, expecting a bounce. To turn that instinct into a rule you need a way to say "fallen hard," and the classic tool is the Relative Strength Index (RSI).

What RSI measures

RSI is an oscillator that compresses recent up-moves versus down-moves into a number between 0 and 100. High readings (traditionally above 70) mean recent candles have been mostly up — "overbought." Low readings (below 30) mean mostly down — "oversold." The mean-reversion claim is that an oversold reading marks exhaustion: sellers are spent, and a bounce is due. It is one of the first indicators nearly every new trader meets, and "buy when RSI drops below 30" is one of the first rules they try.

The mechanical rule we tested

We tested exactly that beginner's rule, with a sensible exit attached:

Run on SPY, five years of daily bars. No discretion, no extra filters — the rule as it is usually first taught.

The honest verdict

Robustness verdict — SPY · 5y · as of 2026-07-13
Failed robustness testingThis backtest did NOT survive robustness testing. The headline numbers overstate it.
Backtest return 6.8%Buy & hold 72.73%Trades 8

Random control

Beat 66.4% of 500 randomly-timed versions of itself (real 6.8% vs random average 3.06%).

FAIL

Out-of-sample

Too few trades to compare (4 in-sample, 4 out).

NOT PROVEN

Significance

t = 0.555 against a 3.5 threshold.

FAIL

Deflated Sharpe

0.6976 — Sharpe of 0.1964.

FAIL

These checks describe how much of this backtest survives statistical scrutiny. Past simulated performance is not a guide to future results, and nothing here is a recommendation to trade.

Over five years the rule fired only 8 times and returned about +7% — against roughly +73% for simply buying and holding SPY. The first thing the numbers say is mundane but important: in a market that trended up for five years, a rule that sits in cash waiting for dips missed almost the entire rise. Mean reversion and a rising trend are natural enemies; buying dips means being flat during the parts that mattered most.

The random control says the sharper thing. The real +7% beat only about 66% of its 500 randomly-timed twins — under the 95% bar, with about one in three scrambled versions doing better. The oversold timing was not measurably better than entering at a random moment and holding the same target. And with only 8 trades, the significance test (t = 0.55) and the out-of-sample split have too little to work with to say anything confidently at all — the engine reports that honestly rather than inventing a verdict from a handful of trades.

What to take from it

"Buy the dip" is not wrong as an instinct — plenty of durable strategies are mean-reverting. But the naïve RSI-below-30 version, on this instrument and window, did three unhelpful things at once: it traded too rarely to prove anything, it underperformed doing nothing by a wide margin, and its entry timing added no measurable edge over random. That is not a reason to dismiss mean reversion; it is a reason to distrust the first, simplest form of any famous idea until it has faced the gates. The same RSI idea on a genuinely range-bound instrument, or with a different threshold, could look different — which is precisely the kind of question the Strategy Editor and the robustness report exist to answer without fooling yourself.

Try it yourself

Try it in Visor →

What to read next