RSI 30 and 70 are round numbers, not results

Short-horizon reversal — securities that fell over days-to-a-month tend to bounce — is a genuine, strongly significant regularity in the academic record. The specific 30/70 oversold and overbought levels are not: they have no derivation, they are Wilder's round numbers from a 1978 trade book. This runs Wilder's own RSI over synthetic prices whose reversal strength you set yourself, then scores every threshold pair on a grid so you can see where 30/70 actually lands.

The rule

RSI lookback14 bars

Computed with Wilder's original smoothing of average gain and average loss — the recursive form from his 1978 book, not a plain simple moving average.

Entry — oversold level30

Buy the synthetic asset when RSI closes below this. Wilder's number was 30; nothing in the literature derives it.

Exit — overbought level70

Max holding bars10 bars

Time stop: leave anyway after this many bars, so trades cannot drift into a long-horizon bet.

Costs & market

One-way cost10 bps

Charged on entry and on exit. Lehmann (1990) himself flagged that reversal profits are cost-sensitive.

Reversal strength60

Scales how strongly each day's return pushes back against the last 5 days' cumulative move — i.e. the negative short-horizon autocorrelation that Jegadeesh (1990) and Lehmann (1990) documented. At 0 there is no reversal at all: a pure random walk, where any RSI edge you see is luck.

Sample#1

Same generator, new random draw. The generator also has clustered volatility, so calm and turbulent spells alternate — that is what produces oversold episodes at all. Watch which threshold cell wins this time.

Mean net return per trade at your thresholds, after cost

one random synthetic sample

Mean per trade, gross

before any cost

Turnover

Rule, net of cost, annualised

Vs buy & hold, annualised

Rank of the 30/70 cell

Best cell, this sample

entry / exit

Times the best threshold pair has changed as you moved things0

Synthetic price & Wilder RSI · last 600 bars of the sample

Price RSI (Wilder smoothing) Entry Exit In a position Oversold zone Overbought zone

Cumulative index, net of cost (100 = start of sample)

RSI rule, net of cost Buy & hold

Threshold heatmap · mean net return per trade (bps), entry × exit

Wilder's 30/70 cell Best cell in this sample Under 3 trades, not scored

What you are looking at: the same 72 threshold pairs scored on one synthetic sample at your current lookback, time stop and cost. Every figure here is hypothetical output from randomly generated data — not an account, not a projection, not a return anyone earned. Three things are worth doing. First, redraw the sample: 30/70 is one cell on a noisy surface, it is usually not the best cell, and the best cell moves somewhere else on almost every draw — which is what it looks like when a threshold has no derivation behind it. Second, set reversal strength to 0. There is then no short-horizon reversal in the data at all, so no oscillator can have genuine predictive power, and whatever green survives is one draw's luck. Third, raise the cost and watch the gap open between the gross figure and the net headline — that gap is the whole documented objection. The low entry levels stay grey because RSI almost never gets that far down: there is nothing there to trade.

What the research actually found

Support level: partial. The phenomenon is documented; the thresholds are not.

“RSI, ATR and ADX come from J. Welles Wilder's 1978 trade book — engineering heuristics, never peer-reviewed at birth. The phenomenon they gesture at, short-horizon reversal, IS academic: securities that fell over days-to-a-month tend to bounce (Jegadeesh 1990; Lehmann 1990).”

Supported.

“Short-term reversal is a real, strongly significant return regularity (Jegadeesh 1990; Lehmann 1990) — the statistical soil oscillators grow in. On 60 years of the London FT30, mechanical RSI and MACD rules beat buy-and-hold before costs in most specifications (Chong & Ng 2008).”

Not supported.

“The reversal profits live in small, high-turnover trades that transaction costs consume — Lehmann himself flagged the cost sensitivity — and the specific 30/70 thresholds have no derivation; they are Wilder's round numbers. No robust study validates RSI levels as a standalone profitable signal in modern, cost-realistic conditions.”

Note carefully what the FT30 result is and is not: Chong & Ng (2008) report those rules beating buy-and-hold before costs. That is the claim in the record, and it is the only one made here.

Wilder, J. Welles (1978). New Concepts in Technical Trading Systems. Trend Research. [Origin of RSI/ATR/ADX — a practitioner book, not peer-reviewed; listed for provenance.]

Jegadeesh, Narasimhan (1990). “Evidence of Predictable Behavior of Security Returns.” Journal of Finance, 45(3), 881–898.

Lehmann, Bruce N. (1990). “Fads, Martingales, and Market Efficiency.” Quarterly Journal of Economics, 105(1), 1–28.

Chong, Terence Tai-Leung, & Ng, Wing-Kam (2008). “Technical analysis and the London stock exchange: testing the MACD and RSI rules using the FT30.” Applied Economics Letters, 15(14), 1111–1114.