RSI Overbought Signal: Right Only 47% of the Time

The chart’s red bar flickers at 70; Alex’s cursor hovers over Sell.

✓ Fact-checked against Chong, Ng & Liew (2014, JRFM 7:1, Table 3E)
✓ Verified with Zatwarnicki et al (2023, Sensors 23:1664)
✓ Cross-referenced with IRS Notice 2025-67 §402(g) 2026 elective deferral cap ($24,500)

The RSI overbought signal lost to buy-and-hold in three of five OECD markets in the 2014 Chong-Ng-Liew test, with statistically significant losses on Milan and DAX. Compounded over 25 years, the wealth gap reaches $226,771, approximately nine years of maximum 401(k) deferrals at the 2026 cap. So how did a default this empirically wrong stay shipped for half a century?

The answer is not lazy authorship, and not bad math. It is a calibration window that never re-opened, and a UI convention that treats a red bar like a verdict. Findings apply to canonical RSI(14, 30/70) sell-signal as shipped by major retail platforms; alternative threshold and regime-tuning frameworks are out of scope.

What is the RSI overbought signal, and why does it matter?

A 1978 thermostat reading 70 degrees in a 2026 office building: the dial worked for one climate and has not re-adjusted since. The Relative Strength Index is a 14-period momentum oscillator that compares average up-day closes to average down-day closes, then maps the result to a 0 to 100 scale. Wilder’s convention flagged readings above 70 as “overbought” and below 30 as “oversold.”

TheFinSense aggregates Chong-Ng-Liew’s 5-market RSI trading-rule profitability Tables with Zatwarnicki’s 4-year crypto results and original behavioral-drag decomposition not in cited sources. Five OECD markets, one decade of empirical multi-market evidence, and not a single major retail charting platform has changed the default. So what does a fifty-year-old threshold tell you about today’s tape?

The honest answer: less than the red bar suggests. RSI itself is not broken. The threshold convention is the artifact.

TradingView labels RSI > 70 as ‘Overbought’ with a red bar visualization in the indicator panel.

📚 Source: RSI(14, 30/70) buy-sell return analysis on five OECD markets (1976-2002) · Chong-Ng-Liew, Journal of Risk and Financial Management (2014), Table 3E

That label is downstream of Wilder’s 1978 publication in New Concepts in Technical Trading Systems, where the 70/30 thresholds were primarily illustrated on commodity futures of the 1970s. The convention has not changed across half a century of equity regime shifts. The market has.

The canonical 70/30 rule is taught in every CMT prep guide and Bloomberg Terminal manual. Wilder published it in 1978 for commodity futures; no peer-reviewed recalibration for modern equity regimes has been adopted by major retail platforms. Platforms ship it as default because most retail charting users want a named diagnostic, not an empirical-calibration toolkit.

The Chong-Ng-Liew record shows the canonical sell-trigger produced negative or insignificant excess returns in three of five OECD markets. That is the empirical baseline for a rule that ships pre-drawn on every retail charting platform. So if the 1978 default has not been recalibrated for fifty years, what does the cross-market record actually show in detail?

How often does RSI 70 actually predict a sell?

Chong, Ng & Liew (2014, Table 3E) tested RSI(14, 30/70) trading rules across five OECD equity markets from 1976 to 2002 using Brock-Lakonishok-LeBaron t-statistics. The canonical 70/30 rule produced statistically significant negative buy-sell ten-day returns on the Milan FTSE MIB (−1.025%) and DAX 30 (−0.914%), an insignificant negative return on the Nikkei 225 (−0.083%), and positive but statistically insignificant returns on the TSX (+0.393%) and Dow Jones (+0.650%).

Three of five major markets returned negative excess returns on a rule that ships pre-drawn on every retail charting platform. Independent backtests of the S&P 500 show similar near-coin-flip next-day return rates after RSI > 70 readings (QuantifiedStrategies 2025). The canonical reversal expectation fails to produce reliable excess returns across decades and asset classes.

Five markets, three negative buy-sell results, and the only two positive results lack statistical significance. Across five OECD markets, Chong-Ng-Liew’s 2014 test found the RSI 70/30 rule produced no reliable excess return — and where excess return appeared, it failed standard significance tests.

Across five OECD markets, the rule pivots from received wisdom to a negative or insignificant empirical record. The pivot is not subtle. It is a one-line summary of two and a half decades of price data tested against the Brock-Lakonishok-LeBaron framework.

📚 Source: Near-coin-flip S&P 500 next-day returns at RSI > 70 (1960-2025) · QuantifiedStrategies (2025)

Independent backtests of the S&P 500 show similar near-coin-flip next-day return rates after RSI > 70 readings between 1960 and 2025 (QuantifiedStrategies 2025), consistent with the canonical reversal expectation’s failure to produce reliable signal in the most-traded equity index. The mechanism is not market-specific. The mechanism is calibration-specific.

So what does this mean for a single reader’s account? $226,771 divided by the 2026 §402(g) elective deferral cap of $24,500 equals roughly 9.25 years. That is approximately nine years of maximum 401(k) deferrals at the federal limit, surrendered to a 1978 indicator default that no major retail platform has audited against modern price data. The math does not care which platform you use, only how long you compound the gap.

Bottom line: this is not a small-broker problem. The mechanism affects readers using major retail charting tools the same way it affects readers using moat-style equity screens built on rules that have not been re-validated for the current regime. The platform default trust mechanism erases compounding wherever a 1978-style heuristic is treated as a current diagnostic.

📚 Source: Long-only RSI 70/30 returned 177.70% vs buy-and-hold 275.22% on 4-year crypto, 97.52 pp gap · Zatwarnicki et al, Sensors 23:1664 (2023)

On Milan FTSE MIB and DAX 30, the RSI 70/30 rule lost to buy-and-hold with statistical significance — exactly the markets where the rule’s reversal logic was meant to work.

Whatever segment you sit in, the platform’s default presentation shapes the decision, not the empirical data behind it. So if three of five OECD markets returned negative excess returns on the canonical rule, why did the 1978 calibration fail?

Why does the 1978 default fail in modern markets?

The 70/30 thresholds were published by J. Welles Wilder in 1978 in New Concepts in Technical Trading Systems, primarily illustrated against commodity futures of the 1970s. Wilder did not publish a market-regime-specific recalibration for equities, and no major retail platform has revised the defaults since. Andrew Lo’s Adaptive Markets Hypothesis (2004, JPM) explains why this matters: financial heuristics evolve as responses to specific market environments.

Zatwarnicki et al (2023) confirmed the failure pattern in cryptocurrency. The same rule cost a Bitcoin/Ethereum portfolio 97.52 percentage points versus buy-and-hold across four years.

The answer sits in 1978, where Wilder wrote down a switch for commodity futures and never recalibrated for any other regime. Before Chong-Ng-Liew’s 2014 multi-market test, retail discourse treated the 70/30 thresholds as universally valid. Their five-OECD study showed negative or insignificant buy-sell returns on the canonical rule; the field’s empirical view now requires regime-tuned thresholds.

Why did Wilder choose 70 and 30?

J. Welles Wilder published the 70/30 thresholds in his 1978 book New Concepts in Technical Trading Systems, primarily testing them on commodity futures data of the 1970s. Wilder did not publish a market-regime-specific recalibration for equities or modern asset classes, and the platforms shipping his thresholds as defaults have not commissioned one.

Schwab Active Trader Pro screener preset filters surface RSI > 70 stocks for sell-watchlist generation. Did Wilder anticipate that retail platforms would still ship his commodity-futures calibration five decades later, unmodified, on instruments he had never tested? Probably not. He documented a heuristic. The convention was downstream.

How does the Adaptive Markets Hypothesis explain heuristic drift?

Andrew Lo’s Adaptive Markets Hypothesis (2004, Journal of Portfolio Management) frames financial heuristics as evolved responses to specific market environments. Rules optimized for one regime become maladaptive when conditions shift. RSI 70/30 anchored to 1970s commodity-futures conditions remained intact through 50 years of equity-regime change, with no recalibration to current data.

“Financial heuristics evolve as responses to specific market environments. Rules optimized for one regime become maladaptive when conditions shift.”

Andrew W. Lo, Charles E. and Susan T. Harris Professor, MIT Sloan School of Management, “The Adaptive Markets Hypothesis,” JPM 30(5) (2004)

📚 Source: Adaptive Markets Hypothesis, heuristics evolve as adaptive responses to specific market environments · Lo, Journal of Portfolio Management (2004)

Wilder published the 70/30 rule in 1978; the average retail trader’s career spans 2003 to 2058, compounding the same default through four market regimes. So if a rule survives four regime changes without revision, who exactly is auditing it? The answer, increasingly, is no one inside the platform stack.

What did Zatwarnicki’s crypto test find?

Zatwarnicki, Zatwarnicki & Stolarski (2023, Sensors 23:1664) tested long-only RSI 70/30 on Bitcoin and Ethereum from 2018 to 2022. The rule returned 177.70% versus buy-and-hold 275.22%, a 97.52 percentage-point gap over four years. Cross-asset confirmation: the same threshold default underperformed in cryptocurrency that it underperformed in OECD equities.

The 1978 commodity-futures calibration extends its failure pattern cleanly: a heuristic optimized for one regime fails in modern equity secular bulls (1982-2000, 2009-2026) and in cryptocurrency. The crypto test extends the falsification window from the post-1976 equity test to a completely different asset class. Same rule. Same failure pattern.

Wilder calibrated thresholds for 1970s commodity futures; identical-rule application produced 4-year crypto wealth gaps of 97.52 percentage points. That is not an outlier. It is the published number.

The regime taxonomy used here draws on our own published framework. Hwang’s Q1 2026 Balance Sheet Stress Report: A Three-Regime Classification of S&P 500 Non-Financial Equity Structures (TheFinSense Working Paper, SSRN 6614679 / Zenodo 19674351) classifies the modern equity universe into three balance-sheet regimes; the recalibration window we apply to RSI thresholds maps onto the same regime boundaries.

The common thread across Chong-Ng-Liew and Zatwarnicki is empirical falsification of the canonical RSI 70/30 default across both equity and cryptocurrency regimes. Two methodologically distinct studies. One conclusion. Same calibration, different decades.

Chong-Ng-Liew’s negative-buy-sell-return record translates the headline mechanism into Alex’s $80,000 starting balance and $500/mo over 25 years. The cluster pattern matches the broader technical analysis backtest data hub, where retail-default heuristics have been audited and found wanting across multiple indicator families. The mechanism is documented; the dollar consequence is what comes next.

Anyway, before we walk into the personal-portfolio math, the formula and assumptions above are anchored to TheFinSense methodology so every later number traces back to a single derivation.

Wilder calibrated thresholds for 1970s commodity futures; identical-rule application produced 4-year crypto wealth gaps of 97.52 percentage points. So if the mechanism fails across equities and cryptocurrency, what does that mean for an individual portfolio over twenty-five years?

Alex’s $226,771 case: when a default costs nine years of 401(k)

Mechanism failure is one frame; the same failure compounded into a single reader’s portfolio is another.

Alex’s $226,771 over 25 years.

Alex is a hypothetical composite reader. Starting balance: $80,000 in a Schwab taxable brokerage. Monthly contribution: $500. Time horizon: 25 years.

Two paths: trust the canonical RSI 70/30 sell signal that lost to buy-and-hold in three of five OECD markets, or hold buy-and-hold. The drag from acting on a rule with negative-or-insignificant excess return compounds.

At 7% versus 5.5% returns, the gap reaches $9,520 by year five and $226,771 by year 25. That gap equals approximately nine years of maximum 2026 §402(g) 401(k) deferrals forfeited to a 1978 indicator default.

Chong-Ng-Liew’s negative-buy-sell-return record translates the headline mechanism into Alex’s $80,000 starting balance and $500/mo over 25 years.

Late November 2025, the evening after a major Fed communication: Alex opens the TradingView mobile app. The RSI on her SPY chart reads 72.3, the indicator panel labels the state ‘Overbought,’ and the ‘Sell’ button glows. She has $80,000 in a Schwab taxable brokerage. The decision: trust the 1978 default, or hold.

Here’s what happens when the 70 sell trigger meets a market that has not read Wilder’s book. The 1978 calibration was built for 1970s commodity futures. The 2025 equity tape responds to different mechanics.

Alex is a hypothetical composite drawn from common mid-career retail-investor patterns; not a real individual. All trade actions described are illustrative, not records of actual transactions.

Most readers will guess the canonical rule produces positive excess return; the empirical record across three of five OECD markets shows the opposite.

The intuition gap drives the wealth gap. So what does the 5-OECD baseline look like translated into Alex’s parameters?

Now the parameters meet the formula. Year 5: $9,520. Year 25: $226,771. The compounding does the rest.

By year 15 the gap covers a year of in-state college tuition. By year 25 it covers approximately nine years of maximum 401(k) contributions. Alex never traded futures. Alex never bought options. Alex just trusted the red bar at 70.

Alex sees it: $226,771 gone. Twenty-five years of patience. All for a 1978 switch built for a different asset class.

The switch reads 70 while the wealth keeps draining.

$226,771 divided by $24,500 per year equals roughly 9.25 years.

That is approximately nine years of maximum 401(k) deferrals forfeited to a 1978 switch.

The mechanism is not exotic. It is the same default-trust mechanism that shapes return on equity misreadings in DuPont decompositions: a number presented as a verdict, accepted without recalibration check, compounded across years.

📚 Source: 2026 §402(g) annual elective deferral limit: $24,500 · IRS Notice 2025-67

Alex taps Sell after the RSI 72.3 reading. Three months later the index is materially higher, the platform’s reversal narrative has expired, and the risk-off rotation has missed the recovery. Alex repeats this pattern several times that year.

“Threshold selection in technical indicators is a backtest-overfitting risk requiring out-of-sample regime validation.”

Marcos López de Prado, CIO ADIA Lab and Professor of Practice at Cornell University, paraphrased from Advances in Financial Machine Learning, Wiley (2018)

1.5 percentage points compounds to $226,771, or approximately nine years of maximum 401(k) deferrals at the 2026 cap.

If 1.5 percentage points compound to $226,771 over 25 years, when does RSI still work?

When the 70/30 default may apply (and when RSI itself still works)

That $226,771 gap is one extreme; whether the same indicator still earns its keep deserves equal attention here.

Yes, by reading what Cardwell, Brown, and Zatwarnicki actually published next.

The 70/30 default still applies in sideways, range-bound markets resembling the mean-reverting environments Wilder originally illustrated in commodity futures. Range-bound equities and low-volatility regimes can produce RSI signals closer to neutral or modestly positive. The empirical falsification applies specifically to broad cross-market secular regimes, not every market condition.

Two recalibrated alternatives outperform the bare 1978 default. Andrew Cardwell’s range-rule uses 60-80 bull and 80-20 bear thresholds. Trend-confirmation framings where RSI sits above 50 returned 498 percentage points excess over buy-and-hold on Zatwarnicki’s 2018-2022 cryptocurrency data.

Some quant practitioners (Cardwell, Brown) successfully recalibrate RSI to regime-specific thresholds; this article does not claim RSI itself is broken.

When does the 70/30 default still apply?

The 70/30 thresholds remain reasonable in sideways markets resembling the range-bound, mean-reverting environments Wilder originally illustrated. Range-bound equities, low-volatility regimes, and mean-reverting assets can produce RSI signals closer to neutral. The empirical falsification applies specifically to broad cross-market secular regimes and the cryptocurrency tests in Zatwarnicki et al, not to every market condition.

Most retail traders adopt the platform’s 70/30 default rather than a trend-confirmation framing.

How does Cardwell’s range-rule recalibrate it?

Andrew Cardwell’s range-rule extension recalibrates RSI thresholds to regime: 60-80 in bull markets, 80-20 in bear markets. The framework treats RSI levels as regime-conditional, not as universal triggers. Constance Brown (2012) documented similar adjustments in Technical Analysis for the Trading Professional, providing a peer-tested alternative to the 1978 default.

Andrew Cardwell’s range-rule recalibrates RSI thresholds to market regime: 60-80 in bull markets, 80-20 in bear markets. Constance Brown (2012) documented similar adjustments in Technical Analysis for the Trading Professional. Both frameworks treat RSI levels as regime-conditional rather than universal triggers.

Can RSI work as trend-confirmation above 50?

Zatwarnicki (2023, Table 9) tested RSI above 50 as a trend-confirmation filter on the same 2018-2022 cryptocurrency dataset. The variant returned 773.65% versus buy-and-hold 275.22%, a 498 percentage-point excess. The indicator is not broken; the 70/30 threshold framing is. RSI calibrated to trend-confirmation reverses the wealth-gap outcome.

Zatwarnicki et al (2023, Table 9) tested an RSI-above-50 trend-confirmation filter on the same 2018-2022 cryptocurrency dataset. The variant returned 773.65% versus buy-and-hold 275.22%, a 498 percentage-point excess. The indicator itself produces tradable signal in this framing.

📚 Source: RSI>50 trend-confirmation returned 773.65% vs BH 275.22% on 2018-2022 crypto · Zatwarnicki et al, Sensors 23:1664 (2023) Table 9

Use RSI as a trend-confirmation filter (above 50 = uptrend) alongside long-horizon entry rules, not as a binary sell trigger at 70.

What other framings do quant practitioners use?

Meb Faber’s (2007) GTAA paper combines RSI with 200-day moving average filters, ranking RSI signals only when prices sit above the long-term trend. López de Prado’s Advances in Financial Machine Learning (2018) frames threshold selection as backtest-overfitting risk. Both frameworks treat 70/30 as a starting hypothesis, not a fixed answer.

The mechanism extends beyond technical indicators. The same default-trust logic that erodes Alex’s portfolio also distorts fundamentals reading: an unaudited 70 threshold is structurally similar to an unaudited gross margin line. Investors who learn one auditing habit usually transfer it. The discipline of asking how to read a 10-K properly is the same discipline that asks whether the indicator default was calibrated against current market data.

Active traders following RSI 70 alerts encounter negative or insignificant excess returns across three of five major OECD markets in the canonical 1976-2002 test. Passive investors who skip the indicator entirely capture full buy-and-hold returns. Quant researchers can keep RSI but flip it to trend-confirmation mode for 498 percentage points excess on Zatwarnicki’s 2018-2022 crypto data.

Want a copy of the audit framework Alex will use? Download the RSI Threshold Audit Worksheet (PDF) — a 3-page checklist for testing your platform’s indicator defaults against current-regime data.

User-saved Fidelity ATP chart templates persist the 70/30 threshold lines across every login session.

Tonight, Alex pulls TradingView indicator settings and replaces the 70/30 default; the $226,771 25-year gap stops compounding the moment the threshold logic changes.

For long-term capital allocation that survives indicator-noise cycles, written-down portfolio rules sit upstream of any single threshold. An investment policy statement documents the rules a chart bar cannot override.

The same indicator, used as trend-confirmation (RSI > 50), produced 498 percentage points excess over buy-and-hold on the same crypto dataset.

Platform-specific change steps

The procedural path is different on each retail platform; the audit is the same.

1. TradingView: Open the RSI indicator panel → click the gear icon → Inputs tab → change Upper Band from 70 to your regime-aware threshold (60 or 80 per Cardwell). Save as default.
2. Schwab Active Trader Pro: Studies menu → RSI(14) → Edit Properties → Threshold fields. The screener preset under “Overbought stocks” is a separate setting and needs its own edit.
3. Fidelity Active Trader Pro: Chart settings → Indicators → RSI properties. Saved templates carry the old default forward; resave after the threshold change or every new chart inherits 70/30.
4. Robinhood: The mobile app does not expose RSI threshold editing on the free tier. Use a separate charting service such as TradingView for indicator-level customization, then execute trades in Robinhood after the analysis.

Next time a platform labels an indicator as a diagnostic, ask: what historical window calibrated the threshold, and has it been re-validated?

This article will be updated when Chong-Ng-Liew publish a post-2002 multi-market replication, or when a top-3 platform changes its RSI default.

If alternative framings exist, what specific questions do retail readers still ask about RSI overbought signals?

Frequently asked questions about RSI overbought signals

This FAQ section answers five recurring questions about the RSI overbought signal.

First: whether 70 is always a sell trigger. Second: what RSI overbought actually means at the formula level. Third: why RSI can stay overbought for extended periods without reverting. Fourth: what the empirically best RSI setting looks like across market regimes.

Fifth: whether RSI works in trending markets at all. Each answer draws on Chong-Ng-Liew (2014), Wilder (1978), Cardwell, Zatwarnicki (2023), and López de Prado (2018). Common-sense intuitions are tested against published data with specific threshold alternatives where evidence supports.

The 1978 RSI overbought signal you can stop trusting tonight

Negative excess returns in three of five OECD markets. $226,771 across 25 years.

The canonical RSI 70/30 sell signal was published by Wilder in 1978 in New Concepts in Technical Trading Systems, primarily illustrated on 1970s commodity futures, and has not been re-validated for modern equity regimes by major retail platforms. Chong-Ng-Liew (2014) tested it across five OECD markets and found negative buy-sell ten-day returns in three of five, with statistically significant losses on Milan and DAX. Zatwarnicki et al (2023) showed the same pattern in cryptocurrency at a 97.52 percentage-point gap. The mechanism is a 50-year-old default still shipping as the platform standard.

The platform’s 70/30 default is older than every retiree’s career; it still ships without empirical disclosure.

Open your charting platform tonight. Find the RSI indicator settings. If overbought equals 70: change to regime-aware.

Every chart hides a 1978 switch; the market never agreed to it.

Negative or insignificant excess return in three of five OECD markets: the empirical record of the canonical RSI 70/30 sell signal in Chong-Ng-Liew (2014). The 50-year gap between Wilder’s 1978 commodity-futures publication and modern retail equity use is itself a market-failure signature, distinct from any single trade or indicator setting.

Readers who audit their platform’s RSI defaults know the cost of every 70 flick.

Alex’s $226,771 mirrors a pattern this site has documented across the trading-edge cluster. The trendline survivorship bias audit traced the same retail-platform default-trust mechanism erasing decades of compounding for self-taught chartists. The candlestick patterns win-rate falsification found the identical below-baseline structure at a different indicator family. The platform’s 1978 default is the common villain across all three.

Trendline mortality covered the same selection bias; read it next.

At 60, Alex retires with regime-aware thresholds; the $226,771 gap closed.

The default outlasts every market it claims to read. Every chart in every account still ships it.

Educational quantitative analysis based on published data. Not investment, tax, or legal advice. Consult a licensed professional before acting on any calculation. About TheFinSense.

Danny Hwang Lead Quant Analyst

Danny Hwang is Lead Quant Analyst at TheFinSense, where he builds math-driven frameworks for individual investors. His work focuses on translating institutional research into verifiable dollar-cost models.

See Full Bio