Waiting for RSI Confirmation Cut Returns in Six-Market Study

Waiting for the RSI cross-back confirmation produced no statistically significant gain in any of 12 market-and-setting comparisons, and it significantly reduced returns in four of them. That's the core result of a TrendAdvantage study published on TradingView, covering 293 large stocks across six markets from 30 October 2006 to 30 October 2026.

The study tested two rules on the same stocks. Rule A buys as soon as RSI(14) drops into oversold and sells when it rises into overbought. Rule B waits for confirmation: buy only after RSI drops into oversold and then climbs back out; sell only after it enters overbought and falls back below. A earlier study on 107 US stocks found the confirmation step reduced returns; this repeat asks whether the result holds elsewhere.

What did the test cover?

• Markets: Spain, Mexico, Brazil, Germany, Japan and Korea, each studied separately
• Stocks: the 50 largest eligible by market capitalisation (43 in Mexico)
• Settings: RSI(14) at 30/70 and 35/65, fixed in advance, no optimisation
• Trading: long-only, daily bars, 0.10% transaction cost per side, dividend-adjusted local-currency prices
• Validation: Wilcoxon signed-rank tests with Benjamini–Hochberg correction, plus a calendar-time block bootstrap

Execution used signals at the close and fills at the next open, with 95% of equity per trade. Stocks needed at least 10 years of data and no more than 5% of trading days missing.

Where did waiting cost the most?

Korea showed the clearest penalty. At 30/70, Cross beat Cross-Back on 43 of 50 stocks; at 35/65, on 42 of 50. Both were statistically significant, and waiting cost roughly 1.4 to 1.5 percentage points a year.

Japan followed. Cross won on 36 of 50 stocks at 30/70, significant, with waiting on the buy alone costing 0.73 points a year at 30/70 and 0.93 at 35/65 — both significant. Mexico at 30/70 was the fourth significant comparison: Cross won 29 of 43 stocks, with buy-side waiting costing 0.50 points.

Spain, Brazil and Germany showed no significant difference. Buy and hold beat both RSI rules in every market, with median CAGRs ranging from 6.07% in Spain to 11.39% in Japan, against RSI-rule medians of roughly 1.4% to 5.5%.

What drives the cost of waiting?

Most of the damage came from waiting on the buy, not the sell. Buy-side waiting carried a significant cost in five comparisons; sell-side waiting did so only once (Korea at 30/70). This matches the earlier US study, which put the cost at about 0.9 points a year at 30/70 and 1.2 at 35/65.

Waiting also rarely bought cheaper. Matching each Cross trade with the Cross-Back trade from the same oversold episode, Cross-Back bought at a lower price in only 29% to 36% of paired trades — roughly three times in ten. The US figure was about 31%.

Risk didn't improve either. Cross-Back had a significantly deeper drawdown in one case (Korea at 30/70) and was significantly worse on the Martin ratio in six of 12 comparisons and on the Sharpe ratio in five. It was significantly better on none.

How robust are the results?

Equal-weight portfolios favoured Cross in five of six markets at 30/70 — Germany was the exception by 0.03 points — and in all six at 35/65. Sub-periods (test start–2013, 2014–2019, 2020–2026) showed Cross-Back trailing in every earlier period, though the gap narrowed after 2020, when Cross-Back led in three of 12 comparisons. Varying transaction costs between 0% and 0.25% per side changed the estimated cost of waiting by 0.02 points a year or less.

The authors flag real limits: survivorship bias from using today's largest listed companies, idle cash earning 0% in high-interest markets like Brazil and Mexico, a single uniform transaction cost despite Korea's sell-side tax, and nominal local-currency returns. These are historical results, not forecasts.

The bottom line, per the study: "Across the 293 large stocks tested in six markets, every statistically significant difference between the two rules favoured acting when the threshold was first crossed." For traders set on confirmation, the data says wait on the sell, not the buy — and test any rule on your own instruments before trusting it.