Ten Oscillators Meet Thirty Years of SPY: Williams %R Takes the Robustness Crown

RSI may be the oscillator most traders recognize, yet recognition is not evidence. The narrator asks whether it truly finds short-term pullbacks best, and answers with a frozen experiment: ten oscillators, more than three decades of SPY history, identical entry timing and identical exits. A signal at the close buys SPY at the next open, one position at a time, and the headline comparison forces every oscillator through the same QS exit rule — a position closes at the next open once SPY finishes above the prior day's high. One-, three-, and five-day exits get their own separate checks.

The ranking refuses to crown the biggest profit factor alone. Profit factor divides gross profits by gross losses, so a reading above one means winners out-earned losers across the historical sample. According to Backtest, win rate alone misleads, because a handful of large losses can overturn a pretty hit rate — which is why trade count, average trade, drawdown, neighboring parameter settings, alternative exits, and survival across market eras all enter the verdict.

One market, one rulebook: the test framework

The countdown opens at the bottom. StochRSI lands tenth with a 1.81 factor across 488 trades; CCI follows at 1.97 across 285. Connors RSI takes eighth at 2.12 across 298 trades with a 12.4% maximum drawdown — and QuantifiedStrategies describes that gauge as a short-horizon derivative of Wilder's classic 14-period RSI, built from a two-period lookback blended with trend duration and the size of price moves. The Ultimate Oscillator sits seventh at 1.96 across 479 trades, while IBS, the busiest signaler of the bunch with 631 trades, posts 2.12 in sixth.

Fifth goes to the Chande Momentum Oscillator : the seven-day version under minus 60 produces 219 trades at a 2.42 factor with an average trade near 0.70%. Fidelity's guide presents the CMO as Tushar Chande's momentum gauge swinging between minus 100 and plus 100, where plus 50 flags overbought, minus 50 flags oversold, and divergences that price alone does not draw can warn of a turn. Fourth belongs to the gauge everyone waited for — RSI. A four-day lookback under 30 yields 465 trades at a 1.99 factor, a 0.51% average trade, and 71.8% winners. Context from Guanalyser sharpens the picture: across 483 S&P 500 names from 2006 to 2026, RSI(2) finished first and Stochastic RSI a close second, which explains why short lookbacks dominate dip-buying.

Entering the leaders' group

Third place goes to Stochastic : the smoothed seven-day version under 25 delivers 393 trades at a 2.48 factor, a 0.66% average trade, and a 14.3% maximum drawdown — one of the cleanest blends of strength and sample size in the test. SentimenTrader defines Stochastic as a momentum lens on where the close sits inside its recent range, read through the fast %K line and its %D moving average, with 80 marking stretched upside and 20 stretched downside. TrendInvestorPro adds the close-only StochClose variant, where an overbought spell begins above 80 and ends on a drop back under 50.

Runner-up is QSRSI , and the irony is numerical: its 2.63 headline factor across 271 trades is the highest in the field, paired with a 0.84% average trade and 76.8% winners. It still finishes second, because the title was never awarded on one number. The gap between the glossiest figure and the most survivable one is the entire argument of the test.

The summit and the robustness trial

The winner is Williams %R : the seven-day version under minus 95 manages 288 trades at a 2.61 factor, a 0.82% average trade, 76.4% winners, and about 12.1% maximum drawdown. Alphrex traces the gauge to Larry Williams's 1973 book, noting its inverted zero-to-minus-100 scale — very negative means very depressed — which makes it the mirror image of unsmoothed Stochastic %K. The crown rests on three breakage attempts: the factor reads 3.01 for 1993-2003, 2.51 for 2004-2014, and still 2.28 for 2015-2026; neighboring Williams %R settings stay profitable; and the idea keeps earning across one-, three-, and five-day exits.

The narrator attaches a firm boundary to all of this: the verdict covers SPY only, long side only, short-horizon dip-buying inside one frozen framework, and other markets or styles can shuffle the order. Every figure is historical and measured before trading costs — a record of tendency, not a forecast. The genuine surprise is not that Williams %R works but that RSI works too and still lands fourth: QSRSI's 2.63 tops the sheet while Williams %R's 2.61 takes the title, because the judges scored the whole robustness package.