Mean reversion williams_r

Williams %R

Range: -100 to 0. Bullish: %R crosses above oversold (-80) from below. Bearish: %R crosses below overbought (-20) from above.

Signal family

Mean reversion — Oscillator-based signals that fire at overbought or oversold extremes — typically fade the prevailing move.

Parameters

Name Description Default Range
period Lookback period 14 5–50
overbought Overbought level -20 -30–-10
oversold Oversold level -80 -90–-70

Historical context

5,522,997 triggers on 24,258 tickers, 1995-09-06 → 2026-05-01. Universe: US large-cap (mcap ≥ $100,000,000, price ≥ $1). Long-only convention: BUY at open T+1, hold the horizon, compare to S&P 500 Equal Weight over the same window.

Methodology footnotes

Benchmarks shown in the detail tables: spxew (S&P 500 Equal Weight — primary, median-stock view, avoids the 2020+ megacap-concentration distortion), spx (S&P 500 cap-weighted, distorted post-2020), msci (MSCI World USD). Per-stock regime tags: trending = ADX(14) ≥ 25, high vol = 20d realized annualized vol ≥ 20%. 1d return = intraday T+1 open→close; 20d = open T+1 to close T+20.

At a glance — alpha vs S&P 500 Equal Weight, US-only

Holding-period sensitivity. Bullish columns: positive = signal worked (long the trigger beat the index). Bearish columns: negative = signal worked (the flagged stock underperformed).

Horizon Bullish α Bearish α
5-day -0.07% +0.00%
20-day -0.12% +0.15%
60-day +0.17% +0.40%
1-year +1.54% +2.40%

Sign flip across horizons. Bullish triggers go from -0.07% (5d) to +1.54% (1y) — short-term fade but longer holding recovers and wins.

Random-date null check (20-day): Bullish: worse than random (p=1.000).
Bearish: worse than random (p=1.000).

Where does WILLIAMS_R actually fire?

The bucket distribution often reveals what the signal really is, regardless of its textbook label. Heavy concentration in "non-trending + high vol" = it's mostly a chop-market event. Heavy in "trending + low vol" = it picks up the smooth grinds. Read the chart before the alpha numbers — context shapes everything that follows.

Williams %R (williams_r) — trigger count distribution by per-stock regime quadrant (trending/non-trending × high/low realized volatility) for , US-only universe

Does it work in every regime?

Trigger alpha split by the host stock's own regime on the trigger date — trending or ranging, high-vol or low-vol. The 20d alpha you'd actually capture if you took the trade. Bars matching your direction's "right" sign (green) = the signal worked in that regime; opposite sign = avoid it there. A signal with one strong-positive bar and three flat ones isn't a "20d alpha" signal — it's a "20d alpha when the stock is X" signal.

Williams %R (williams_r) — mean 20-day alpha versus S&P 500 Equal Weight by per-stock regime quadrant, side by side
Trending + Low vol
Stock in a clean directional move with low realized volatility. Textbook "trend-following paradise" — smooth grind with little whipsaw risk.
Trending + High vol
Violent directional moves — parabolic rallies, crisis selloffs. Trend exists but the path is noisy. Signal timing may be imprecise.
Non-trending + Low vol
Quiet chop, summer doldrums, consolidations. No directional bias but also no big swings — small edges become reliable if they exist at all.
Non-trending + High vol
Choppy and violent — the classical "whipsaw zone" for momentum signals. Crossovers and breakouts fire repeatedly without follow-through.

Does it work in every era?

A multi-year average can hide major instability. The sample splits into three windows: 2015–2019 (pre-COVID), 2020–2022 (pandemic + 2022 bear), and 2023+ (post-ZIRP + AI megacap rally). All three matching your direction's "right" sign = the signal is durable. One era doing all the work = a regime-specific edge that may not repeat. The bigger the variance across eras, the smaller the position you should run.

Williams %R (williams_r) — 20-day alpha split by historical sub-period (2015-2019, 2020-2022, 2023+) to check consistency across market regimes

↑ Bullish triggers

Bench Metric 1d 5d 20d 60d 252d
spx Stock % -0.01% +0.16% +0.96% +2.94% +12.11%
Bench % +0.01% +0.24% +1.30% +3.33% +14.24%
Alpha % -0.04% -0.11% -0.33% -0.41% -2.17%
Median alpha -0.11% -0.35% -1.19% -2.63% -10.71%
Hit rate (α>0) 47.3% 46.2% 44.1% 42.8% 37.5%
p (naive) <0.001 <0.001 <0.001 <0.001 <0.001
p (HAC) <0.001 <0.001 <0.001 <0.001 <0.001
N 2,679,499 2,593,038 2,581,973 2,510,619 2,269,996
msci Stock % -0.01% +0.16% +0.96% +2.94% +12.11%
Bench % +0.06% +0.26% +1.14% +2.97% +11.85%
Alpha % -0.07% -0.11% -0.20% -0.01% -0.07%
Median alpha -0.14% -0.37% -1.07% -2.28% -8.51%
Hit rate (α>0) 46.4% 46.0% 44.6% 43.7% 39.8%
p (naive) <0.001 <0.001 <0.001 0.2549 0.0273
p (HAC) <0.001 <0.001 <0.001 0.5771 0.5800
N 2,662,426 2,572,370 2,548,646 2,491,323 2,244,934
spxew Stock % -0.01% +0.16% +0.96% +2.94% +12.11%
Bench % +0.04% +0.21% +1.09% +2.76% +10.48%
Alpha % -0.06% -0.07% -0.12% +0.17% +1.54%
Median alpha -0.12% -0.30% -0.94% -1.99% -6.89%
Hit rate (α>0) 47.0% 46.8% 45.3% 44.4% 41.5%
p (naive) <0.001 <0.001 <0.001 <0.001 <0.001
p (HAC) <0.001 <0.001 <0.001 <0.001 <0.001
N 2,662,708 2,572,563 2,546,037 2,483,251 2,239,567
Distribution of all 20d alpha outcomes for this direction. Median and winsorized mean shown.
Williams %R (williams_r) — bullish 20-day alpha histogram showing distribution of per-trigger returns
Observed 20d alpha (vertical line) against the null distribution of random-date firing. If the line is deep inside the null cloud, the signal adds no information. If it sits in a tail, the signal is doing real work in that direction.
Williams %R (williams_r) — bullish 20-day observed alpha versus random-date permutation null (200 iterations)
Permutation null detail — all horizons × both benchmarks
200-iteration null: for each ticker, sample N random dates from its history (matching observed trigger count) and compute the same alpha. Both observed and null are baseline-centered per ticker (each ticker's own baseline alpha is subtracted), so the null distribution is centered on ~0 and the comparison tests signal effect alone — not the universe-selection lift that all surviving large-caps share. pperm = one-sided fraction of null iters with mean in the "signal was right" tail (right for bullish, left for bearish).
Horizon Bench Observed lift Null mean Null 95% CI pperm
1d spx +0.08% +0.09% [+0.08%, +0.09%] 1.000
1d msci +0.08% +0.09% [+0.09%, +0.09%] 1.000
1d spxew +0.08% +0.08% [+0.08%, +0.08%] 0.005
5d spx +0.33% +0.37% [+0.36%, +0.38%] 1.000
5d msci +0.32% +0.38% [+0.37%, +0.38%] 1.000
5d spxew +0.33% +0.36% [+0.35%, +0.36%] 1.000
20d spx +1.03% +1.19% [+1.17%, +1.20%] 1.000
20d msci +1.04% +1.20% [+1.19%, +1.22%] 1.000
20d spxew +1.01% +1.16% [+1.14%, +1.17%] 1.000
60d spx +2.66% +2.55% [+2.52%, +2.57%] 0.005
60d msci +2.64% +2.57% [+2.54%, +2.60%] 0.005
60d spxew +2.49% +2.48% [+2.46%, +2.51%] 0.174
252d spx +5.32% +5.16% [+5.11%, +5.21%] 0.005
252d msci +5.15% +5.11% [+5.06%, +5.15%] 0.050
252d spxew +4.97% +4.82% [+4.77%, +4.86%] 0.005

Example triggers on US large-caps (2023+, mcap ≥ $30B)

Six recent bullish WILLIAMS_R triggers on US mega-caps. Top three: the signal's best outcomes. Bottom three: the worst. Catalyst-driven outliers (|α| > 25%) excluded so what's left is the signal's own typical good and bad days, not earnings shocks.

Strongest outcomes (what WILLIAMS_R looks like when it works)
Williams %R (williams_r) — strongest bullish trigger #1 on a US mega-cap, showing the stock's OHLC candles, moving averages, signal-specific overlays, and the 20-day forward window Williams %R (williams_r) — strongest bullish trigger #2 on a US mega-cap, showing the stock's OHLC candles, moving averages, signal-specific overlays, and the 20-day forward window Williams %R (williams_r) — strongest bullish trigger #3 on a US mega-cap, showing the stock's OHLC candles, moving averages, signal-specific overlays, and the 20-day forward window
Weakest outcomes (what WILLIAMS_R looks like when it fails)
Williams %R (williams_r) — weakest bullish trigger #1 on a US mega-cap, showing the 20-day forward window where the signal failed Williams %R (williams_r) — weakest bullish trigger #2 on a US mega-cap, showing the 20-day forward window where the signal failed Williams %R (williams_r) — weakest bullish trigger #3 on a US mega-cap, showing the 20-day forward window where the signal failed
Stock-regime quadrants (2×2 per-stock, 20d alpha detail table)
Each quadrant groups triggers by the stock's own ADX(14) and RV(20) at the trigger date — the textbook conditioning variable (not market-level). Stock %, bench %, alpha %, and HAC p-value shown for each benchmark.
Quadrant N Stock % (spx) Bench % (spx) Alpha % (spx) p (HAC) Stock % (msci) Bench % (msci) Alpha % (msci) p (HAC) Stock % (spxew) Bench % (spxew) Alpha % (spxew) p (HAC)
Trending + Low vol Clean directional grind, low whipsaw 168,188 +0.26% +1.05% -0.77% <0.001 +0.26% +0.90% -0.61% <0.001 +0.26% +0.78% -0.47% <0.001
Trending + High vol Crisis selloff or parabolic rally 912,248 +1.45% +1.48% -0.05% 0.0017 +1.45% +1.26% +0.12% <0.001 +1.45% +1.18% +0.19% <0.001
Non-trending + Low vol Quiet chop, summer doldrums 331,385 +0.27% +1.02% -0.75% <0.001 +0.27% +0.90% -0.62% <0.001 +0.27% +0.76% -0.47% <0.001
Non-trending + High vol Classical "whipsaw zone" for momentum 1,347,279 +0.91% +1.28% -0.34% <0.001 +0.91% +1.15% -0.25% <0.001 +0.91% +1.10% -0.18% <0.001
Sub-period breakdown table (20d alpha)
Historical clustering check. If alpha concentrates in one era, the signal's robustness is questionable.
Period N Alpha % (spx) p (HAC) Alpha % (msci) p (HAC) Alpha % (spxew) p (HAC)
2015-2019 2015-01-01 → 2020-01-01 824,394 -0.45% <0.001 -0.27% <0.001 -0.34% <0.001
2020-2022 2020-01-01 → 2023-01-01 835,687 -0.14% <0.001 +0.01% 0.5361 -0.36% <0.001
2023-2026 2023-01-01 → 2099-01-01 1,098,355 -0.38% <0.001 -0.31% <0.001 +0.26% <0.001

↓ Bearish triggers negative alpha = signal was right (stock underperformed market)

Bench Metric 1d 5d 20d 60d 252d
spx Stock % +0.03% +0.21% +0.83% +2.62% +12.31%
Bench % +0.02% +0.23% +0.92% +3.02% +13.74%
Alpha % +0.00% -0.01% -0.05% -0.39% -1.46%
Median alpha -0.06% -0.29% -0.97% -2.60% -9.87%
Hit rate (α>0) 48.3% 46.8% 45.1% 42.8% 38.2%
p (naive) 0.0037 0.0035 <0.001 <0.001 <0.001
p (HAC) 0.0038 0.0069 <0.001 <0.001 <0.001
N 2,682,061 2,597,938 2,569,361 2,516,507 2,223,847
msci Stock % +0.03% +0.21% +0.83% +2.62% +12.31%
Bench % +0.04% +0.21% +0.80% +2.59% +11.32%
Alpha % -0.01% +0.02% +0.07% +0.06% +0.90%
Median alpha -0.08% -0.27% -0.85% -2.18% -7.49%
Hit rate (α>0) 48.0% 47.1% 45.7% 43.9% 40.8%
p (naive) <0.001 <0.001 <0.001 <0.001 <0.001
p (HAC) <0.001 <0.001 <0.001 0.0388 <0.001
N 2,663,410 2,579,722 2,562,480 2,500,319 2,213,542
spxew Stock % +0.03% +0.21% +0.83% +2.62% +12.31%
Bench % +0.04% +0.22% +0.74% +2.26% +10.00%
Alpha % -0.02% +0.00% +0.15% +0.40% +2.40%
Median alpha -0.07% -0.26% -0.73% -1.78% -5.87%
Hit rate (α>0) 48.2% 47.2% 46.3% 45.0% 42.5%
p (naive) <0.001 0.3638 <0.001 <0.001 <0.001
p (HAC) <0.001 0.3999 <0.001 <0.001 <0.001
N 2,655,513 2,557,182 2,544,325 2,485,306 2,196,052
Distribution of all 20d alpha outcomes for this direction. Median and winsorized mean shown.
Williams %R (williams_r) — bearish 20-day alpha histogram showing distribution of per-trigger returns
Observed 20d alpha (vertical line) against the null distribution of random-date firing. If the line is deep inside the null cloud, the signal adds no information. If it sits in a tail, the signal is doing real work in that direction.
Williams %R (williams_r) — bearish 20-day observed alpha versus random-date permutation null (200 iterations)
Permutation null detail — all horizons × both benchmarks
200-iteration null: for each ticker, sample N random dates from its history (matching observed trigger count) and compute the same alpha. Both observed and null are baseline-centered per ticker (each ticker's own baseline alpha is subtracted), so the null distribution is centered on ~0 and the comparison tests signal effect alone — not the universe-selection lift that all surviving large-caps share. pperm = one-sided fraction of null iters with mean in the "signal was right" tail (right for bullish, left for bearish).
Horizon Bench Observed lift Null mean Null 95% CI pperm
1d spx +0.11% +0.08% [+0.08%, +0.08%] 1.000
1d msci +0.13% +0.09% [+0.08%, +0.09%] 1.000
1d spxew +0.11% +0.07% [+0.07%, +0.08%] 1.000
5d spx +0.37% +0.34% [+0.33%, +0.34%] 1.000
5d msci +0.40% +0.34% [+0.33%, +0.35%] 1.000
5d spxew +0.35% +0.32% [+0.31%, +0.33%] 1.000
20d spx +1.13% +1.08% [+1.06%, +1.09%] 1.000
20d msci +1.14% +1.09% [+1.08%, +1.10%] 1.000
20d spxew +1.10% +1.05% [+1.03%, +1.06%] 1.000
60d spx +2.17% +2.34% [+2.32%, +2.37%] 0.005
60d msci +2.19% +2.36% [+2.34%, +2.39%] 0.005
60d spxew +2.20% +2.27% [+2.25%, +2.30%] 0.005
252d spx +4.47% +4.79% [+4.73%, +4.84%] 0.005
252d msci +4.50% +4.74% [+4.68%, +4.79%] 0.005
252d spxew +4.25% +4.44% [+4.38%, +4.49%] 0.005

Example triggers on US large-caps (2023+, mcap ≥ $30B)

Six recent bearish WILLIAMS_R triggers on US mega-caps. Top three: the signal's best outcomes. Bottom three: the worst. Catalyst-driven outliers (|α| > 25%) excluded so what's left is the signal's own typical good and bad days, not earnings shocks.

Strongest outcomes (what WILLIAMS_R looks like when it works)
Williams %R (williams_r) — strongest bearish trigger #1 on a US mega-cap, showing the stock's OHLC candles, moving averages, signal-specific overlays, and the 20-day forward window Williams %R (williams_r) — strongest bearish trigger #2 on a US mega-cap, showing the stock's OHLC candles, moving averages, signal-specific overlays, and the 20-day forward window Williams %R (williams_r) — strongest bearish trigger #3 on a US mega-cap, showing the stock's OHLC candles, moving averages, signal-specific overlays, and the 20-day forward window
Weakest outcomes (what WILLIAMS_R looks like when it fails)
Williams %R (williams_r) — weakest bearish trigger #1 on a US mega-cap, showing the 20-day forward window where the signal failed Williams %R (williams_r) — weakest bearish trigger #2 on a US mega-cap, showing the 20-day forward window where the signal failed Williams %R (williams_r) — weakest bearish trigger #3 on a US mega-cap, showing the 20-day forward window where the signal failed
Stock-regime quadrants (2×2 per-stock, 20d alpha detail table)
Each quadrant groups triggers by the stock's own ADX(14) and RV(20) at the trigger date — the textbook conditioning variable (not market-level). Stock %, bench %, alpha %, and HAC p-value shown for each benchmark.
Quadrant N Stock % (spx) Bench % (spx) Alpha % (spx) p (HAC) Stock % (msci) Bench % (msci) Alpha % (msci) p (HAC) Stock % (spxew) Bench % (spxew) Alpha % (spxew) p (HAC)
Trending + Low vol Clean directional grind, low whipsaw 209,277 +0.45% +0.69% -0.22% <0.001 +0.45% +0.54% -0.06% 0.0015 +0.45% +0.38% +0.11% <0.001
Trending + High vol Crisis selloff or parabolic rally 1,042,153 +1.01% +0.97% +0.08% <0.001 +1.01% +0.84% +0.23% <0.001 +1.01% +0.76% +0.32% <0.001
Non-trending + Low vol Quiet chop, summer doldrums 337,323 +0.49% +0.75% -0.25% <0.001 +0.49% +0.60% -0.10% <0.001 +0.49% +0.50% +0.01% 0.5731
Non-trending + High vol Classical "whipsaw zone" for momentum 1,175,111 +0.89% +0.97% -0.05% 0.0011 +0.89% +0.88% +0.06% 0.0001 +0.89% +0.85% +0.08% <0.001
Sub-period breakdown table (20d alpha)
Historical clustering check. If alpha concentrates in one era, the signal's robustness is questionable.
Period N Alpha % (spx) p (HAC) Alpha % (msci) p (HAC) Alpha % (spxew) p (HAC)
2015-2019 2015-01-01 → 2020-01-01 832,308 -0.48% <0.001 -0.28% <0.001 -0.25% <0.001
2020-2022 2020-01-01 → 2023-01-01 822,478 +0.38% <0.001 +0.51% <0.001 +0.14% <0.001
2023-2026 2023-01-01 → 2099-01-01 1,108,442 -0.04% 0.0276 +0.04% 0.0328 +0.47% <0.001

Methodology and caveats

How to read. Entry at open of T+1 (one trading day after the signal fires on close of T). 20d = open T+1 to close T+20. Alpha = stock return − benchmark return over the same window (Convention A, single-sided, textbook). For bullish triggers, POSITIVE alpha = signal was right. For bearish triggers, NEGATIVE alpha = signal was right (stock underperformed market). No sign-flipping; the direction of the bet determines what "good" looks like. Per-stock regime is each stock's own ADX(14) and RV(20) at the trigger date — not market-wide state.

Three p-values, three robustness tests. (a) p_naive: scipy one-sample t-test on winsorized alphas. Optimistic because overlapping 20d windows on the same ticker inflate effective N. (b) p_hac: Newey-West HAC with lag = horizon — corrects for the overlap and is the academic-finance standard. (c) p_perm: fraction of 200 random-date null iterations with mean ≥ observed. Tests whether the signal beats random date selection at all. A signal that clears all three (pnaive, phac, pperm all < 0.05) has real information; a signal that fails pperm has zero edge even if the t-test says "significant."

Caveats. (i) Universe reflects today's active tickers; delisted losers pruned → survivorship bias. (ii) Mcap ≥ $100M filter uses today's snapshot, not point-in-time — mild lookahead on which stocks enter the sample, not on returns. (iii) Means and p-values use winsorized alphas (1/99 percentile) to prevent data errors from dominating. Medians and hit rates use raw data. (iv) Zero transaction costs assumed. Realistic bid-ask + commissions remove 20–40bps from 20d alpha on US large-caps, more on small-cap. Sub-20bps alpha is noise in practice. (v) Past performance does not predict future results.

How to use this

1 · When to reach for this signal

Caution recommended. Bullish 20d alpha is -0.33% and worse than random — triggering on random dates would have produced better long-side returns. Bearish 20d alpha is -0.05% and worse than random . Either direction fails the "beats random" test. Don't use Williams %R as a standalone entry trigger. It may still be useful as part of a composite (section 4).

2 · When it works — the setups that drive it

  • Best bullish setup: Trending + High vol — alpha -0.05% / 20d on 912,248 historical triggers.
  • Best bearish setup: Trending + High vol — alpha +0.08% / 20d on 1,042,153 historical triggers.
  • Best era for bullish: 2020-2022 — alpha -0.14% / 20d.
  • Best era for bearish: 2020-2022 — alpha +0.38% / 20d.

3 · When it fails — common false positives

  • Weakest bullish cell: Trending + Low vol — alpha -0.77% / 20d on 168,188 triggers.
  • Weakest bearish cell: Non-trending + Low vol — alpha -0.25% / 20d on 337,323 triggers.
  • Worst era for bullish: 2015-2019 — alpha -0.45% / 20d.
  • Worst era for bearish: 2015-2019 — alpha -0.48% / 20d.

Signal-specific failure patterns

Bullish fails systematically, deepens over horizon
Williams %R bullish (exit of oversold < −80 zone) produces α=−0.17 at 20d (p(HAC)<1e-10, p_perm=1.000), widening to −0.29 at 60d. 416,045 triggers is a very large sample; the failure is statistically robust, not noise. The overbought/oversold framing doesn't map onto directional forward returns in this universe.
evidence: bullish vs SPX: 20d α=−0.17 N=416k p_perm=1.000; 60d α=−0.29 p_perm=1.000
Bearish is the real signal — compounds strongly at 60d
Williams bearish (exit of overbought > −20 zone) at α=−0.22 at 20d (p(HAC)<1e-18) widens to −0.58 at 60d (p<1e-17). Both horizons have p_perm=0.005. The signal isolates a persistent 1-3 month underperformance in stretched-high stocks. Largest N (449k) of any bearish oscillator — reliable because of scale.
evidence: bearish vs SPX: 20d α=−0.22 p_perm=0.005; 60d α=−0.58 p_perm=0.005
2020-2022 was the noise period (like every oscillator)
2015-2019 bearish α=−0.36, 2020-2022 flips to +0.13, 2023-2026 back to −0.38. Same structure as other oscillators — QE/rate-cut driven liquidity rendered overbought signals non-predictive. Bearish Williams should be down-weighted during obvious liquidity surges.
evidence: bearish 20d vs SPX: 2015-19 −0.36, 2020-22 +0.13, 2023-26 −0.38

4 · Pairing inside a screen

The statements below describe how this signal relates to others by construction — which indicator family it belongs to, and where same-family redundancy might reduce the independence of evidence inside a Daily Report. These are taxonomic classifications drawn from standard technical-analysis texts; they are not pairing backtests. A multi-signal convergence backtest is planned but not yet run.

Oscillator-family redundancy

Williams %R belongs to the momentum-oscillator family alongside RSI, Stochastics, and CCI — each is constructed from closing price over a short lookback, normalised to a bounded range (Murphy, Technical Analysis of the Financial Markets, 1999; Pring, Technical Analysis Explained, 5th ed. 2014; Kirkpatrick & Dahlquist, Technical Analysis, 3rd ed. 2015). Stacking two or more of these in the same direction within a single Daily Report produces correlated rather than independent evidence.

What would likely rescue this signal

This block calls out the data or conditions that could turn a technically weak signal into a usable one in a composite screen. Based on signal mechanics and the observed failure patterns above; individual combinations are not yet backtested.

  • Pair with trend filter — Williams bearish alone has edge; Williams bearish + close below 50DMA (downtrend filter) would likely concentrate the alpha. Testable without new data.
  • Extend holds — 60d alpha is nearly 3x the 20d alpha. Holding the full signal window captures substantially more. Time stops > price stops.

See also Why technical-only signals don't survive on their own for the broader argument.

5 · Before you act — a 5-point checklist

  1. Normal trading day? Rule out earnings (within ±3 days), ex-dividend, or known corporate-action dates — the signal is almost certainly reading noise, not momentum, in those windows.
  2. Where is price vs its own 50 / 200 DMA? A mean-reversion signal firing against the long-term trend (e.g. oversold in a clean uptrend) is much more reliable than one firing with it.
  3. What's the sector breadth doing? An isolated signal in a broadly down-trending sector is a lower-confidence setup than one firing with the rest of its peer group.
  4. Is ADV20 enough for your size? If the trigger is on a $500M name and you want to move $1M notional, you're the tape. Consider adv20d ≥ 5% of your intended position.
  5. What invalidates you? Define a price level (for longs: a close below the trigger-day low; for shorts: close above the trigger-day high) and honor it. The backtest alpha is an average; any one trade can be at either tail.

Execution notes

Bearish is the tradable side. 449k triggers over 2015-2026 — very large dataset. Alpha compounds from 20d to 60d (−0.22 → −0.58). Entry open T+1. Bullish Williams %R has not produced positive alpha in any horizon, any regime, any sub-period — skip.