failed_double_top
Failed Double Top Breakout
Bearish reversal: price broke above resistance (double top breakout) but then falls back below the resistance level. Bulls are trapped. Failure threshold normalized by daily volatility.
Signal family
Pattern — Formal chart-pattern detectors (double tops / bottoms, failed breakouts, HH/HL structure).
Parameters
| Name | Description | Default | Range |
|---|---|---|---|
| peak_order | Peak detection window | 15 | 5–25 |
| tolerance_zscore | Tolerance (z-scores of daily vol) | 1.5 | 0.5–3.0 |
| failure_window | Max days for failure after breakout | 60 | 20–120 |
Historical context
72,982 triggers on 19,843 tickers, 1989-02-28 → 2026-05-01. Universe: global — all covered exchanges (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, global universe
Holding-period sensitivity. Bullish columns: positive = signal worked (long the trigger beat the index). Bearish columns: negative = signal worked (the flagged stock underperformed).
| Horizon | Bearish α |
|---|---|
| 5-day | +0.31% |
| 20-day | +0.53% |
| 60-day | +0.37% |
| 1-year | +3.43% |
Failed Double Top Breakout is a single-direction signal — only the bearish side is meaningful.
Where does FAILED_DOUBLE_TOP 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.
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. This is the 20-day alpha a trade taken on the trigger would have captured. This signal is bearish-only, so negative bars mark the regimes where it worked — a bearish trigger is right when the stock underperforms. Positive bars mark regimes to avoid. One strong bar beside three flat ones is not a "20-day alpha" signal — it is a "20-day alpha when the stock is X" signal. Bar labels carry the sample size; a cell built on a handful of triggers is noise, not a regime finding.
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 carrying the sign that favours the trade means the signal is durable; one era doing all the work means a regime-specific edge that may not repeat. The greater the variance across eras, the smaller the position it justifies.
↓ Bearish triggers negative alpha = signal was right (stock underperformed market)
| Bench | Metric | 1d | 5d | 20d | 60d | 252d |
|---|---|---|---|---|---|---|
| spxew | Stock % | +0.13% | +0.48% | +1.25% | +2.89% | +13.78% |
| Bench % | +0.03% | +0.17% | +0.72% | +2.46% | +10.21% | |
| Alpha % | +0.10% | +0.31% | +0.53% | +0.37% | +3.43% | |
| Median alpha | +0.03% | -0.04% | -0.56% | -2.10% | -6.03% | |
| Hit rate (α>0) | 50.5% | 49.6% | 47.4% | 44.4% | 42.8% | |
| p (naive) | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 | |
| p (HAC) | <0.001 | <0.001 | <0.001 | 0.0002 | <0.001 | |
| N | 70,547 | 68,303 | 67,835 | 64,962 | 57,602 | |
| spx | Stock % | +0.13% | +0.48% | +1.25% | +2.89% | +13.78% |
| Bench % | +0.01% | +0.25% | +1.11% | +3.33% | +14.32% | |
| Alpha % | +0.12% | +0.23% | +0.17% | -0.53% | -0.72% | |
| Median alpha | +0.04% | -0.13% | -0.94% | -3.12% | -10.38% | |
| Hit rate (α>0) | 50.8% | 48.8% | 45.7% | 41.9% | 38.3% | |
| p (naive) | <0.001 | <0.001 | 0.0002 | <0.001 | 0.0009 | |
| p (HAC) | <0.001 | <0.001 | 0.0003 | <0.001 | 0.2296 | |
| N | 71,065 | 68,947 | 68,569 | 65,687 | 58,223 | |
| msci | Stock % | +0.13% | +0.48% | +1.25% | +2.89% | +13.78% |
| Bench % | +0.02% | +0.24% | +0.94% | +3.00% | +11.75% | |
| Alpha % | +0.11% | +0.26% | +0.30% | -0.06% | +1.32% | |
| Median alpha | +0.03% | -0.10% | -0.81% | -2.70% | -8.01% | |
| Hit rate (α>0) | 50.5% | 49.1% | 46.3% | 42.9% | 40.6% | |
| p (naive) | <0.001 | <0.001 | <0.001 | 0.4441 | <0.001 | |
| p (HAC) | <0.001 | <0.001 | <0.001 | 0.5331 | 0.0294 | |
| N | 70,813 | 68,650 | 67,998 | 65,408 | 57,727 |
Permutation null detail — all horizons × each benchmark
| Horizon | Bench | Observed lift | Null mean | Null 95% CI | pperm |
|---|---|---|---|---|---|
| 1d | spxew | +0.23% | +0.08% | [+0.06%, +0.10%] | 1.000 |
| 1d | spx | +0.23% | +0.09% | [+0.07%, +0.11%] | 1.000 |
| 1d | msci | +0.25% | +0.09% | [+0.07%, +0.11%] | 1.000 |
| 5d | spxew | +0.68% | +0.35% | [+0.30%, +0.40%] | 1.000 |
| 5d | spx | +0.64% | +0.37% | [+0.32%, +0.42%] | 1.000 |
| 5d | msci | +0.67% | +0.37% | [+0.33%, +0.42%] | 1.000 |
| 20d | spxew | +1.56% | +1.13% | [+1.04%, +1.22%] | 1.000 |
| 20d | spx | +1.42% | +1.16% | [+1.07%, +1.25%] | 1.000 |
| 20d | msci | +1.44% | +1.17% | [+1.08%, +1.26%] | 1.000 |
| 60d | spxew | +2.29% | +2.37% | [+2.21%, +2.52%] | 0.149 |
| 60d | spx | +2.13% | +2.45% | [+2.30%, +2.60%] | 0.005 |
| 60d | msci | +2.17% | +2.46% | [+2.31%, +2.61%] | 0.005 |
| 252d | spxew | +5.02% | +4.65% | [+4.33%, +4.92%] | 0.995 |
| 252d | spx | +4.91% | +5.00% | [+4.67%, +5.30%] | 0.284 |
| 252d | msci | +4.70% | +4.92% | [+4.61%, +5.22%] | 0.095 |
Example triggers on US large-caps (2023+, mcap ≥ $30B)
Six recent bearish FAILED_DOUBLE_TOP triggers on US mega-caps, ranked within this sample of six: the top three and the bottom three, with extreme outliers excluded. Ranking is relative, not absolute — where the signal did well across the sampled names, even the bottom three can have beaten the benchmark, so the alpha printed on each panel is what settles it. Both groups are tail outcomes by construction; read them as the range, not the typical result.
Best three of the six sampled
Weakest three of the six sampled — not necessarily losses
Stock-regime quadrants (2×2 per-stock, 20d alpha detail table)
| Quadrant | N | Stock % (spxew) | Bench % (spxew) | Alpha % (spxew) | p (HAC) | Stock % (spx) | Bench % (spx) | Alpha % (spx) | p (HAC) | Stock % (msci) | Bench % (msci) | Alpha % (msci) | p (HAC) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Trending + Low vol Clean directional grind, low whipsaw | 2,228 | +0.54% | +0.51% | -0.00% | 0.9945 | +0.54% | +1.03% | -0.46% | 0.0040 | +0.54% | +0.81% | -0.25% | 0.0983 |
| Trending + High vol Crisis selloff or parabolic rally | 39,552 | +1.64% | +0.70% | +0.98% | <0.001 | +1.64% | +1.11% | +0.58% | <0.001 | +1.64% | +0.96% | +0.72% | <0.001 |
| Non-trending + Low vol Quiet chop, summer doldrums | 4,189 | +0.28% | +0.66% | -0.36% | 0.0005 | +0.28% | +1.11% | -0.80% | <0.001 | +0.28% | +0.91% | -0.61% | <0.001 |
| Non-trending + High vol Classical "whipsaw zone" for momentum | 27,012 | +0.91% | +0.72% | +0.12% | 0.0940 | +0.91% | +1.07% | -0.18% | 0.0105 | +0.91% | +0.88% | -0.07% | 0.3463 |
Sub-period breakdown table (20d alpha)
| Period | N | Alpha % (spxew) | p (HAC) | Alpha % (spx) | p (HAC) | Alpha % (msci) | p (HAC) |
|---|---|---|---|---|---|---|---|
| 2015-2019 2015-01-01 → 2020-01-01 | 17,841 | +0.03% | 0.6911 | -0.19% | 0.0151 | +0.04% | 0.5605 |
| 2020-2022 2020-01-01 → 2023-01-01 | 22,296 | +0.26% | 0.0053 | +0.33% | 0.0004 | +0.50% | <0.001 |
| 2023-2026 2023-01-01 → 2099-01-01 | 32,817 | +1.01% | <0.001 | +0.27% | 0.0004 | +0.32% | <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: one-sided fraction of 200 random-date null iterations falling in the “signal was right” tail (mean ≥ observed for bullish; mean ≤ observed for bearish). 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 not beaten random timing whatever the t-test says — and because the test is one-sided, a pperm up at its 1.000 ceiling is not "no edge" but inverted edge: every random draw served the claimed direction better than the trigger dates did.
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
Not a standalone entry trigger at 20 days. This signal fires bearish-only — there is no bullish variant. Bearish 20d alpha is +0.53% — worse than random : firing on random dates would have done better. Fires are screening context inside a composite (section 4), not entries.
These verdicts are 20-day holds vs S&P 500 Equal Weight. Longer horizons can differ in either direction — check the permutation detail tables below before extrapolating.
2 · When it works — the setups that drive it
- Best bearish setup: Non-trending + Low vol — alpha -0.36% / 20d on 4,189 historical triggers.
- Least-bad era for bearish: 2015-2019 — alpha +0.03% / 20d on 17,841 triggers — still wrong-signed; no era produced negative alpha.
3 · When it fails — common false positives
- Weakest bearish cell: Trending + High vol — alpha +0.98% / 20d on 39,552 triggers.
- Worst era for bearish: 2023-2026 — alpha +1.01% / 20d on 32,817 triggers.
Signal-specific failure patterns
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. Measured pair results — same-day co-fires put through the pair backtest — follow under “Measured pairings” below.
Sequential with completed pattern
Failed-double-top and double-top-breakout fire on the same underlying pattern structure at different points: the breakout signal fires when price closes above the twin-peak resistance level by the breakout margin; failed_double_top fires when, after that breakout, price falls back below the level by a volatility-scaled threshold within the failure window (Edwards & Magee, Technical Analysis of Stock Trends, 11th ed. 2018; Bulkowski, Encyclopedia of Chart Patterns, 3rd ed. 2021). They are sequential rather than concurrent — one signal replacing the other as the setup evolves, not two independent pieces of evidence.
Measured pairings — Bonferroni survivors
Beyond the literature pairings above, these are the same-day co-fire combinations involving Failed Double Top Breakout that cleared the pair backtest's Bonferroni cut on the full 2016–2026 sample (549 pairs × 5 horizons = 2,745 hypotheses), on universes filtered to ADV ≥ $5M, price ≥ $5 and market cap ≥ $100M. That cut is two-sided: it asks only whether the co-fire's α is reliably different from zero, in either direction, so a pair can survive by reliably underperforming — every row below happens to be positive on the full sample, but that is not what the test asked. The same run holds out 2023+: the Test columns are that held-out window, printed for every row with enough held-out co-fires to measure, so a survivor that did not repeat out of sample is visible rather than hidden. All α figures here are for holding the stock long after the co-fire — no shorting assumed, and no sign flip for bearish legs. So positive α means the co-fire was followed by outperformance and negative α by underperformance, whichever way either leg points — a bearish leg does not flip the reading. Survivors are rare by design — absence of a pair here means it did not clear the cut, not that it was untested. Ranked by held-out (2023+) α. Historical tendencies, not recommendations.
China A-shares
| Pair (same-day co-fire, long) | Full α | Test α (2023+) | Test N | p_perm test |
|---|---|---|---|---|
| failed_double_top bearish + weekly_change bearish | +3.04% | +4.34% | 1,298 | 0.002 |
| failed_double_top bearish + rsi bearish | +2.12% | +2.86% | 749 | 0.002 |
| failed_double_top bearish + new_20d_low bearish | +2.36% | +2.20% | 1,860 | 0.002 |
| failed_double_top bearish + hh_hl_streak bearish | +1.65% | +2.18% | 939 | 0.002 |
| failed_double_top bearish + macd bearish | +1.57% | +1.38% | 887 | 0.002 |
| failed_double_top bearish + vwap_cross bearish | +1.05% | +1.20% | 1,850 | 0.002 |
6 of this universe's 138 surviving pairs involve this signal · α vs 83188.HK.
China A-share survivor α runs large but skews toward small-caps, where trading costs and thin liquidity claim a large share of any measured edge — screening context, not a capturable spread.
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.
- Volume-gate the failure day — A failure day on heavy volume reads as distribution — size exiting through a level the crowd is watching; on light volume it reads as noise that often reverts. The filter is derivable from OHLC plus volume and testable within the platform's own screens.
- Demand a decisive failure — The default failure threshold is one z-score of daily volatility below the resistance (clamped to 1-8%). Requiring a deeper close below the level, or several consecutive closes below it, filters one-day shakeouts at the cost of an even later entry — a trade-off worth testing rather than assuming.
- Condition on trend context — A failed breakout inside a deteriorating trend structure (for example, a bearish HH/HL reading) is plausibly a different population from a failed breakout inside an intact long-term uptrend, where dips below old resistance tend to get bought. Pairing the fire with a trend-structure filter is testable within the platform.
See also Why technical-only signals don't survive on their own for the broader argument.
5 · Before acting — a 5-point checklist
- 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.
- Where is price vs its own 50 / 200 DMA? Pattern signals carry their own structural context; check that the implied support/resistance levels have historical relevance, not just the most-recent 3-month range.
- 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.
- Is ADV20 enough for the intended size? A $1M notional order in a $500M name moves the tape by itself. A useful floor is adv20d ≥ 5% of the intended position.
- What invalidates the trade? Define a price level in advance (for longs: a close below the trigger-day low; for shorts: a close above the trigger-day high) and honour it. The backtest alpha is an average; any single trade can land at either tail.
Execution notes
Direction is definitionally bearish, but treat tradability as an empirical question: if the bearish side beats the random-date null in the current tables at your horizon, the signal can serve as a short-side screen tile; if it does not, treat fires as context — a note that a widely watched breakout just failed — rather than a short entry. If traded, entry at the open T+1; the natural invalidation is a close back above the resistance level, which would mean the failure itself failed and the breakout re-asserted. Everything here describes historical tendency, not a recommendation.