double_top_breakout
Double Top Breakout
Bullish: two peaks test the same resistance level, then price breaks out above. Tolerance is normalized by daily volatility (z-scores). Requires minimum 8% retracement between peaks.
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 |
| min_separation | Min days between peaks | 25 | 10–60 |
| max_separation | Max days between peaks | 252 | 60–504 |
Historical context
19,045 valid triggers on 3,171 distinct tickers between 2015-06-30 and 2026-04-21. Universe: us_only · mcap ≥ $100,000,000 · price ≥ $1 (3,175 tickers). Entry at open T+1. 1d = intraday T+1; 20d = open T+1 to close T+20.
Benchmarks: spxew (S&P 500 Equal Weight — the primary benchmark here; a median-stock view that avoids the 2020+ megacap-concentration distortion), spx (S&P 500, cap-weighted), and msci (MSCI World USD). Per-stock regime: trending = ADX(14) ≥ 25, high vol = 20d ann. vol ≥ 20%.
At a glance (20d alpha vs S&P 500 Equal Weight, US-only)
Double Top Breakout is a single-direction signal — only the bullish side is meaningful. (The trigger condition only describes one side of the move.)
Reading this: the random-date null is: for each ticker, sample N random dates and compute the same alpha — what alpha does a signal with no information produce? If the signal's observed alpha beats the null (pperm≤0.05), it's adding real information. If it's inside or worse than the null, the signal doesn't add value over random firing — any observed alpha is either noise or a universe artifact.
How often does DOUBLE_TOP_BREAKOUT fire in each regime?
The signal's bucket distribution is itself informative. If 50%+ of all DOUBLE_TOP_BREAKOUT triggers fire in the "non-trending + high vol" quadrant, the signal is structurally a chop-market event — regardless of what its textbook definition claims. Bullish and bearish are shown separately; counts are across the full US-only sample after the mcap and price floor.
Per-stock regime quadrant — 20d alpha
Each trigger is tagged with the host stock's own technical regime on the trigger date: is the stock itself in a trend (ADX(14) ≥ 25) or ranging? And is its realized 20-day volatility high (≥ 20% annualized) or low? This is the textbook conditioning variable — "does this signal work better in trending stocks?" — answered at the level of the individual stock, not the market. Positive bars are good for the signal; negative bars mean alpha vanishes into the benchmark or worse.
Sub-period check — does the signal work in every era?
A multi-year average can hide major instability. We split the sample into three non-overlapping windows: 2015–2019 (pre-COVID, normalized monetary policy), 2020–2022 (pandemic crash + recovery + rate-shock bear), and 2023+ (post-ZIRP, AI megacap rally). If a signal's alpha is positive overall but comes entirely from one era, that's a red flag — the conditions that produced it may not repeat. A robust signal shows a consistent sign across all non-empty buckets.
↑ Bullish triggers
| Bench | Metric | 1d | 5d | 20d | 60d | 252d |
|---|---|---|---|---|---|---|
| spx | Stock % | -0.10% | -0.14% | -0.09% | +1.29% | +8.83% |
| Bench % | +0.00% | +0.13% | +0.77% | +2.49% | +12.96% | |
| Alpha % | -0.10% | -0.26% | -0.80% | -1.14% | -4.15% | |
| Median alpha | -0.10% | -0.42% | -1.32% | -2.45% | -10.34% | |
| Hit rate (α>0) | 47.0% | 45.6% | 43.0% | 43.0% | 37.7% | |
| 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 | 19,043 | 18,955 | 18,826 | 18,398 | 16,270 | |
| msci | Stock % | -0.10% | -0.14% | -0.09% | +1.29% | +8.83% |
| Bench % | +0.05% | +0.14% | +0.64% | +1.99% | +10.17% | |
| Alpha % | -0.15% | -0.28% | -0.69% | -0.68% | -1.38% | |
| Median alpha | -0.16% | -0.41% | -1.19% | -1.97% | -7.40% | |
| Hit rate (α>0) | 46.0% | 45.6% | 43.6% | 44.2% | 40.8% | |
| p (naive) | <0.001 | <0.001 | <0.001 | <0.001 | 0.0002 | |
| p (HAC) | <0.001 | <0.001 | <0.001 | <0.001 | 0.0618 | |
| N | 18,937 | 18,845 | 18,697 | 18,274 | 16,159 | |
| spxew | Stock % | -0.10% | -0.14% | -0.09% | +1.29% | +8.83% |
| Bench % | +0.01% | +0.10% | +0.53% | +1.64% | +8.68% | |
| Alpha % | -0.11% | -0.23% | -0.54% | -0.26% | +0.32% | |
| Median alpha | -0.10% | -0.33% | -1.01% | -1.69% | -5.96% | |
| Hit rate (α>0) | 47.3% | 46.1% | 44.2% | 45.2% | 42.4% | |
| p (naive) | <0.001 | <0.001 | <0.001 | 0.0970 | 0.3786 | |
| p (HAC) | <0.001 | <0.001 | <0.001 | 0.1383 | 0.6617 | |
| N | 18,970 | 18,829 | 18,695 | 18,272 | 16,114 |
Permutation null detail — all horizons × both benchmarks
| Horizon | Bench | Observed α | Null mean | Null 95% CI | pperm |
|---|---|---|---|---|---|
| 1d | spx | -0.10% | +0.04% | [-0.05%, +0.04%] | 1.000 |
| 1d | msci | -0.15% | +0.01% | [-0.07%, +0.01%] | 1.000 |
| 1d | spxew | -0.11% | +0.01% | [-0.08%, +0.01%] | 1.000 |
| 5d | spx | -0.26% | +0.24% | [-0.04%, +2.58%] | 1.000 |
| 5d | msci | -0.28% | +0.25% | [-0.04%, +2.61%] | 1.000 |
| 5d | spxew | -0.23% | +0.27% | [-0.02%, +2.62%] | 1.000 |
| 20d | spx | -0.80% | +0.51% | [+0.02%, +1.82%] | 1.000 |
| 20d | msci | -0.69% | +0.64% | [+0.14%, +1.95%] | 1.000 |
| 20d | spxew | -0.54% | +0.71% | [+0.20%, +2.03%] | 1.000 |
| 60d | spx | -1.14% | +1.14% | [+0.31%, +2.76%] | 1.000 |
| 60d | msci | -0.68% | +1.59% | [+0.73%, +3.23%] | 1.000 |
| 60d | spxew | -0.26% | +1.81% | [+0.94%, +3.42%] | 1.000 |
| 252d | spx | -4.15% | +3.06% | [+0.98%, +5.16%] | 1.000 |
| 252d | msci | -1.38% | +5.40% | [+3.32%, +7.67%] | 1.000 |
| 252d | spxew | +0.32% | +6.70% | [+4.70%, +8.85%] | 1.000 |
Example triggers on US large-caps (2023+, mcap ≥ $30B)
Six recent bullish DOUBLE_TOP_BREAKOUT triggers on US mega-caps, filtered to |alpha| ≤ 25% to exclude catalyst-driven outliers (earnings surprises, M&A, binary events). The first three are the strongest outcomes — what the signal looks like when it works. The last three are the weakest — what the signal looks like when it fails. Each chart shows the stock's price with signal-appropriate technical overlays (e.g. MACD subpanel on MACD pages, Bollinger Bands on Bollinger pages, the 52-week trailing max line on 52w-high pages), a dot marking the trigger date, and the forward window shaded (green when the signal was right, red when it wasn't). Click any chart to open full-size.
Strongest outcomes (what DOUBLE_TOP_BREAKOUT looks like when it works)
Weakest outcomes (what DOUBLE_TOP_BREAKOUT looks like when it fails)
Stock-regime quadrants (2×2 per-stock, 20d alpha detail table)
| 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 | 1,706 | -0.28% | +0.47% | -0.70% | <0.001 | -0.28% | +0.25% | -0.48% | 0.0014 | -0.28% | -0.00% | -0.21% | 0.1639 |
| Trending + High vol Crisis selloff or parabolic rally | 6,698 | +0.01% | +0.93% | -0.85% | <0.001 | +0.01% | +0.83% | -0.76% | <0.001 | +0.01% | +0.66% | -0.53% | 0.0024 |
| Non-trending + Low vol Quiet chop, summer doldrums | 1,803 | -0.71% | +0.34% | -1.02% | <0.001 | -0.71% | +0.11% | -0.79% | <0.001 | -0.71% | -0.01% | -0.65% | <0.001 |
| Non-trending + High vol Classical "whipsaw zone" for momentum | 8,838 | +0.12% | +0.80% | -0.65% | <0.001 | +0.12% | +0.68% | -0.55% | <0.001 | +0.12% | +0.66% | -0.50% | <0.001 |
Sub-period breakdown table (20d alpha)
| Period | N | Alpha % (spx) | p (HAC) | Alpha % (msci) | p (HAC) | Alpha % (spxew) | p (HAC) |
|---|---|---|---|---|---|---|---|
| 2015-2019 2015-01-01 → 2020-01-01 | 5,694 | -1.15% | <0.001 | -0.96% | <0.001 | -0.75% | <0.001 |
| 2020-2022 2020-01-01 → 2023-01-01 | 5,292 | -0.44% | 0.0107 | -0.38% | 0.0275 | -0.67% | <0.001 |
| 2023-2026 2023-01-01 → 2099-01-01 | 8,059 | -0.78% | <0.001 | -0.67% | <0.001 | -0.29% | 0.0515 |
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 -0.80% and worse than random — triggering on random dates would have produced better long-side returns. Either direction fails the "beats random" test. Don't use Double Top Breakout 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: Non-trending + High vol — alpha -0.65% / 20d on 8,838 historical triggers.
- Best era for bullish: 2020-2022 — alpha -0.44% / 20d.
3 · When it fails — common false positives
- Weakest bullish cell: Non-trending + Low vol — alpha -1.02% / 20d on 1,803 triggers.
- Worst era for bullish: 2015-2019 — alpha -1.15% / 20d.
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. A multi-signal convergence backtest is planned but not yet run.
Reversal-pattern family
Double top and double bottom are canonical two-swing reversal patterns (Edwards & Magee, Technical Analysis of Stock Trends, 11th ed. 2018; Bulkowski, Encyclopedia of Chart Patterns, 3rd ed. 2021). Their statistical properties have been studied in peer-reviewed work (Lo, Mamaysky, and Wang, "Foundations of Technical Analysis", Journal of Finance 55(4), 2000). A completed double-top breakout and a failed-double-top signal on the same stock fire in sequence rather than concurrently — they represent different stages of the same pattern.
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 + sector gate — A double-top breakout on 2× volume in a strong sector is plausibly a different population than the default. Filter cost is ~60% of triggers; concentrates the remaining sample. Testable.
- Use as opportunistic short setup — If the breakout fails (closes back below the double-top level within 5-10 days), THAT is the signal — it's captured by the failed_double_top module. Skip the straight breakout; wait for the failure confirmation.
- Require structural consolidation pre-breakout — Tight-range consolidation (<5% range over 20d) before a breakout is structurally different from a straight-line rally to new highs. Filter derivable from OHLC.
See also Why technical-only signals don't survive on their own for the broader argument.
5 · Before you act — 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 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.
- 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
Pattern fires on only ~5 times per ticker per decade, but the signal's forward returns are consistently negative. Treat it as a documentation-only signal; skip as a primary trigger. If traded, entry open T+1 with tight invalidation (close back below breakout level within 5 days = exit). Raw signal as entry trigger underperforms passive SPX by 80-114 bps per month.