SignalSync

SignalSync scans every S&P 500 constituent on daily bars for four bullish chart patterns and reports only confirmed, risk-defined setups with an entry, a structural stop and a reference target.

It is a heuristic screener, not trading advice. Every hit should be checked on a chart before acting.

The page: yanivil.github.io/SignalSync, rebuilt after every scan, in four parts: the buy signals now with the buy limit, stop loss and take profit and one instruction on how to act; why each stock was triggered, with its chart; the open trades with where each stands between its stop and its target; and every closed trade with its result in price terms, scored with the backtest's accounting.

Pattern	Type	Confirmation trigger	Stop-loss	Reference target
Cup & Handle	continuation	daily close above the handle high on ≥ 1.4× average volume	handle low − 0.25 ATR	entry + (right rim − cup bottom)
Inverse Head & Shoulders	reversal	daily close above the neckline on ≥ 1.3× average volume	right-shoulder low − 0.25 ATR	entry + (neckline at the head − head)
Bullish Wolfe Wave	reversal	daily close back above the 1-3 line after point 5	point-5 low − 0.25 ATR	line 1-4 at the ETA
Double Bottom	reversal	daily close above the peak between two lows within 3 % of each other, the rally between them ≥ 10 %	second low − 0.25 ATR	entry + (peak − lower low)

Four rules shape every detector: no forced patterns (strict geometry plus a 0–100 quality score, minimum 60), respect the wider trend (SMA50/SMA200 filters), enter only after confirmation (close-based triggers with volume, CONFIRMED vs WATCHLIST, no chasing beyond 5 %), and always define risk (setups with a stop more than 12 % away for cups, 15 % for the others, are rejected). Two more rules came from the eleven-year replay: a confirmed row is retired after its sixth session on the list (a Wolfe after its fifth), because entries later than that showed no edge; and the Cup & Handle is watch-only, listed but never a buy signal, because its breakouts made +0.03 R per signal with 18 % reaching the target. The rules follow the engine specification adopted on 2026-09-05; the previous rule set is kept as the legacy profile (--profile legacy) so the two can be replayed side by side.

Pipeline

flowchart LR
    A[S&P 500 constituent CSV<br/>pinned upstream commit] --> B[Symbol normalisation<br/>BRK.B → BRK-B]
    B --> C[yfinance Ticker.history<br/>8 threads · retry 5 s / 10 s]
    C --> D[fill_missing_close<br/>last-trade quote → newest close]
    D --> E[adjust_ohlc<br/>split / dividend ratio]
    E --> F[align_last_bar<br/>majority's newest complete bar]
    F --> G{per symbol}
    G --> H[Cup & Handle]
    G --> I[Inverse H&S]
    G --> J[Bullish Wolfe Wave]
    H & I & J --> K[score ≥ 60<br/>CONFIRMED / WATCHLIST<br/>de-duplicate]
    K --> L[output/signals.json]
    K --> M[output/report.md]
    L & M --> N[GitHub Actions commit<br/>→ pages workflow<br/>yanivil.github.io/SignalSync]

Loading

The whole scanner is one module, scan.py: constants at the top, then data loading, indicators, the four detectors, and reporting. See the wiki for the full walk-through.

Quickstart

Requirements: Python 3.12, pandas, numpy, yfinance. requirements.txt is hash-pinned (compiled from requirements.in); requirements-dev.txt adds pytest and ruff. No API keys: the constituent list comes from a pinned public GitHub dataset and prices from Yahoo Finance via yfinance.

python3 -m venv .venv && . .venv/bin/activate && pip install --require-hashes -r requirements.txt -r requirements-dev.txt

python scan.py                              # full S&P 500 scan, 2 years of daily bars

python scan.py --tickers AAPL,MSFT,NVDA -v  # subset, debug logging

python scan.py --min-score 70 --max-age 2 --csv my_universe.csv --out-dir out

python -m pytest -q                         # whole test suite, offline, about two seconds

bash run_daily.sh wraps the scan in a self-healing virtualenv and keeps dated logs under logs/.

Flag	Default	Meaning
--tickers	S&P 500	comma-separated symbols instead of the index
--csv	pinned GitHub CSV	local constituents file with a Symbol column
--period	2y	yfinance history period (about 500 daily bars)
--min-score	60	minimum quality score to report
--max-age	3	max bars since the confirming close (H&S and Wolfe get +5 for pivot lag)
--profile	spec	rule profile: spec (the specification), tuned (spec with four rules relaxed and a minimum reward:risk of 1.0 plus a 60-bar watchlist limit added, all on replay evidence; what the nightly job runs) or legacy (rules until 2026-09-05). See the tuning page for the replay comparison.
--out-dir	output	where signals.json and report.md are written

Exit codes: 0 ok, 2 no price data at all (network problem).

Sample output

output/report.md from the 2026-09-04 run (abridged):

# S&P 500 pattern scan — 2026-09-04 08:25

Scanned 502 of 503 symbols (daily bars, last bar 2026-09-03). Min quality score 60.
Breakouts older than the per-pattern limit are dropped (bars: Cup & Handle 3,
Inverse Head & Shoulders 8, Bullish Wolfe Wave 8; ...).
Data errors: 1

## Confirmed breakouts (actionable): 5

| Ticker | Pattern                  | Entry  | Max buy | Stop   | Risk % | Target | R:R   | Score | Age | Vol× | Trend                                   | Details                                              |
|--------|--------------------------|--------|---------|--------|--------|--------|-------|-------|-----|------|-----------------------------------------|------------------------------------------------------|
| CL     | Bullish Wolfe Wave       | 90.09  | 90.80   | 88.67  | 1.58   | 107.98 | 12.6  | 84    | 8/8 | 0.97 | close above SMA200, SMA50 > SMA200, ... | 1 2026-07-23 @89.25, 2 2026-07-28 @95.46, ... 5 2026-08-21 @89.16; line 1-3 now 88.95; first target 95.46 (point 4) |
| HAL    | Inverse Head & Shoulders | 37.29  | 37.83   | 32.64  | 12.47  | 42.34  | 1.09  | 82    | 4/8 | 0.8  | close above SMA200, SMA50 < SMA200, ... | LS 2026-07-02 @32.44, head 2026-07-30 @30.84, RS 2026-08-26 @32.88, neckline 36.03->35.91 |
| VRTX   | Cup & Handle             | 557.96 | 572.51  | 528.86 | 5.22   | 626.63 | 2.36  | 77    | 1/3 | 1.16 | close above SMA200, SMA50 > SMA200, ... | left rim 2026-07-07 @533.67, bottom 2026-07-23 @465.00 (depth 13%), ... trigger 553.47 |

## Watchlist (pattern complete, waiting for a close above trigger): 17

| BDX    | Cup & Handle             | 191.76 | 195.69  | 183.9  | 4.1    | 236.51 | 5.69  | 84    | -   | -    | close above SMA200, ...                 | left rim 2026-02-24 @184.86, ... handle low 2026-09-03 @184.91 (depth 4.2%), trigger 191.76 |

## Closed since the last report (2026-09-03 08:25): 2

| Ticker | Pattern      | Was       | Outcome        | Entry  | Stop   | Target | Detail                                          |
|--------|--------------|-----------|----------------|--------|--------|--------|-------------------------------------------------|
| GPC    | Cup & Handle | CONFIRMED | TARGET_REACHED | 138.08 | 133.59 | 164.92 | high 165.10 on 2026-09-03 reached target 164.92 |
| DG     | Cup & Handle | WATCHLIST | FAILED         | 134.13 | 116.6  | 161.68 | close 116.20 on 2026-09-03 at or below stop 116.6 |

Max buy is the open above which the setup no longer qualifies: the trigger plus 5 %, or lower where a fill would already carry 1.5× the planned risk (CL above: a Wolfe stop sits 1.6 % under the entry, so the chase allowance is 0.8 %, not 5 %). R:R is the reward per unit of planned risk, (target − entry) / (entry − stop); it shrinks with every session the entry drifts above the trigger, which is why HAL, four sessions past its breakout, shows 1.09. F&G (added after this sample) is the stock's own fear-and-greed reading at the last close, 0 to 100, the average of RSI 14, the MACD histogram's percentile within the trailing year and Bollinger %B; above 80 the stock is stretched, and over ten years of replay such breakouts paid least. Information only. The last table explains every row of the previous report that is gone today (the two rows above are illustrative). output/signals.json carries the same rows as records plus a meta block (last_bar, per-symbol bar histogram, effective breakout-age limits, the market context: SPY regime, VIX and breadth, informational only) and the closed list. The schema is documented in the wiki.

Documentation

Page	Contents
Architecture and Data Pipeline	universe source, ingestion, throttling, the "which bar is scanned" logic, scheduling
Pattern Catalog	exact geometric criteria, formulas for entry / stop / target / score, known edge cases
Configuration and Tuning	every threshold, what loosening or tightening it does, false-positive filters
Testing and Contributing	running the suite, fixtures, adding a pattern, code style

docs/wiki/ is the source of truth for the GitHub wiki; the sync-wiki workflow mirrors it there on every change to main. Coverage is printed in the tests workflow log rather than published as a badge, which would need an external service.

Layout

scan.py                              detectors, data loading, reporting, CLI (single module)
test_scan.py                         original suite: textbook fixtures, random-walk sweep, last-bar handling
test_patterns.py                     primitive precision, formula verification, negative controls, boundaries
test_pipeline.py                     retry policy, universe loading, end-to-end mini universe
test_evaluate.py                     outcome classification and the git signal log
test_backtest.py                     walk-forward replay: no look-ahead, first-seen signals, fills, breakdowns
test_build_site.py                   the web page: grades by day on the list, levels, track record, HTML
test_site_charts.py                  the bars behind the page's charts, cut at the scan's bar
test_universe_history.py             point-in-time membership from a git history
conftest.py                          shared fixtures and the offline yfinance stand-in
tools/debug_last_bar.py              per-symbol last-bar diagnostics (also a manual GitHub workflow)
tools/evaluate_signals.py            replay past CONFIRMED signals against later prices (manual workflow)
tools/backtest.py                    walk-forward replay of the scanner over the last N sessions (manual workflow)
tools/universe_history.py            index membership as of any past date, from the constituent dataset's git history
tools/build_site.py                  builds the web page from signals.json and the live track record
tools/site_charts.py                 fetches the recent bars behind the page's charts (pages workflow)
site/                                stylesheet and script that tools/build_site.py inlines into the page
.github/workflows/daily-scan.yml     01:17 UTC daily: tests, scan, commit output/
.github/workflows/pages.yml          after each successful scan: score the signal log, fetch chart bars, build the page, deploy to GitHub Pages
.github/workflows/tests.yml          lint + tests + coverage on pull requests and pushes to main
.github/workflows/sync-wiki.yml      mirrors docs/wiki/ into the GitHub wiki
run_daily.sh                         local wrapper (venv, dependency checksum, dated logs)
output/                              latest signals.json + report.md, committed by CI
requirements.in / requirements.txt   runtime deps: lower bounds, and the hash-pinned compile used by every install
requirements-dev.in / -dev.txt       pytest, pytest-cov, ruff, likewise hash-pinned
SECURITY.md                          how to report a vulnerability
docs/wiki/                           documentation, mirrored into the GitHub wiki

Known limitations

• Pattern recognition is heuristic. Thresholds follow common practice (O'Neil, Bulkowski, Wolfe) but there is no industry standard; expect some false positives and misses.
• Swing points are only recognised 5 bars after they print, so Inverse H&S and Wolfe confirmations can be reported up to 5 bars late.
• Cup bases must be explained at least as well by a parabola as by a two-legged V; the rule is calibrated on reference shapes, not on market data (see the pattern catalog).
• The spec profile requires breakout volume (1.4× for cups, 1.3× for H&S) and caps a cup at half its preceding advance, so it confirms far fewer setups than the legacy rules; breakouts without volume appear on the watchlist marked as such.
• Signal quality is measured by replay, not proven live. tools/backtest.py (the backtest workflow) runs the scanner walk-forward over past sessions and tools/evaluate_signals.py scores the signals the nightly job actually committed. Every rule value was chosen on the year to 2026-09-04. Replayed on the index as it was in each year from 2016 to 2026 (--constituents-asof, which removes most of the survivorship bias; symbols that no longer trade still have no Yahoo history), the tuned profile made +0.23 R per trade, positive in ten of eleven years, with drawdowns of 36 to 41 R inside four of those years against an expected yearly gain near 35 R; the Cup & Handle detector was zero-expectancy over the same years. The tuning page has the tables and the check every further change must pass.
• The quality score is a gate, not a ranking: its buckets do not order outcomes in replay.
• Yahoo Finance data is unofficial. Symbols with fewer than 60 bars are skipped and counted in meta.errors.
• There is no persistent price cache: every run re-downloads two years of history for the whole universe.

Community

Have questions, ideas for new patterns, or setups to share? Join the conversation in GitHub Discussions:

• Announcements — project updates and community guidelines.
• Ideas — suggest new patterns, additional universe filters, and scoring improvements.
• Q&A — setup help, environment debugging, and algorithm questions.
• Show and tell — share live signals, backtests, and custom forks.

License

MIT. The scanner is a heuristic screener provided as is, without warranty; nothing it outputs is trading advice.