docs: ten years on the index as it was, with the ledger and the open questions updated

[yanivil]

Description

Documentation only: the multi-year evidence from the eleven point-in-time replays dispatched after #107 merged.

• Tuning page, new subsection "Ten years on the index as it was". One run per calendar year 2016 to 2026 (runs 34246882614 to 34247722648; tuned, asof=true, bars=500, horizon=60, grid=false), the per-year table with coverage, hit rate, mean R, month-block intervals and drawdowns, and the pooled numbers: +0.23 R on 1629 traded signals, positive in 10 of 11 years; inverse H&S +0.23 R on 1096, Wolfe +0.45 R on 228 and the best detector in 2018, 2019 and 2022, cups +0.03 R on 305; bear regime +0.59 R against bull +0.16; VIX above 25 +0.50 in eight of eight years; a stock's SMA50 more than 5 % under its SMA200 +0.35; score buckets flat in mean R; drawdowns of 36 to 41 R inside 2018, 2019, 2022 and 2023.
• What the ten years overturned from the two-year window, stated as such: breakout age, reward:risk above 4, low VIX, the SMA band U-shape, and the Wolfe verdict.
• Ledger: the tuned profile held over ten years; the Wolfe rhythm value held after failing its 12-row unseen window, with the note that the 0.30 alternative has not been replayed over the same years.
• Open questions replace the hypothesis list: Stocks with SMA50 more than 5 % below SMA200 ran +0.35 R over ten years against +0.15 to +0.20 elsewhere #99 and Market context at the scan day: bear regime +0.59 R and VIX above 25 +0.50 R over ten years, against +0.16 in bull regimes #101 (rewritten), Yearly drawdowns of 36 to 41 R in four of eleven years against an expected yearly gain near 35 R #108 drawdowns and Cup & Handle: +0.03 R on 305 signals over ten years with an 18 % hit rate; decide whether it stays a confirmed pattern #109 cups (opened), Per-signal probability and expected value: premature at about 120 resolved trades a year #97.
• Market context section, the profiles paragraph, the testing page and the README limitations point at the ten-year numbers instead of the two-year ones; changelog.

Issue updates made alongside: #98 and #100 closed as not supported, #102 closed as held, #99 and #101 rewritten, #82 and #97 commented, all with the numbers and run IDs.

Related Issues

Related: #95, #96, #82, #97, #98, #99, #100, #101, #102, #108, #109

Checklist

• Code follows existing project style and type annotations (no code changed)
• Threshold changes include a one-line "why" comment in scan.py (no thresholds changed)
• Offline unit tests pass (python -m pytest -q)
• Random-walk false-positive rate stays under 5% (detection logic untouched)
• Documentation updated in docs/wiki/
• CHANGELOG.md updated

🤖 Generated with Claude Code