Rule v4.4, 3 Oct 2026 night (Bangkok). 70 iterations, and every rule now also run on Oct 2019 to Sep 2022 point in time
baskets (DeFi summer 2020, the 2021 bull, the 2022 bear), data that was never used to design the rules.
Out of sample check: v4.1 +1188%, v4.2 +1245%, v4.3 +1398%, v4.4 +2168% vs BTC +125% (worst drop BTC -72%, rules -24 to -27%).
Run length (breakout to top, 60 day window): 2025-26 median 15 days (+17%), 2023-24 27 days (+37%), 2020-21 bull 50+ days (+76%);
 after the top the next 30 days fell a median -23% to -30% in every period.
Breakout day: 21 definitions; the close above the 30 day high (20 day when half the baskets are risk on) stayed best overall.
Extended: 22 definitions incl. Fear and Greed; best = first HOT close (20 day z score 2+, top of the 90 day range) but sell only 25%.
 Fear and Greed as a trigger flipped between thresholds (65/70/75): not robust.
Turn down: 21 definitions (5/10/20 day lows, EMA10/20, 50 day avg, swing low structure breaks, failed breakout, trailing 15 to 30%,
 wait for a bounce, chandelier, two closes, vs BTC). Best = 15 day low close, with the 20% trailing stop armed only after +30%.
 Waiting for a bounce helped in 2020-22 but made worse drops in 2023-26 (-35% to -42%).
Age rotation: new coin volume share (listed under 30 days, top 1000): in the top fifth of weeks the next 4 weeks were worse in all 4
 periods (new coins -7 to -20%, alt market -2 to -12%). Solana launchpad share of DEX volume high: pump/meme baskets -6 to -14%
 next 14 days in 2024 and 2025, no effect in 2026. Rule: no new meme buys when the share is in the top 20% of its last year.
Breadth thrust 1.5x sizing: +100 CAGR points in 2023-24, nothing in 2020-22 or 2025-26: not used.
Files: bull/ (fetch_old.py, build_old.py, bk_old.json, px_old.json, age.py, age_weekly.json, age_out.txt, trench_share.py, trench_out.txt),
 it100/ (x50.py, it20*.py, it20_*.txt/json, runlen.py, runlen_*.txt, rule4_v44.py).