Memory chain — every decision, 2006–2026

Micron (MU), log scale. ▲ entries · ▼ exits — placed by the locked rules using ONLY data available at each moment (Taiwan monthly prints, published ~10th). Each card shows the exact math at decision time. No hindsight, no exceptions.
$3$10$30$100$300$1,000 1234567
closed fires: 6 · mean +88% · win rate 100% · every ▲▼ pair reproduced by pit_replay.py from the raw archive

The compounding wedge — combined book vs SPY, same graph, $100k start (log scale)

1x2x5x10x20x Jul 2013 — DRAM consolidates to 3 players Micron absorbs Elpida; oligopoly cycle begins — the wedge opens here combined 20.4x SPY 7.5x
hover for date & values · combined book = idle capital sits in SPY, system fires overlay on top (the investable product) · SPY = the benchmark. In idle years the book IS the market — the widening gap is fire alpha, compounding. Series stored at data/combined_vs_spy_monthly.csv · per-fire figures reproduced by pit_replay.py.

Annual returns — combined book vs SPY, year by year

-25%+0%+25%+50%0708091011121314151617181920212223242526
■ combined book · ■ SPY · matched or beat the market in 14 of 20 years · one down year the market didn't share (2007, first fire mistimed the cycle bottom) · idle years ≈ market years by construction — the system only diverges from SPY when it has a signal

Drawdowns — how deep, and who fell further

0%-20%-40%
combined book worst -55% · SPY worst -51% · the book holds SPY when idle, so it inherits every market drawdown (plus a little overlay in '08) — the edge is return-side, not downside protection. Risk-capped mandates run the fire sleeve alone: max drawdown −15% (pit_replay).

Risk & return — the numbers an allocator asks for

COMBINED BOOKSPY
CAGR (2007–2026)+16.8%+11.0%
Growth of $100k$2,035,406$753,976
Annualised volatility19.2%15.5%
Sharpe ratio (0% rf)0.900.75
Max drawdown-55.4%-50.8%
Best year+56.7%+32.3%
Worst year-38.6%-36.8%
beta to SPY 1.09 · annualised alpha +4.7pp · positive years 16/20 · monthly replay from the local archive; per-engagement record (81% win, +42% mean, 13.4-mo hold) in QUANT_CRITIQUE.md

Live forward record — published on a 21-day delay

The live record began July 6, 2026. First delayed disclosure: July 27, 2026.
2 decisions logged in the last 21 days — not yet public. Positions and decisions appear here 21 days after they are logged.
Every decision is logged with its rule and prediction BEFORE the outcome is known, then scored. The delayed log below is the unedited discipline record — including mistakes.

Disclosures

This page documents a PAPER (simulated) portfolio run under pre-registered mechanical rules. No real capital is represented unless explicitly labeled REAL. All pre-2026 figures are BACKTESTED/HYPOTHETICAL results computed point-in-time from archived statutory filings and adjusted prices; backtested performance has inherent limitations, including hindsight and survivorship effects, and is NOT indicative of future results. Nothing on this page is investment advice, an offer, or a solicitation of any investment. Positions shown are 21 days stale by design. © 2026.

The decision cards — what the system knew, when

Fire 1 · 2009-09+7%
signal print: Nanya +39% · ADATA +79% · Phison +87%
rule: ≥2 names >+20% after 6-month contraction ✓
tier at that moment: T2 · 22.5% — clean (-35% off high, no recent spike)
entry 2009-10 @ $6.62 → exit 2010-11 @ $7.09 — cycle cooled — median <+20%
Fire 2 · 2013-03+204%
signal print: Nanya +22% · ADATA +24% · Phison -8%
rule: ≥2 names >+20% after 6-month contraction ✓
tier at that moment: T2 · 22.5% — clean (-17% off high, no recent spike)
entry 2013-04 @ $9.19 → exit 2014-05 @ $27.88 — cycle cooled — median <+20%
Fire 3 · 2016-06+199%
signal print: Nanya -11% · ADATA +26% · Phison +25%
rule: ≥2 names >+20% after 6-month contraction ✓
tier at that moment: T3 · 30% — deep glut: -62% below 24-month high
entry 2016-07 @ $13.40 → exit 2017-12 @ $40.10 — cycle cooled — median <+20%
Fire 4 · 2018-08+5%
signal print: Nanya +89% · ADATA +30% · Phison +8%
rule: ≥2 names >+20% after 6-month contraction ✓
tier at that moment: T2 · 22.5% — clean (-21% off high, no recent spike)
entry 2018-09 @ $44.11 → exit 2019-10 @ $46.37 — cycle cooled — median <+20%
Fire 5 · 2020-02+105%
signal print: Nanya +34% · ADATA +22% · Phison +48%
rule: ≥2 names >+20% after 6-month contraction ✓
tier at that moment: T2 · 22.5% — clean (-27% off high, no recent spike)
entry 2020-03 @ $41.02 → exit 2021-04 @ $83.94 — cycle cooled — median <+20%
Fire 6 · 2023-10+11%
signal print: Nanya -4% · ADATA +40% · Phison +23%
rule: ≥2 names >+20% after 6-month contraction ✓
tier at that moment: T2 · 22.5% — clean (-17% off high, no recent spike)
entry 2023-11 @ $75.42 → exit 2024-12 @ $83.85 — cycle cooled — median <+20%
Fire 7 · 2025-06+804%
signal print: Nanya +21% · ADATA +59% · Phison +16%
rule: ≥2 names >+20% after 6-month contraction ✓
tier at that moment: T2 · 22.5% — clean (-17% off high, no recent spike)
entry 2025-07 @ $108.98 → still held @ $984.75 — OPEN — exit-watch live
Rule set: STRATEGY_LOCKED.md v1.5/1.6 · validation battery: QUANT_CRITIQUE.md · data: statutory Taiwan monthly revenue (MOPS/FinMind archive) + adjusted prices. Backtest caveats (instigation bias, N=7) documented and priced into planning numbers.