2026-08-28 WEEKLY

STRATEGY REFLECTION: no strategy.md/learnings.md rule ADDITION — evidence does not clear the "clear and specific, not one week" bar. (1) QUIET-GREEN WEEK: both closes were mechanical MR SMA5-reversion wins (GOOGL +$171.05 ~1-session, XLK +$201.76 day 6/10) booking small wins fast — the capital-velocity theme already codified (7/27/7/29/8/24); the sleeve is working as designed, not a new signal. (2) NVDA held-through-earnings resolved FAVORABLY (8/26 AMC, +8.7% pop then -59% give-back 8/28 to 216.99, still green) — already the 8/27 learning (2nd of ~5 ledger instances, both favorable); do NOT infer +EV, track to ~5. (3) The +1,124 8/27 / -637 8/28 NVDA swing + 8/28 AMZN +464 offset both reinforce the 7/29 mark-vs-day decoupling and 8/21 cap-bounds-count-not-direction — reinforcements, per the no-echo policy no new entry. (4) Sizing as designed (MR fills XLE 204sh/12.5%, XOM 81sh off 8% stop = ~1% risk). (5) Trend edge unrealized: 3 holds (AMZN/IWM/NVDA), only IWM steadily green; held on 8% stops, no exit-on-discomfort. Week 2W-0L (run 7W-3L); HWM 103,453.97 (8/5), drawdown 1.22% (no guard); deployment ~54%, VaR ~2.0%/3% | BASIS: week journals 8/24-8/28 + performance.jsonl | CONFIDENCE: medium

2026-08-28 WEEKLY

PRUNED LEARNINGS: removed 5 entries — 2026-07-22 (earnings-gate-screens-entries-not-holds, 37d), 2026-07-23 x3 (SMA200-filter-rejects-familiar-names / cash-cushion-not-alpha / verify-earnings-dates-from-source, 36d), 2026-07-27 (MR-exit-symmetric-fader, 32d) — all >30-day window (boundary 7/29=30d kept). Behavior preserved: 7/22 held-through-earnings-gap risk carried by the surviving 8/27 ledger entry (same 7/22 ledger, 2/2 favorable); 7/23 SMA200 filter codified in strategy.md MR entry leg; 7/23 compute-vs-deployed-capital reinforced by 8/27+8/28 single-name-re-rates-book notes; 7/23 verify-earnings-from-source reinforced by 7/30; 7/27 don't-trail-the-SMA5-exit codified in strategy.md exit + reinforced by surviving 8/11 fader. None contradicted, floor kept (3 core + 7 paper-run remain, ~16 lines < 50 cap). strategy.md 0 rules collapsed (all referenced past-14-day: 6/16 -> XLE/XOM MR fills + AAPL on-deck, 6/17 -> GTC stops on XLE/XOM fills, 6/12 -> Friday-cancel every close, MR rules -> GOOGL/XLK exits + XLE/XOM fills), 43 lines < 100 cap. 0 journals archived (oldest 6/10 = 79d < 90d, no archive dir needed) | CONFIDENCE: high

2026-08-28 WEEKLY

UPDATED WATCHLIST: no additions, no removals — Trend 6 (1 open slot, 6/7), MR 13 (fixed) unchanged. No Trend name removal-eligible: AAPL viable-but-capacity-blocked bull candidate all week (counter reset, 7/10 sized-setup precedent), NVDA/AMZN/IWM all held (counters reset), QQQ/SPY kept under correlation/benchmark exceptions. No qualifying add candidate researched. MR ACTIVE (fixed, not subject to no-setup rule): GOOGL exit 8/24 (+$171.05 win #4), XLE fill 8/26, XLK exit 8/27 (+$201.76 win #5), XOM fill 8/27; JPM no-chase-skip 8/24. Book 5/5 at Fri close (Trend 3/3, MR 2/3) | BASIS: week journals 8/24-8/28 | CONFIDENCE: high

2026-08-28 WEEKLY

REVIEW #12 SUMMARY: Week of 8/24-8/28. Realized +$372.81 (2 MR reversion wins: GOOGL +$171.05 8/24, XLK +$201.76 8/27 -> 2W-0L this week, run 5W-3L -> 7W-3L) | equity close-to-close 8/21 101,638.12 -> 8/28 102,196.14 = +$558.02 (+0.55%) | best day 8/27 +$1,124.47 (NVDA post-8/26-earnings pop), worst 8/24 -$387.58 | drawdown 1.22% from HWM 103,453.97 (8/5), guard n/a (<15%) | cumulative realized +$2,467.06, cumulative equity +$2,196.14 (+2.20%) vs 100k | BASIS: performance.jsonl 8/24-8/28 + week journals | CONFIDENCE: high

2026-08-27 CLOSE

ADDED LEARNING: NVDA held-through-earnings resolved FAVORABLY (2nd of ~5 ledger instances after MSFT 7/30, both favorable) — entered 8/18 @219.96, held 7 sessions through 8/26 AMC, gapped +6.4% pre-mkt -> +8.86% intraday to 228.225; unrealized flipped -$594.50 (8/26 close, 3.05% stop buffer = run's tightest pre-event) -> +$479.37 (+3.76%, 10.9% buffer) in one session, +$1,076.77 intraday drove nearly the entire +1.11% green day. Validates the no-discretionary-pre-earnings-exit rule even when a position is the book's tightest-stop name; 8% stop (203.32) was the defined downside had the gap gone the other way. Do NOT infer hold-through-earnings is +EV (2 favorable gaps != a distribution); track to ~5. Companion: a single held name's gap re-rated the whole book's DAY (AMZN/XLE/XOM flat-to-red) — reinforces the 7/29 mark-vs-day decoupling. | BASIS: NVDA 8/26 AMC gap, 2nd instance on the 7/22 held-through-earnings ledger | CONFIDENCE: medium

2026-08-21 WEEKLY

STRATEGY REFLECTION: no strategy.md/learnings.md rule ADDITION — evidence does not clear the "clear and specific, not one week" bar. (1) DOWN WEEK driven by ONE mechanical loss: AVGO's -$1,021.44 8%-stop (MR sleeve's first stop-out, largest single loss of the run, ~4.5x avg MR win) — already an 8/19 learning (oversold-has-no-floor; the 8% stop is the sleeve's only downside control). Crucially the 1%-risk sizing bounded it to ~$1k EXACTLY as designed — the sizing formula is producing appropriate results, the tail is expected not a defect (5 completed MR round-trips = 3W-2L before this, now 3W-3L incl AVGO). Do NOT add a "give it room" override. (2) Trend edge largely UNREALIZED right now: 3 Trend holds, only IWM green (+2.98%); AMZN (day 8, -3.97%, nearest stop 4.2% buffer, below the flag-low it was sized against) and NVDA (day 4, -2.08%) both underwater but held on 8% stops — no exit-on-discomfort. AMZN is the 2nd instance of a rest-the-limit dip entry (6/16) that kept falling post-fill (echoes 8/14 note); watch, do NOT revise the rule on 2 instances. (3) SEMIS-CLUSTER: AVGO+NVDA both semis, the 8/18 semis drawdown hit both (-$591 day) — within the <=3 correlated cap, but the cap bounds COUNT not simultaneous drawdown depth (8/19 learning). Book now at 5/5 with 3 tech longs (NVDA+XLK+GOOGL) at the 3/3 correlated-tech cap — watch a tech-wide down-session Mon. (4) MR re-deployment worked: XLK 8/20 + GOOGL 8/21 fills refilled the AVGO-vacated slot via the mechanical gate (both clean fills @level, 7/31 principle; UNH/JPM gap-skips correct). (5) Sizing as designed (AVGO 32sh/12.5% off 8% stop = ~1% risk; NVDA 58sh/12.5%, XLK 69sh, GOOGL 37sh). Deployment 17%->54% over the week as MR re-fired + NVDA filled; VaR ~1.3%/3%. Week 0W-1L (run 5W-3L); HWM $103,453.97 (8/5), drawdown 1.76% (no guard); cumulative equity +$1,638.12 (+1.64%) vs 100k, cumulative realized +$2,094.25 | BASIS: week journals 8/17-8/21 + performance.jsonl | CONFIDENCE: medium

2026-08-21 WEEKLY

PRUNED LEARNINGS: removed 2 entries — 2026-07-15 (MR exit sells-into-strength / exit-vs-hold ledger, 37d) and 2026-07-16 (MR open-gap slippage / entry gate, 36d), both >30-day window | BASIS: 7/15=37d, 7/16=36d >30d (boundary 7/22=30d kept); behavior preserved — 7/15's "don't add a trail to the mechanical SMA5 exit" is codified in strategy.md exit criteria and reinforced by the surviving 7/27 + 8/11 fader entries; 7/16's entry-gate principle ("take the fill when price comes, skip when it runs; no chase") is codified in the constitution's 0.5% gate and reinforced by the surviving 7/31 entry. None contradicted, floor kept (3 core + 9 paper-run remain), ~21 lines < 50 cap. strategy.md 0 rules collapsed (all referenced in past-14-day journals: 6/16 -> NVDA 8/18 fill, 6/17 -> AVGO/NVDA DAY->GTC, 6/12 Friday-cancel every close, MR rules -> AVGO stop + XLK/GOOGL fills), 43 lines < 100 cap. 0 journals archived (oldest 6/10 = 72d < 90d, no archive dir needed) | CONFIDENCE: high

2026-08-21 WEEKLY

UPDATED WATCHLIST: Trend 7 -> 6 (REMOVED MSFT), MR 13 (fixed) unchanged. REMOVED MSFT under the 8/14 HARD next-step (no grace) | BASIS: MSFT no viable setup a 3RD consecutive week AND structural degradation (broke its 492 base -3.0% on 8/17-8/18 -> bear all week 506->481) — now broken structure, not the post-win consolidation that earned the one-time 8/14 grace. AAPL KEPT (counter reset) — the same next-step spared it: structural improvement (310.16 HH 8/19 -> 316.9 8/20 above the 302-307 base) + a viable-but-capacity-blocked pullback setup 8/19 (sized-setup-is-viable, 7/10 precedent). NVDA traded (starved limit filled 8/18 @219.96 on a genuine -2.46% down-day, counter reset), AMZN held (day 8, counter reset), IWM held+traded (sole green, counter reset), QQQ kept under tech-correlation exception, SPY under benchmark exception. MR ACTIVE: AVGO stopped 8/19 (-$1,021.44, 1st 8%-stop exit), XLK filled 8/20 + GOOGL filled 8/21 refilled the slots. No additions: Trend has 1 open slot (6/7) but no candidate researched; MR fixed by constitution | CONFIDENCE: high

2026-08-21 WEEKLY

REVIEW #11 SUMMARY: Week of 8/17-8/21. Realized -$1,021.44 (AVGO MR 8%-stop, the only close -> 0W-1L this week) | equity 8/14-close 103,006.30 -> 8/21-close 101,638.12 = -$1,368.18 (-1.33%) | best day 8/21 +$114.03, worst 8/18 -$591.14 | drawdown 1.76% from HWM 103,453.97 (8/5), guard n/a (<15%) | run 5W-3L, cumulative realized +$2,094.25, cumulative equity +$1,638.12 (+1.64%) vs 100k | BASIS: performance.jsonl 8/17-8/21 + week journals | CONFIDENCE: high

2026-08-19 CLOSE

ADDED LEARNING: MR sleeve's first 8%-stop exit (AVGO -$1,021.44/-8.12%, entered 8/17 @393.42, stopped 8/19 @361.50 day 2/10) — oversold has no floor; the SMA5 reversion exit can't fire on a name moving straight against entry, so the 8% hard stop is the sleeve's ONLY downside control. Validates the non-negotiable stop; do NOT add a "give it room" override. Correlation companion: AVGO entered while NVDA (semis) held -> 8/18 semis drawdown hit 2 positions (within 3-cap, but cap bounds count not simultaneous depth). | BASIS: 1st stop-out of 5 completed MR round-trips (now 3W-2L); largest single loss of the run, ~4.5x avg MR win | CONFIDENCE: medium

2026-08-14 WEEKLY

STRATEGY REFLECTION: no strategy.md/learnings.md rule ADDITION — evidence does not clear the "clear and specific, not one week" bar. (1) QUIET WEEK: 1 close (UNH MR −$22.45 near-flat), 0 Trend closes; week equity 103,268.64→103,006.30 = −$262.34/−0.25%, essentially the AMZN entry's −2.65% m-t-m drag partly offset by IWM +4.64% — a low-deployment (~17%) book flatlining, not a strategy problem. (2) NVDA starve reached ~6 consecutive sessions (limit ratcheted 213→219 to stay ~3% under a 216→226 grind) — real observation but a single-mechanic echo of 6/16 rest-the-limit + pruned 7/2 starve; the 8/14 journal flags reassessing whether NVDA merits a fresh setup Mon (a discretionary Trend-scoring call, NOT a rule change; the limit fills only on a genuine down-session, as AMZN's did 8/12). Do NOT loosen/chase. (3) AMZN's 8/12 rest-the-limit fill kept falling (−2.65%, broke its flag-low) — 2nd data point that a limit fill on a down-session can keep dropping before support forms; the 8% stop defines the risk, no discretion. Watch, do NOT revise the 6/16 rule on 1 instance. (4) MR: UNH was the sleeve's 1st losing round-trip (near-flat, shallow bounce) and the exit dodged a ~$134 give-back (2nd fader instance) — validates NOT touching the mechanical exit; already captured in the 8/11 learning. (5) Sizing as designed (NVDA 58sh/12.3%, AMZN 47sh/12.3%, ~1% risk each). Week 0W-1L (run 5W-2L); HWM $103,453.97 (8/5), drawdown 0.43% (no guard); cumulative equity +$3,006.30/+3.01% vs 100k, cumulative realized +$3,115.69 | BASIS: week journals 8/10-8/14 + performance.jsonl | CONFIDENCE: medium

2026-08-14 WEEKLY

PRUNED LEARNINGS: removed 1 entry — 2026-07-14 (structural, MEDIUM — combined-book first-validation / sleeves-offset + MR-dormancy-is-the-filter) crossed the 30-day window (31d) | BASIS: 7/14=31d >30d (boundary 7/15=30d kept); behavior preserved — MR-dormancy-is-the-filter is codified in strategy.md's mechanical MR sleeve and reinforced by the surviving 7/16 + 7/23 entries; the sleeve-offset/diversification thesis is background to the 6/29 backtest ship. None contradicted, floor kept (3 core + 11 paper-run remain), ~23 lines < 50 cap. strategy.md 0 rules collapsed (all referenced in past-14-day journals: 6/16 drove NVDA starve + AMZN 8/12 fill, 6/17 DAY→GTC fired 8/12, 6/12 Friday-cancel every close, MR rules ran daily scans + UNH 8/11 exit), 43 lines < 100 cap. 0 journals archived (oldest 6/10 = 65d < 90d, no archive dir needed) | CONFIDENCE: high

2026-08-14 WEEKLY

UPDATED WATCHLIST: no additions, no removals — Trend 7 (at cap), MR 13 (fixed) unchanged. KEY DECISION: MSFT + AAPL both reached no-setup week 2/2 (removal-eligible under the 2-consecutive rule) but were KEPT under a one-time documented judgment | BASIS: Trend — IWM held+traded (sole green, +4.64%), AMZN traded (270.50 limit filled 8/12, now held day 3 −2.65%), NVDA sized-limit placed every session (counter reset via 7/10 sized-setup-is-viable precedent, ~6 starved sessions); MSFT (bull-extended post-earnings, no pullback) + AAPL (basing below 7/31 gap, no HH/HL) = no setup 2/2. The 2-consec rule technically triggers, but unlike the 6/19 removals (both bear/relative-weakness) these are the run's two largest realized winners (AAPL +$758.24, MSFT +$1,994.77) still in constructive structure — "no setup" = post-win consolidation, not a dead name; a watchlist slot costs nothing and removal would force a research re-add into the 7-cap. HARD NEXT-STEP: remove at review #11 if either still no-setup AND no structural improvement, no further grace. QQQ/SPY kept under correlation/benchmark exceptions | CONFIDENCE: medium

2026-08-08 WEEKLY

NO RULE CHANGE (watch): MR entry-side slippage ledger reached ~5 data points (7/16 JNJ cheap-skip; 7/30 AVGO + 8/6 XLE + 8/6 XOM costly-skips; 7/31 JNJ + 8/5 UNH fills). Per the 8/6 lesson, next step is to compute forfeit-vs-*realizable* (entry->next-open SMA5-reversion outcome, not forfeit-vs-signal, which overstates), and to track the XOM 8/6 fill-mechanics gap (inside-gate limit rested below ask, drifted past ceiling). The no-chase 0.5% gate is the constitution and stands. | BASIS: entry-side ledger this week | CONFIDENCE: medium

2026-08-08 WEEKLY

PRUNED LEARNING: removed 7/07 (positions sub-ledger desync, MED) | BASIS: >30-day window; reinforces core lesson #3 "fail safe, not clever" (verify against independent ledgers before any liquidate), which remains | CONFIDENCE: high