Weekly Results — Sells PAPER TRADING

ALL was a stop-loss trade, plain and simple — two trading days, -5.9%, kicked out by the fixed -5% rule before I could even see what the thesis would have done given time. The model's confidence here was never high to begin with: 0.6058 composite, rank 10, and the reasoning leaned almost entirely on "upcoming catalysts" and a Catalyst pillar score of 80 against a Momentum pillar of only 48.18. That's a lopsided setup — the model was betting on an event, not on price behavior already confirming the move. When Momentum is the weakest pillar and Catalyst is doing all the lifting, that's a trade that needs the catalyst to actually land favorably, and I don't have anything in the dossier showing it did.

What I can see: the regime flipped from bull (0.84 confidence) to neutral (0.50) over the two-day hold — flagged explicitly in deep_stats as regime_flipped_during_hold. VIX rose from 15.03 to 16.5, HY OAS ticked up, and SPY's 20-day return decayed sharply from 4.2% to 1.5%. Market breadth also weakened — advancers/decliners flipped negative on 7/14. That's a broad de-risking move, not something specific to Allstate. The CPI print landed on 7/14, right in the middle of the hold, and macro conditions worsened right after — that's circumstantial but plausible as a trigger.

Sector context makes this look less idiosyncratic: peer insurance names averaged -1.867% over the same window, so the group was soft. ALL's -5.9% is still much worse than peer average, meaning stock-specific weakness compounded a sector-wide wobble — I don't have technicals or sentiment detail at entry/exit to pin down what specifically broke down inside the name.

Headlines don't give a clean smoking gun: the CFO announcement was rated positive, and the only negative-scored headline was a generic "hidden insurance stock" piece from the day before entry — not company-specific bad news. No earnings surprise, no news_events entries. I don't have a single catalyst that explains the drop; I have a macro regime flip plus sector softness plus a stock that fell further than its peers, and no data to explain the incremental idiosyncratic loss.

Lesson for the future: when Catalyst pillar score is materially higher than Momentum (here, 80 vs 48), that spread should itself be treated as a risk flag — the model is pricing in an event that hasn't been confirmed by price action. Going forward I want to test whether trades with Catalyst-Momentum spread above some threshold (e.g., >25 points) show higher stop-out rates in the fixed -5% regime, and if so, either widen the stop or require Momentum >55 before entry regardless of composite score.

# OII Post-Mortem: -0.69% over 4 days

The model ranked OII #2 with a composite of 65.6, confidence 0.6563. Looking at the pillar breakdown, this was a sentiment-and-technical story more than anything else — Sentiment scored 77.92, Technical 60.48, while Momentum and Catalyst sat right at the middle (50.67, 50.0). Smart Money and ML pillars were null, so the score was built on a narrower base than the framework's average pick, and the reasoning field explicitly leaned on "upcoming catalysts" without naming one — a soft justification I should weight less going forward.

The entry technicals actually look stretched, not primed. RSI at 67, CCI at 169.7, MFI at 78.9, Williams %R at -7.6 — every momentum oscillator was near overbought at the moment of purchase. Price was already sitting right at the upper Bollinger band (43.09 vs entry 43.02). That's a profile of a stock that had already run, not one about to break out. By exit, RSI had cooled to 61 and Williams %R dropped to -27.8 — consistent with the position simply mean-reverting off an overbought entry rather than continuing higher.

The bigger structural problem: the regime flipped from bull (0.84 confidence) to neutral (0.50 confidence) during the four-day hold. Market breadth confirms the wobble — advancers/decliners flipped negative on the CPI print day (7/14: 70 advancers vs 93 decliners) and again on exit day (55 vs 108). VIX rose from 15.03 to 16.73, SPY's 20-day return decayed from 4.2% to 0.3%. This wasn't an OII-specific breakdown; the tape underneath it went from tailwind to headwind mid-trade.

Sector context makes the loss look worse in relative terms. Energy peers averaged +1.7% over the same window (16 names) while OII lost 0.69% — so this was idiosyncratic underperformance layered on top of a decent sector tape, not a sector-wide drag. The move itself was small in magnitude: only 0.2x ATR, meaning this was noise-level drift, not a real directional break.

**Lesson for the future:** when entry RSI/CCI/MFI/Williams %R are all simultaneously flashing overbought, treat that as a red flag for chasing a bull-regime confirmation trade — regardless of composite score — and specifically test whether entries with overbought technical stacks +confidence >0.60 underperform in this framework's history. If that pattern holds, add an explicit overbought veto rather than let Sentiment (77.92) outweigh four converging technical warnings.

NAVN — Post-Mortem, week of 2026-07-13

The model put NAVN at rank 7 with a composite of 60.47, driven mostly by catalyst-timing logic ("high probability events within 30 days") plus a Sentiment pillar of 62.37 and Technical of 60.94. Fundamental came in weak at 47.92, and both Smart Money and ML pillars were null — meaning two of the seven inputs contributed nothing, and the model's cross-sectional percentile (78.33) was leaning on the remaining five pillars to do all the work. That's a thinner basis for a rank-7 pick than the composite number alone suggests.

The trade lost 1.96% over four days, exiting on the scheduled Friday sell. Two things stand out. First, the regime flipped — bull with 0.8398 confidence at entry to neutral at 0.5 confidence by exit. That's a genuine change in the backdrop, not noise: SPY's 20-day return decayed from 4.24% to 0.30% over the same four days, and VIX rose from 15.03 to 16.73. The tape cooled off while I was holding. Second, this wasn't NAVN-specific weakness — the sector average return for the week was -1.252% across 9 peers, and NAVN's -1.96% is only modestly worse than that. This reads as sector/market drag more than an idiosyncratic breakdown in the name itself.

The headlines don't explain the loss — if anything they lean positive (BNP Paribas raised its price target to $30, Cisco added to its Navan stake). A CPI print landed on 7/14, inside the holding window, and macro softened afterward (yield curve widened slightly, HY OAS ticked up), consistent with a market that got more cautious mid-week. No single catalyst stands out beyond that; I don't have technicals or move-vs-ATR data stored for this trade to say how unusual the drawdown was in NAVN's own volatility terms.

Lesson: the model weighted a catalyst/sentiment setup without any market-regime confirmation gate at entry, and rank-7 picks with two null pillars (Smart Money, ML) should probably be down-weighted or flagged as lower-conviction rather than scored at face value. A falsifiable fix: track whether regime-flip-during-hold correlates with underperformance across all trades with confidence >0.80 at entry — if flips predict losses at a meaningfully higher rate, add a same-week regime-stability check before sizing into catalyst-driven picks.

# SHC Post-Mortem: -1.88% over 4 days

The model ranked SHC 9th with a composite of 60.59 and 71.7th percentile cross-sectionally — not a top-conviction pick, but a reasonable one. The reasoning cited "upcoming catalysts" with high probability within 30 days, plus a Technical pillar of 62.33 and Sentiment at 61.12. Momentum was the weak link at 48.17, and Smart Money/ML pillars were null — meaning two of the seven scoring dimensions simply weren't contributing signal here. That's worth flagging up front: this was a below-full-information decision by the model's own admission.

What actually happened: I bought into a bull regime with 83.98% confidence and sold out of a neutral regime with confidence. The regime flipped during the hold — that's logged explicitly in deep_stats. This wasn't a slow bleed; something shifted in the broader market read between July 13 and July 17. Macro context lines up: VIX rose from 15.03 to 16.73 (+11%), spy_return_20d decelerated hard from 4.24% to and a CPI print landed on July 14, right inside the hold. Market breadth also cracked mid-week — advancers/decliners flipped negative on the 14th (70/93) and again sharply on exit day the 17th (55/108), even though breadth had bounced on the 16th. That's a choppy, indecisive tape, not a clean trend.

Sector context cuts the other way: peers in Life Sciences Tools & Services averaged +20.2% over the same window. SHC lost money while its own sector ran hard. That's the most damning data point here — this wasn't sector-wide weakness dragging the name down, it was idiosyncratic underperformance against a favorable peer backdrop. I have no headlines, no news_events, no earnings surprise, and no options_flow data to explain why SHC specifically diverged from a strong sector — the dossier is empty on all of those. I can't point to a catalyst that didn't fire or a story that broke. No single driver stands out in what I have on file.

Fundamentally, SHC also carried a rich valuation (PE 67.4, PB 8.3) with eps_growth reported at 277.46% — a number extreme enough that I treat it as likely distorted by a small or negative prior-year base rather than genuine operating momentum. The Momentum pillar score (48.17) already seemed to be flagging that skepticism before the trade even started.

**Lesson for the future:** when regime confidence at entry is high (>80%) but Momentum is sub-50 and Smart Money/ML pillars are null, the composite score is leaning on stale or incomplete inputs to justify a mid-pack rank. A concrete rule to test: down-weight or exclude entries where more than two of the seven pillars are null, since this trade shows the model can still generate a "buy" signal on partial information and get blindsided by a regime flip it had no room to detect mid-week.