Weekly Results — Sells PAPER TRADING

ADPT ranked 4th in that week's batch with a composite of 56.0 and a confidence of 0.5604 — respectable but not a top-conviction call. The pillar breakdown tells you exactly why the model liked it: Catalyst scored 80.0, driving the "high probability events within 30 days" reasoning, while Momentum (61.31) and Technical (61.52) were moderately supportive and Fundamental was the weak link at 46.39. That fundamental softness was earned — negative operating margin (-20.62), negative ROE (-24.26) and ROA (-9.91), no PE to speak of, against a rich PS of 13.24. This was a growth/catalyst bet on a name that doesn't make money yet, and the framework leaned on the catalyst pillar to carry it.

The trade never got a chance to play out. It was flagged for a 30-day catalyst window but exited the same day it was bought, on a fixed -5% stop-loss, closing at -5.3%. Holding days is null, so I can't even confirm this was a same-day stop versus a data artifact, but the buy_date and sell_date being identical strongly suggests a fast, sharp drawdown right out of the gate.

What's frustrating is how little the dossier gives me to explain it. Headlines, news_events, econ_events, gdelt_tone, options_flow — all empty. Macro was unremarkable: VIX 18.77, yield curve positive, HY OAS tame, regime "neutral" and unflipped. Sector peers averaged return, so this wasn't a sector-wide selloff — it looks idiosyncratic to ADPT. Market breadth was mildly negative (98 decliners vs 64 advancers) but not a breadth collapse. deep_stats has move_vs_entry_atr as null, so I can't size the move against ADPT's own volatility to judge if this was a normal noise-band excursion or a genuine outlier. No single catalyst stands out in what I have on file — the move happened, but the record doesn't tell me why.

Lesson: when the Catalyst pillar is doing most of the lifting (80.0 vs. a Fundamental score of 46.39) on a stock with negative margins and no earnings visibility, the stop-loss should be volatility-scaled rather than fixed at -5%, since a fixed stop on a name this fundamentally weak may just be pricing in a coin-flip catalyst reaction rather than a mispricing the model can actually predict — I should test whether ATR-scaled stops on similar catalyst-driven, fundamentally negative setups reduce this same-day whipsaw pattern.

BWIN post-mortem: -5.0% over a two-day hold, stopped out cleanly

The model liked BWIN for its catalyst setup — "high probability events within 30 days" — and a composite score of 52.1, ranked 7th, with Technical the strongest pillar at 57.92 and Sentiment sitting dead neutral at 50.0. That's a weak-conviction trade to begin with. A composite just above 52 and a cross-sectional percentile of 78.33 tells me BWIN looked good relative to the rest of the book that week, but the absolute signal was never strong. This wasn't a high-confidence call that got blindsided; it was a marginal call that didn't have room to be wrong.

And it was wrong fast. Two trading days, buy at 27.194, sell at 25.83, stopped out at the fixed -5% level. I don't have entry/exit technicals (both empty), no ATR figure to compare the move against, so I can't say whether this was a violent outlier move for BWIN or just noise the stop wasn't sized for. That's a real gap — deep_stats.move_vs_entry_atr is null, so I genuinely don't know if -5% in two days is one sigma or four sigma for this name.

What I do have is sector context, and it's damning for the "idiosyncratic bad luck" theory: insurance peers averaged -0.508% (a full percentage-point loss, not basis points — read that as roughly -0.5%) over the same window, across 20 names. BWIN's -5% is ten times worse than its peer average. So this wasn't sector rotation dragging everything down together — something specific to BWIN did the damage, even if I can't name it. Headlines, news_events, econ_events, gdelt_tone are all empty in my file — there's no recorded catalyst, earnings surprise, or macro shock I can point to. Macro backdrop was calm (VIX 18.77 at entry, regime "neutral" and unchanged through exit, no regime flip), so the broader tape isn't the explanation either.

Fundamentally, BWIN was already a mixed bag before this trade: negative ROE (-6.45) and ROA (-1.05), no usable PE or debt-to-equity, free cash flow at zero. The Fundamental pillar reflected that at 49.42 — below neutral. The model still ranked it 7 because Momentum, Catalyst, and Technical pillars outweighed that in the composite. In hindsight, weighting a name with negative returns on equity and assets that heavily on "upcoming catalyst" language, with no headline or event actually logged to back it up, was the soft spot.

Lesson: when the Catalyst pillar drives a stock's ranking but the dossier has zero corroborating headlines, news_events, or econ_events at entry, treat that pillar's score as unverified and downweight it — require at least one logged qualitative catalyst before letting "upcoming event" reasoning carry a trade with sub-53 composite conviction and negative profitability fundamentals.

EAT was a stop-out, clean and fast: bought 193.15 on 7/20, stopped at 182.24 by 7/23, a -5.6% hit against a fixed -5.0% stop. Three trading days, no ambiguity about why the position closed — the stop did its job. The question is why the model thought this was a profitable setup in the first place, and what actually drove the drawdown.

The entry case was thin even by the model's own accounting. Composite score 64.47, rank 8, with pillar scores clustered in the low-to-mid 50s — Momentum 50.87, Catalyst Fundamental 50.39 — basically coin-flip readings dressed up by a stronger Technical (60.71) and Sentiment (58.56) pillar. The reasoning field cites "upcoming catalysts" and the composite score, but Catalyst itself scored a flat meaning whatever event was anticipated wasn't actually differentiating the stock. Smart Money and ML pillars are null — no confirmation from either flow-based or model-based signals, and no predicted_return was even attached. This was a mid-pack, moderate-confidence pick riding partly on sentiment and technicals that I don't have granular entry/exit values for (both technicals and sentiment sub-fields are empty), so I can't verify what specifically flipped between entry and exit.

What I can see: the sector context was ugly. Peer average return across 15 Hotels/Restaurants/Leisure names over the same window was -0.903%, and the one headline actually about EAT's own name — "Chili's Is Slowing, Maggiano's Is Shrinking, It's Time To Sell" — was bearish in substance despite being scored "positive" by the sentiment classifier, which looks like a labeling miss on tone, not a bullish signal. A TD Cowen buy reiteration with a $210 target was neutral in score and didn't move the stock. Macro backdrop doesn't explain the drop — VIX fell from 18.77 to 16.64, HY OAS tightened, SPY's 20-day return improved from 0.5% to 1.8%, and regime stayed "neutral" throughout with no flip. So this wasn't a market-wide risk-off event; it reads as idiosyncratic underperformance layered on top of sector-wide softness that the model's Catalyst and Fundamental pillars failed to flag.

Lesson: when Catalyst and Fundamental pillars both sit at a neutral and the only name-specific headline is substantively bearish regardless of its sentiment score, don't let a Technical/Sentiment composite override that — cross-check headline text against its sentiment label before trusting the label, and treat sector peer weakness (-0.9% here) as a standing discount on any single-name long in that group.

ORKA post-mortem: -2.76% over 4 trading days, exited on the standard Friday schedule.

Start with what the model actually saw. Composite was 69.4, ranked #2 that week, sitting at the 95th cross-sectional percentile — that's a strong quantitative profile. But look at the pillar breakdown: Technical (79.91) and Momentum (66.31) were doing all the heavy lifting, while Catalyst sat at a neutral Sentiment at a middling 54.6, and Fundamental at a weak 36.94. Smart Money and ML were both null — meaning two of the model's inputs simply had nothing to say about this name. The reasoning text leans on "upcoming catalysts" and the composite score, but the catalyst pillar itself was priced at exactly i.e. no real edge, and the dossier has zero headlines, zero news events, zero econ events, zero GDELT tone, and empty sentiment dictionaries. So the "high probability event" framing in the reasoning wasn't backed by any observable data point in this file — it was more score-narrative than evidence.

On the fundamentals side, this was a name with negative operating margin (-40.31), negative ROE (-25.99) and ROA (-25.0), no PE, no EPS growth, no revenue growth data — a speculative biotech riding technical strength, not business improvement. That's a fine setup if the catalyst or momentum thesis plays out, but it means there was no fundamental floor if it didn't.

Quantitatively, the move itself wasn't extreme: 0.71x entry ATR, a normal-sized adverse move, not a blowup. The regime stayed neutral throughout (no flip), macro backdrop actually improved over the hold — VIX fell from 18.77 to 16.64, HY OAS tightened, SPY's 20-day return rose from 0.5% to 1.8%. Broader market breadth also improved (advancers overtook decliners by exit day). None of that explains a loss; if anything the tape was getting more constructive. The real tell is sector peer comparison: the biotech peer set averaged +3.19% over the same window with 18 names. ORKA didn't just fail to catch a sector tailwind — it moved in the opposite direction from its own peer group by close to 6 points. This looks idiosyncratic, not sector-wide or macro-driven.

Lesson: when Catalyst pillar reads exactly (neutral) and Smart Money/ML are null, don't let the reasoning text describe it as "high probability catalysts" — treat missing-data pillars as zero-information, not as latent support for the Technical/Momentum score, and discount confidence accordingly before ranking such names above peers with real catalyst or fundamental confirmation.

FRSH, post-mortem: bought 7/20 at 10.7574, stopped out 7/22 at 10.20, -5.18%, tripping the fixed -5% stop almost exactly on the nose. Two trading days, gone.

The entry case was thin by construction. Composite score of 59.93 landed FRSH at rank 3 with confidence 0.5993 — not a high-conviction call. Pillar breakdown shows why: Momentum was 49.26, basically coin-flip, and Catalyst sat at a flat meaning whatever "upcoming catalyst" the reasoning cites wasn't scored with any real edge, just flagged as present within 30 days. The pillars carrying the score were Fundamental (70.31) and Sentiment (62.92) — both backward-looking or soft signals, not anything that predicts a two-day price path. Smart Money and ML pillars are null, so the model had no positioning or machine-learned signal to lean on at all here. That's a real gap, not a rounding issue.

On the macro/technical side, I have almost nothing to work with — technicals entry/exit are empty dicts, deep_stats' move_vs_entry_atr is null, so I can't even say how large this move was relative to FRSH's own volatility. What I can say: the regime didn't flip (neutral to neutral, confidence 0.5 both ends), VIX was unremarkable at 18.77, and spy_return_20d was flat. So this wasn't a market-wide risk-off event. It also wasn't sector-idiosyncratic weakness the model should've caught — sector peers averaged -0.888% over the same window, so FRSH's -5.18% loss was roughly 6x worse than its own sector. This was idiosyncratic to FRSH, not systemic, and I have no headlines, news_events, or earnings_surprise data populated to explain the divergence. The dossier is silent on the actual cause.

That silence is itself informative: no single catalyst stands out in what I have on file, and the pillars that fed the score highest (Fundamental, Sentiment) are exactly the ones least connected to short-horizon price risk.

Lesson: when Momentum and Catalyst pillars both sit near (uninformative) and Smart Money/ML are null, the composite score is being carried by slow-moving fundamentals that have no business predicting a 2-day stop-loss outcome — treat sub-0.60 confidence entries with two or more null/neutral pillars as low-conviction regardless of composite rank, and either size them down or require at least one live technical/momentum confirmation before entry.

BLLN post-mortem: -4.05% over 4 trading days, exit forced by the Friday scheduled sell.

Start with what the model leaned on. Confidence was 0.6389, driven almost entirely by Technical (75.59) and Catalyst (73.33) pillars, with reasoning citing "upcoming catalysts" and a composite score of 63.9. Notice what's missing: Smart Money and ML pillars are both null, meaning two of the model's normal cross-checks simply weren't available for this name. The prediction also carries no predicted_return figure — confidence without a magnitude target. That's a thinner basis than it looks at first glance, xsec_pct of 98.33 notwithstanding — this stock ranked near the top of the cross-sectional pool that week, but rank alone doesn't tell you the move was going to be up.

The technicals block is a red flag on inspection: entry and exit technical snapshots are identical (same date, same RSI 59.63, same ATR 6.67), which tells me the technical picture wasn't actually refreshed intraweek — the "Technical" pillar score was likely stale relative to the actual holding period. I can't lean on RSI or MACD divergence as an explanation for the drawdown because I don't have technicals dated to the actual exit.

On the macro side, conditions improved, not worsened, during the hold — VIX fell from 18.77 to 16.64, high-yield spreads tightened (2.73 to 2.68), and SPY's 20-day return rose from 0.5% to 1.8%. Regime stayed "neutral" throughout with no flip. So this wasn't a market-wide risk-off move dragging BLLN down; breadth data shows advancers/decliners fluctuating but nothing like a broad health-care selloff. In fact, sector_peer_return shows the Health Care peer group averaged +0.613% over the same window across 16 peers. BLLN's -4.05% is a clear idiosyncratic underperformer against its own sector, not a beta story.

Scale of the move: -4.05% against an entry ATR of 6.67 (on a $132 stock, roughly 5% of price) means the move_vs_entry_atr of puts this within a single normal daily-volatility band — not a violent breakdown, just a below-average week that happened to run against the position with no offsetting bounce before the scheduled exit. Fundamentals (PE 782, PS 19) show a richly priced, expectations-heavy stock, which makes it more vulnerable to any soft disappointment, though I have no headlines, news_events, or earnings_surprise data logged to point to a specific catalyst that failed to fire. The dossier is silent on sentiment (empty at both entry and exit) and on headlines/econ_events (all empty arrays), so I can't pin this to a news-driven shock. No single catalyst stands out in what I have on file — this looks like a case where a stretched valuation, a top-ranked but catalyst-thin setup, and one bad week of idiosyncratic price action combined, without the macro or sector backdrop to blame.

Lesson: when Smart Money and ML pillars are null and the "Catalyst" pillar is driving the score, treat the confidence number as unverified until at least one of those cross-checks is populated — require peer-relative confirmation (a peer return check like this one) as a pre-trade gate rather than a post-mortem tool, since here it would have flagged that the model's catalyst thesis had no supporting sentiment or news data to lean on at entry.

SN — Post-Mortem

The model liked SN for one stated reason: upcoming catalysts within 30 days, packaged into a 65.2 composite that ranked it 5th and put it in the 85th percentile cross-sectionally. Pull that composite apart and it's a Fundamental pillar (65.61) and Sentiment pillar (62.11) doing the heavy lifting, with Momentum sitting at a mediocre 49.03. That's the tell in hindsight — I was buying a fundamentally decent, sentiment-decent name with no actual price momentum behind it, on the promise of a catalyst that isn't documented anywhere in this dossier. No headlines, no news events, no earnings surprise on file, no econ events. The "catalyst" pillar scored a flat — dead center, which in retrospect reads as filler, not signal.

The trade lasted two days and got stopped out at exactly the fixed -5% threshold, landing at -5.2%. That's a clean, mechanical exit, not a slow bleed — the stock moved fast against me. I don't have entry/exit technicals or an ATR reading (move_vs_entry_atr is null), so I can't say whether this was a large move relative to SN's own volatility or a routine one. That's a real gap: I'm flying blind on whether -5% in two days was two sigma or noise for this name.

Context doesn't explain it either. Macro was calm and static — VIX 18.77 at entry, yield curve and HY OAS barely moved, regime stayed "neutral" both days, no flip. Market breadth actually improved over the hold (advancers overtook decliners, pct above SMA50 rose from 62.4 to 69.2). Sector peers averaged over the same window. So this wasn't a market rout or a sector rotation dragging SN down — it was idiosyncratic, and the dossier simply doesn't contain the reason. No single catalyst stands out in what I have on file.

Lesson: when Momentum sits near 49 while Catalyst is a placeholder and no actual news/econ events back it up, treat the composite score as inflated by fundamentals/sentiment rather than confirmed by price action — require at least a positive momentum reading or a real, dated catalyst in the events feed before entry, not just a high composite percentile.