Weekly Results — Sells PAPER TRADING

DXCM was a rank-5 pick with a composite of 60.1 — not a high-conviction entry to begin with. The reasoning cited "upcoming catalysts" and a solid composite, and the pillar breakdown backs that up loosely: Fundamental (65.61) and Sentiment (64.79) were the strongest legs, Catalyst sat at a flat and Momentum was middling at 54.33. Smart Money and ML pillars were null, so the model was leaning on fundamentals and sentiment carrying a stock that had just printed a 52-week high and broken above its 20-day moving average — both signs of a rally that had already run before I got in.

That's the first flag in hindsight: two headlines flagged the stock as freshly extended (the 52-week-high piece and the 20-day breakout piece, which itself scored negative at -0.5267 despite the "positive" framing of price action). Buying into a name already stretched, on a composite score that wasn't particularly strong, is a lower-margin bet by construction.

The macro and regime backdrop also shifted under me. Entry regime was "neutral" at 0.5 confidence; by exit it had flipped to "bull" at 0.7987 confidence. A regime flip mid-hold isn't inherently bearish, but it means the environment the model scored the trade in doesn't match the environment it exited in — that's a real source of mismatch, not noise. Layered onto that, NFP landed on the sell date itself, and market breadth had already started souring the day before exit: decliners outnumbered advancers 88 to 75 on 8/6, reversing the 4th and 5th's healthy breadth. So the tape turned negative right into the scheduled Friday sell.

Peer comparison makes this look idiosyncratic, not sector-driven: Health Care peers averaged +1.053% over the same window while DXCM fell -2.28%. Whatever hit this stock, it wasn't a sector-wide rotation — twenty peers were positive on average while this one specific name reversed. I don't have ATR or technical entry/exit data logged (both come back null), so I can't quantify how far this move sits relative to DXCM's normal daily range, which limits how confident I can be about whether this was a routine pullback or an outsized move.

Lesson: when Catalyst pillar is a flat (no real edge) and the entry news already shows the stock at a 52-week high breaking out, treat that as exhaustion risk rather than confirmation — cross-check headline-implied extension against sector peer momentum before sizing a rank-5, sub-conviction score as a live position.

DK Post-Mortem — entered 2026-08-03, stopped out 2026-08-05

The model bought DK on a composite score of 60.2, ranked 4th, with a confidence of 0.6018. Looking at the pillar breakdown, this wasn't a conviction trade on fundamentals or technicals — Momentum (50.75), Technical (50.14), Catalyst and Fundamental (54.42) were all sitting near the middle of the range. The only pillar doing real work was Sentiment at 78.66, and the reasoning field confirms it: the stated thesis was "upcoming catalysts" with high probability events in the next 30 days, not a technical or momentum setup. Smart Money and ML pillars are both null, so two potential corroborating signals simply weren't available at entry.

The trade lasted two calendar days and was stopped out via the ATR-based stop-loss at -10.2%, worse than the -10.0% threshold. That's a fast, violent move — a two-day double-digit drawdown in a stock the model rated as only moderately confident. I don't have entry/exit technicals or an ATR value stored (deep_stats.move_vs_entry_atr is null), so I can't quantify precisely how many ATRs this move represented, but the stop-loss design itself tells me the move was large enough to breach a volatility-calibrated band almost immediately after entry.

Sector context is informative here: Energy peers averaged -0.823% over the same window versus DK's -10.18%. This was not a sector-wide rotation — it was idiosyncratic to DK, and sharply so. The regime stayed "neutral" throughout with unchanged confidence (0.5), and macro conditions were calm and stable — VIX ticked up only slightly (15.99 to 16.5), the yield curve barely moved, and market breadth was healthy (advancers exceeding decliners on the entry day). None of the macro or regime data explains a move this violent. Unfortunately, headlines, news_events, econ_events, and gdelt_tone are all empty arrays in this dossier — so whatever idiosyncratic catalyst hit DK in this two-day window isn't captured in what I have on file. No single catalyst stands out in the data I'm holding; I can only say the damage was company-specific and sharp, not macro or sector-driven.

Fundamentally, the setup was already shaky: negative revenue growth (-9.53%), negative ROE and ROA, a current ratio below 1 (0.7577) indicating tight liquidity, and leverage at 1.62x debt-to-equity. None of that shows up in the Fundamental pillar score (54.42), which is oddly neutral given these figures — that pillar may be underweighting balance-sheet fragility in favor of the eps_growth headline number (95.67%), which is likely noise off a small or negative base given net margin is -0.21%.

Lesson: when the Sentiment pillar is the primary driver of a buy (78.66 vs. everything else near and the reasoning cites "upcoming catalysts" without naming a computed edge (predicted_return null, weights null), treat that as a high-variance, event-driven bet rather than a scored conviction trade — and require the Fundamental pillar to actually penalize weak liquidity (current ratio < 1) and negative revenue growth before entry, rather than letting a single distorted growth metric hold it near-neutral.

LPG post-mortem: -8.9% over a two-day hold, stopped out at -8.6% ATR.

The entry case was thin from the start. Composite score 59.1, rank 9, confidence 0.59 — not a high-conviction setup, and the reasoning field is explicit that the pick was driven by "upcoming catalysts" and the composite number itself, not by any specific edge I can point to. Look at the pillar breakdown: Technical (78.71) and Fundamental (73.81) were doing all the work, while Momentum (53.82), Catalyst and Sentiment (52.2) were mediocre to flat. Smart Money and ML pillars are both null — meaning two of the model's inputs simply weren't available for this name, and the score leaned entirely on the two pillars that turned out not to matter once price moved.

The technical picture at entry actually reads as stretched, not just "strong": RSI 67.8, MFI 80.6, CCI 121.8, price sitting right at the upper Bollinger band (47.99 vs. buy price 47.39). That's a name already extended into overbought territory being bought for more upside. When it reversed, it reversed hard — the move was 2.61x the entry ATR, well beyond a normal one-ATR stop-loss band, which tells me this wasn't noise, it was a genuine directional break that blew through where the stop was designed to catch it.

Context matters here too. Sector peers in Energy averaged -0.82% over the same window — so the sector was soft, but LPG's -8.9% is roughly 10x that peer average. This was not sector drag, it was idiosyncratic to the name. Macro was calm (VIX 15.99→16.5, no regime flip, breadth still solidly positive at ~70% of stocks above their 50-day), so the broader tape isn't the culprit either. And critically, there's nothing in headlines, news_events, econ_events, or gdelt_tone — all empty. Whatever moved this stock 8.9% in two days isn't captured anywhere in my data, which I have to be honest about rather than paper over.

Lesson: when Smart Money and ML pillars are null and the composite is being carried almost entirely by Technical/Fundamental scores while the stock is already sitting at RSI>67 and the upper Bollinger band, treat that as reduced conviction, not confirmed conviction — a stretched technical setup with two missing pillars shouldn't score high enough to rank in the top 10 without a real catalyst to justify the extension.