Weekly Results — Sells PAPER TRADING

SPSC, entry to exit: +3.32% over four trading days, no model conviction on record. The prediction pillars — Smart Money, ML, weights, xsec_z — are all null. There's no expected return to check against, which means I can't call this a confirmed thesis. It's a blind buy, held into a scheduled Friday exit, that happened to work.

What I do have is the technical picture at entry, and it's not subtle. RSI at 68.6, MACD histogram positive, ADX at 36.4 (a genuinely strong trend reading), CCI at 126, Williams %R at -7.9 — every momentum gauge was pinned toward overbought/strong-trend territory, and price was sitting right at the upper Bollinger band (83.73 vs. 82.19 entry). This is a stock already extended, not a mean-reversion setup. On volume ratio of 0.45, the move up to that point hadn't even come with heavy participation. So on the technicals alone, this reads as a trade taken into strength, and the market kept giving it strength for another four days.

Where this gets interesting is the peer comparison: sector average return for the week was 5.9% against SPSC's 3.3%. This wasn't a name outperforming its group — it lagged its own sector by roughly half. That reframes the win: this wasn't a stock-specific catalyst, it was a broad Technology tape rally lifting nearly everything, and SPSC caught a smaller share of it. The regime flip (neutral, 0.50 confidence, to bull, 0.83 confidence) over the hold corroborates that — the whole market environment turned more favorable mid-week, and SPY's 20-day return of 3.9% at exit backs that up. VIX dropped to 14.51, HY OAS tightened to 2.63. Risk appetite broadly improved.

Move-vs-ATR came in at 0.97 — basically one full average true range of movement over four days, which is a normal-sized move for this stock, not an outlier. No headlines, no news events, no earnings surprise, no econ events in the file. Nothing idiosyncratic drove this.

So: no expected return was on record, the entry technicals show a stock bought while already extended, the move size was unremarkable relative to its own volatility, and the return underperformed the sector average during a week the whole tape turned bullish. This is a beta win, not a called win — the framework didn't identify SPSC as special, and SPSC didn't behave as special. It rode a rising market and landed a below-average share of the sector's gain.

Takeaway: when the prediction pillars are null and the sector peer return beats the trade's own return, treat the win as regime-driven rather than stock-selection-driven — the next test is whether this same technical profile (overbought RSI, high ADX, price at the upper band, low volume ratio) produces a gain when SPY isn't also up nearly 4% over the trailing 20 days; if it doesn't, the "confirmation" here was really just the market regime flip doing the work.

MSI, week of 8/24: bought at 481.03, sold at 483.98 on the Friday scheduled exit, +0.61%. Before I say anything else about drivers, I have to flag what's actually in the prediction field: both pillars (Smart Money, ML) are null, weights null, xsec_z null. There is no recorded expected return here. This isn't a case of "the model called it and got confirmed" — there's no call on file to confirm. Whatever generated this trade, I don't have its reasoning stored. So this write-up is forensic, not a victory lap.

What actually happened: the move was small in absolute terms and small relative to the stock's own volatility — 0.27x ATR-14 (11.117 at entry). That's a quarter of a normal daily range over four holding days. This is not a trade that "worked" in the sense of catching a big dislocation; it's a trade that drifted gently in its favor inside noise.

Context that could explain the drift: technicals were already extended and got more extended — RSI ran from 66.7 to 68.2, ADX rose from 37.8 to 41.0, MFI jumped from 60.9 to 72.7, price hugging the upper Bollinger band the whole time. That's a stock in a live uptrend, not a mean-reversion setup, so a small continuation isn't surprising. There's also a real catalyst: Evercore raised its price target to $550 on 8/26, mid-hold, plus a completed M&A deal (D-Fend buyout) flagged CRITICAL/BULLISH at entry. Both are genuine positive inputs.

But scale matters: sector peers averaged +4.78% over the same window — MSI's +0.61% badly lagged its own sector. This wasn't sector-wide lift carrying the stock; if anything MSI underperformed peers while still landing green. The regime flipped neutral-to-bull during the hold (VIX dropped to 14.51, spy_return_20d +3.9%), which is a tailwind, but a weak one given the peer gap. Market breadth also deteriorated hard on 8/27 (45 advancers vs 118 decliners) right before the exit, and volume ratio collapsed from 1.57 to 0.55 — conviction dried up into the sell.

Takeaway: with no stored prediction to grade, don't credit the framework for this one — treat it as a small, low-conviction drift trade that happened to close positive despite lagging its sector by over 4 points, and next time a trade like this shows up, check whether the null-pillar cases correlate with worse average outcomes before assuming "no signal" trades are neutral rather than a hidden risk flag.

No narrative available for this trade (generation was disabled, errored, or has not run yet) — the mechanical facts above are real and on file regardless.

SRRK closed the week up 0.63% on a scheduled Friday exit, not a signal-driven one — which is itself worth flagging before anything else. There was no entry prediction to grade here. Both pillars (Smart Money, ML) are null, weights are null, xsec_z is null. The model didn't have a read on this trade going in, positive or negative. So there's no thesis to confirm or contradict — this is a case where the framework had nothing to say, and the tape moved anyway.

What actually happened during the hold looks like beta, not alpha. SPY's 20-day return at exit was +3.94%, the regime flipped from "neutral" (0.50 confidence) to "bull" (0.83 confidence) over the four days, and VIX settled at a calm 14.51. That's a market-wide tailwind. Against that backdrop, SRRK's 0.63% gain is unremarkable — and it's actively weak relative to its own sector: biotech peers averaged +3.512% over the same stretch, more than 5x SRRK's return. If this were a clean market-beta trade I'd expect it to track the rally; instead it barely participated while the group ran. That's a sign the position was closer to inert than it was to catching a real move.

The lone headline — a Scholar Rock investor-conference notice on Aug 27, sentiment score 0.49 — is a mild positive at best, not the kind of catalyst that explains even a small outperformance, and it arrived only a day before exit. Market breadth was choppy all week (advancers/decliners flipped hard on the 27th), consistent with no clean directional signal locally. I don't have move-vs-ATR here, so I can't say how this ranks against SRRK's own volatility, but the magnitude is small enough that "noise" is a live explanation.

Takeaway: a profitable trade with a null prediction and sub-sector-average return isn't validation of anything — treat it as a coin flip that landed heads, and don't let three or four of these accumulate uncorrected in the win column before asking why the model had no view on names moving with the broader tape at all.

EXK, +1.3% over four trading days, exit via Friday scheduled sell. Before I talk about what happened, I have to flag what didn't: there's no prediction on file. Both pillars (Smart Money, ML) are null, weights are null, xsec_z is null. This wasn't a case of the model calling a small gain and getting a small gain — there's no read to check the outcome against at all. So this is squarely a "no thesis, no credit" situation, not a confirmed call.

Given that, the honest exercise is just describing what actually drove 1.3%, not pretending the model saw it coming. The move itself was modest relative to the stock's own volatility — 0.23x ATR, meaning the four-day gain was well within a single day's typical range for EXK. That's a small, unremarkable move by the stock's own standards, which matters: it lowers the bar for what needs to have caused it. Sector context confirms this — Metals & Mining peers averaged +2.06% over the same window across 21 names, so EXK actually underperformed its own sector. This wasn't an idiosyncratic win, it was a sector-wide drift that EXK only partially captured.

The news flow is mixed and doesn't point to a clean catalyst either: a mildly negative "overvalued" headline, a neutral insider-sale note, and a positive but unremarkable Terronera-restart/blockade-removal story cluster around entry. Technicals show RSI cooling from 63 to 59 and MACD histogram compressing from 0.109 to 0.030 — momentum was fading, not building, during the hold, even as price ticked up on rising SMA20. The regime flipped from neutral to bull (confidence 0.83) during the hold, and broader tape (SPY +3.9% over 20 days, VIX down to 14.5) was clearly getting friendlier — that's plausibly the real tailwind, a sector/market-beta effect more than anything stock-specific.

Takeaway: with no stored prediction, this is a coin-flip in the record book that happened to land heads — don't let a small positive print like this get counted as evidence the framework "worked" on EXK; the more useful audit is whether pillars are actually populating for trades like this, because a missing prediction turns every outcome into an unfalsifiable win.