Weekly Results — Sells PAPER TRADING

SPSC returned 0.92% over four trading days against a predicted return of 0.1175% — the model wasn't calling for a loss here, but it also wasn't calling for much of anything. This was a low-conviction pick (rank 7, confidence 0.6752) that happened to land positive, and the gap between predicted and actual is wide enough that I don't think the framework earned this one so much as it collected on a position it was lukewarm about.

Look at what actually moved the stock. The Needham note on August 19 — Buy maintained, price target raised to $100 — landed the same day as a positive-sentiment piece on SPS Commerce's supply chain recognition (score 0.9432). That's a real, identifiable catalyst, and it shows up right at the point where RSI jumped from 63 to 69.6 and CCI ran from 79 to 119 between entry and exit dates. The move itself was 0.25x entry ATR, which is a modest, unremarkable magnitude — this wasn't a blowout day, it was a grind higher on decent news.

Here's the part that argues against the model getting credit: SPSC's sector peers averaged a 3.334% return over the same window across 44 names. SPSC returned less than a third of that. This wasn't SPSC outperforming on a differentiated thesis — it was a stock riding a tech-sector-wide bid while lagging its own peer group. The regime stayed bull, confidence dropped slightly (0.846 to 0.7815), and there was an FOMC event mid-hold that didn't visibly disrupt anything. Volume ratio actually fell from 1.29 to 0.79 into the exit, which doesn't support a story of accelerating conviction — it looks more like drift.

The pillar breakdown backs up the "no real edge" read: Catalyst scored a flat Fundamental 50.39, and Smart Money/ML were both null — meaning the composite score leaned heavily on Sentiment (80.82) and Momentum (63.61) without corroboration from the pillars that would indicate the model actually saw the analyst upgrade coming.

Takeaway: when Catalyst and Fundamental pillars sit at essentially neutral and Smart Money/ML are null, a positive outcome that underperforms its own sector peer average by this much should be treated as noise, not signal — don't let this trade's green P&L raise confidence in the model's catalyst-detection until it can show it caught price-target moves like Needham's before they hit tape, not after.

FRPT: A Small Win That Undersells the Setup It Was Sold On

Let me start with the number that actually matters here: predicted return of 0.2476% against realized return of 1.6228%. The model called this correctly in direction and roughly in the right neighborhood of magnitude—if anything it undersold the move. That's the first thing worth flagging before I get into mechanism, because a lot of these post-mortems end up explaining away a lucky number. This isn't that. This is a case where the framework's read and the outcome actually line up.

Now, what was the read? Composite score 75.94, rank 2, and the pillar breakdown tells a coherent story: Sentiment (73.63) and Technical (67.26) carrying the score, Momentum more middling at 58.92, Catalyst sitting at a flat despite the reasoning text citing "upcoming catalysts" as a driver — that's a soft signal at best, not a hard one. The technicals back up the Technical pillar: RSI at 70.4 on entry, MACD histogram positive and expanding, ADX at 34.5 climbing to 40.1 by exit — that's a trend strengthening into and through the hold, not fading. Price didn't need a catalyst to work; it needed the trend already in place to keep grinding, and it did.

Here's the part that tempers enthusiasm: this move was small relative to the stock's own volatility. Move vs. entry ATR was — well under half a normal daily range banked over four trading days. That's not a name that broke out; it's a name that drifted up gently while already extended (price sitting near the upper Bollinger band at entry, RSI in overbought territory at both entry and exit without ever cooling off). Overbought-and-staying-overbought is actually a decent signal in a strong uptrend, but it also means there wasn't a lot of room, and a good chunk of the "win" could just be beta plus drift rather than the specific catalyst thesis playing out.

Sector context cuts against a clean idiosyncratic story too. Food Products peers averaged a 2.809% return over the same window — FRPT's 1.62% actually lagged its own sector. So this wasn't FRPT getting singled out for anything; it was sector-wide strength, with FRPT participating at a below-average clip. That matters for calibration: if I'm crediting the model's Sentiment/Technical read for the win, I have to acknowledge that most of the peer group moved similarly or better, which weakens the idiosyncratic-pick argument somewhat, even though direction and rough size still checked out.

Macro backdrop was calm and supportive, not a tailwind story: VIX ticked up slightly (14.25 to 14.89), spy_return_20d actually decelerated from 4.4% to 2.8% over the hold, regime stayed bull-confirmed at both ends (0.846 to 0.7815, no flip). An FOMC event landed mid-hold (8/19) and didn't visibly disrupt anything — breadth wobbled day to day (advancers/decliners flipped negative on the 20th) but nothing regime-breaking. So no macro shock explains this; it's a case of a stable regime letting an already-strong technical setup keep working.

Takeaway: when the Catalyst pillar is a placeholder but Sentiment and Technical pillars are both elevated and technicals show ADX expanding through the hold, trust that combination for direction — it worked here. But don't over-credit magnitude calls when move_vs_entry_atr comes in under 0.5 and the peer group outperformed the pick itself; that's a sign the framework caught a real but generic sector tailwind more than it found something special in FRPT specifically, and next time I should check the peer return before calling a small win "confirmation" of the individual thesis.

TDW: A Small Win the Model Didn't Actually Call

Let me start with the number that matters most: the model predicted -0.1498% on this trade. It went long anyway — rank 1, composite 66.91, confidence 0.6691 — which tells me the selection was catalyst-and-score driven even while the model's own return forecast was negative. That's an odd setup on its face: you don't usually rank something first and then buy it expecting it to lose money, unless the ranking logic and the return forecast are running on different signals that aren't fully reconciled. Worth flagging on its own, separate from how the trade actually went.

The trade closed up 1.0074% over four days, exiting on the scheduled Friday sell. So the direction was right and the magnitude was small — but the model called for a loss and got a modest gain. This is case two: the win contradicts the entry read. I'm not taking credit for this one.

What actually happened, as far as the dossier shows? Sector peers in Energy averaged 1.95% over the same window — TDW's 1.0074% actually underperformed the peer average. That's a meaningful tell: this doesn't look like an idiosyncratic, TDW-specific catalyst playing out. It looks like sector-wide drift that TDW rode partway. The Fundamental pillar was the strongest input at 84.31 (cheap-ish PE of 12.3, ROE near 24%, EPS growth of 78.48% against revenue growth of -0.92% — an odd combination worth its own scrutiny another day), while Catalyst sat at a neutral meaning the "upcoming catalysts" language in the reasoning wasn't backed by a strong catalyst score. There was an FOMC event on 8/19, right in the middle of the hold, and the regime stayed "bull" throughout (confidence dipped from 0.846 to 0.7815, not a flip). VIX ticked up slightly (14.25 to 14.89), SPY's 20-day return decelerated (4.4% to 2.8%), and breadth was choppy — decliners outnumbered advancers on both the entry and exit days. None of that reads like a tailwind; if anything the macro backdrop softened slightly over the hold.

I don't have technicals, sentiment, headlines, or ATR data populated here — no entry/exit technical readings, no news events, no move-vs-ATR ratio to gauge whether this was a big move for TDW or noise. That's a real gap. Given the size of the gain (about 1%) I'd guess this sits within normal daily noise for an energy name, especially since sector peers moved further. Without ATR I can't confirm that, so I'll flag it as unavailable rather than assume.

Takeaway: when the model's predicted_return is negative but it still ranks a stock first, that's a red flag on internal consistency, not a reason to expect a lucky reversal — I should check how often ranking and predicted_return disagree, because if this trade underperformed its own sector while "working," the model likely got the direction right for the wrong reason (or no reason it can articulate) and I should not treat this as evidence the negative-forecast-but-top-ranked pattern is exploitable.

FRSH: A Directionally Right Call That Didn't Deliver on Scale

The model predicted a 0.319% return for FRSH. It got 0.036% — positive, but roughly one-ninth the size of what it called for. This is a case where I have to be careful not to pat myself on the back for a number that technically cleared zero but massively undershot the thesis.

Start with the composite: 80.22, driven by Fundamental (84.31) and Catalyst (75.0) pillars, with Momentum sitting weaker at 54.89. The reasoning cited "upcoming catalysts" and a high composite score — but the dossier has no news_events logged and only one headline, a generic "profitable stock" roundup piece with a positive sentiment score of 0.8555. That headline isn't a catalyst in any specific sense; it's noise that happened to be favorable in tone. So the "catalyst" thesis that justified a chunk of the score isn't visibly confirmed by anything in the record.

What actually happened during the hold is more informative than what was predicted. The regime stayed bull throughout (confidence dipped from 0.846 to 0.7815, but no flip), and there was an FOMC event on 8/19 sitting right in the middle of the four-day hold — that's typically a volatility-suppressing or volatility-inducing wildcard, and macro shows VIX ticking up slightly (14.25 to 14.89) and SPY's 20-day return decelerating (0.044 to 0.028) over the same window. That's a mildly cooling tape, not a tailwind. Market breadth also weakened into the exit — advancers/decliners flipped negative on both 8/17 and 8/20 (62/101 and 56/107), suggesting the broader market was not pulling this stock up.

The sector comparison is the clearest signal here: Technology peers averaged a 3.336% return over the same period, across 44 names. FRSH's 0.036% badly lags its own sector. This wasn't a case of a rising tide; if anything, this stock underperformed its peer group while the peer group had a strong week. That argues against the "catalyst-driven idiosyncratic winner" story and toward "sector had a good week, this name barely participated."

Fundamentals look genuinely solid on paper — 84.96% gross margin, 15.57% revenue growth, 19.5% ROE — but those are structural, slow-moving inputs. They don't explain a four-day price move and they didn't explain much of the eventual return magnitude either, since the model already had this information going in and still overshot the prediction by an order of magnitude.

There's no ATR data available (move_vs_entry_atr is null) so I can't size this move against the stock's own volatility profile, which limits how confident I can be in calling this move "small" in a statistically rigorous sense — I'm inferring smallness mostly from the raw percentage and the sector gap.

Net read: this is case one dressed up as a win. Predicted return and actual return were both positive, so directionally the model was right, but the magnitude miss is large enough, and the underperformance versus sector peers is stark enough, that I don't think the entry thesis was actually confirmed. The catalyst pillar looks like it was doing work in the score that isn't backed by anything concrete in the record.

Takeaway: when the Catalyst pillar drives a big chunk of the composite score but the headlines/news_events fields are empty or generic, treat that pillar with more skepticism going forward — cross-check catalyst-weighted picks against sector peer returns before trusting the composite score, since a "win" that badly lags its own sector isn't the same signal as a win that beats it.

TOST made 4.98% over four trading days on a trade the model priced at -0.34% expected return. That's not a marginal miss — the model's directional read was wrong, and I need to treat this as a case where the win contradicts the entry thesis rather than confirms it.

Look at what the model actually saw at entry: a composite score of 70.09, driven by Technical (73.69) and Sentiment (63.78) pillars, but Momentum sitting at a lukewarm 50.74 and Catalyst flat at 50.0. Smart Money and ML pillars were both null — no signal there at all. The reasoning field cites "upcoming catalysts" as the selection driver, but the predicted return still came out negative. That's an internal contradiction worth flagging on its own: the model ranked TOST #6 and gave it 70th-percentile cross-sectional standing (xsec_pct 81.25), yet still expected it to lose money. Whatever combined those pillars into a final number was pulling against the raw technical picture — RSI at 61, ADX above 31 signaling a real trend, price sitting well above SMA20/50/200 with room to the upper Bollinger band. That's a stock in an established uptrend, not one screaming reversal.

So what actually drove the 4.98%? The move was 1.34x the entry ATR, a clean, single-day-scale breakout, not an explosive gap that dwarfs typical volatility. The RSI, CCI, and MFI all climbed further into strength by exit (RSI 65, CCI 75, MFI 62), and volume ratio stayed below 1 the whole time — this was a low-volume grind higher, not a news-driven surge on heavy participation. The headlines support that: a "High Growth Momentum Meets a Technical Breakout Setup" piece landed on the exit date itself, and a "partnership wins" story hit on entry day with a strong positive score (0.97). None of these are surprise catalysts — no earnings surprise is on file, no news_events entries, no options flow. The FOMC meeting fell inside the hold window (Aug 19) but the regime stayed "bull" throughout with confidence merely softening (0.846 to 0.782), and VIX ticked up marginally (14.25 to 14.89) — nothing regime-breaking. Sector peers averaged 1.31% over the same window across 34 names, so TOST's near-5% gain was meaningfully idiosyncratic, not just financial-services beta.

Net read: this looks like a technical continuation trade that worked because the technical picture was genuinely strong, while the model's own predicted-return number undersold what its own Technical and Sentiment pillars were telling it. The composite score and rank got it right; the point-blank return forecast did not.

Takeaway: when the Technical and Sentiment pillars are both comfortably above 60 and Momentum/Catalyst are merely neutral rather than negative, don't let a negative predicted-return output override the composite — on this evidence, the blended score was the better signal than the point forecast, and that discrepancy is worth tracking across more trades before trusting either one exclusively.