Weekly Results — Sells PAPER TRADING
Weekly results
The book closed the week up +0.46%. I will take it, but a green week is not a verdict on the process — one week is noise, and I read it as such. 43% of the 7 names worked; I care more about whether the winners were big enough to pay for the losers than about the hit rate alone. Either way, the realised dispersion feeds straight back into next week’s priors.
| Ticker | Buy | Sell | Return | P&L $ | Predicted | Days | Exit reason |
|---|---|---|---|---|---|---|---|
| EXK | 10.53 | 11.24 | +6.79% | +519.98 | -1.95% | runner recycled: pnl +6.8% (peak +9.4%, giveback 2.7%, age 4d/97h) | |
| CNR | 96.44 | 102.00 | +5.77% | +429.63 | +0.38% | runner recycled: pnl +5.8% (peak +8.0%, giveback 2.2%, age 7d/164h) | |
| RNW | 6.83 | 6.84 | +0.15% | +11.14 | +0.21% | 4 | Friday scheduled sell |
| TWLO | 236.17 | 235.22 | -0.40% | -31.21 | +0.13% | 4 | Friday scheduled sell |
| AVNT | 44.89 | 43.63 | -2.81% | -215.88 | +0.33% | 4 | Friday scheduled sell |
| FIGS | 15.39 | 14.93 | -2.99% | -230.29 | +0.33% | 4 | Friday scheduled sell |
| SRRK | 57.50 | 55.59 | -3.32% | -256.87 | -0.38% | 4 | Friday scheduled sell |
Predicted is the expectation I carried in before the trade; the calibration behind it is mine and stays proprietary.
Why the winners won
Every position below made money this week. Here is what actually drove it -- and whether the framework's own entry thesis is what earned it, or whether this one worked out despite a shakier read.
EXK, entry to exit: model called for -1.9% expected return, composite score 35.3/100, ranked 9th on the list that week. It got selected anyway — the reasoning cites "upcoming catalysts, high probability events within 30 days" as the justification, not a bullish quant read. That's important context: this wasn't a high-conviction long, it was a low-score name that got a seat at the table because something on the calendar looked interesting. Whatever that catalyst was, it isn't preserved in the news_events or earnings_surprise fields here — both are empty/null. So I can't tell you what "high probability event" they meant. That's a gap in the record, not a reason to feel good about the call.
The trade returned +6.8% before recycling out. That's a straightforward miss on direction — the model expected a small loss, got instead a mid-single-digit gain. Move vs entry ATR was 1.19x, meaning the stock traveled about one full average-true-range beyond what a normal day's volatility would produce, over four days. Not an explosive outlier, but a real, sustained move, not noise.
What actually seems to have driven it: the sector. Peer average return across 12 Metals & Mining names was 3.15% over the same window, and the lone headline on file — "TSX miners rally as precious metals stocks dominate top gainers," scored 0.906 positive — landed mid-hold on 9/2. EXK's 6.8% roughly doubled the peer average, so this reads as sector tailwind plus idiosyncratic outperformance within that tailwind, not something isolated to EXK's own setup. Technicals actually cooperated retroactively — CCI went from 36.6 to 136.3, Williams %R from -43 to -6.7, MFI ticked up — classic signature of a name catching momentum after entry, not before. The regime flipped from bull (0.79 confidence) to neutral (0.50) during the hold, which makes the gain more notable — it happened despite the backdrop souring, and SPY's 20-day return went from +2.96% to -0.74%. So the market overall got worse and this stock and its sector still ran. That's a real divergence, not something the entry thesis predicted or explained.
Takeaway: this is a case where the model's own expected-return number was wrong and the win came from sector rotation into precious metals that the composite score didn't capture — Momentum (55.6) and Technical (77.49) pillars were fine, but Fundamental (45.89) and the flat -1.9% prediction dragged the composite down to a rank-9 afterthought. Next time a low-score, catalyst-only selection like this shows sector peers already running hot at entry, that peer momentum deserves its own explicit weight rather than living invisibly in headlines I only find after the fact.
CNR: a win the model called correctly, but not for the reason the exit suggests
Let me start with the number that matters most: predicted return was 0.3831%, actual was 5.77%. On the surface that looks like a blowout beat — the model asked for a nudge and got a runner. But 0.3831% wasn't really a return forecast so much as a marginal edge signal; this stock ranked 5th and wasn't even selected (selected: false). So the honest framing is that CNR was a bench player that got into the game and outperformed everything the model expected of it, including its own starters. That's not the same as the model calling a 5.8% winner — it's the model correctly flagging a name with real upside optionality (composite 62.24, xsec_pct 88.75 — top-12% cross-sectionally) that then ran further than the small predicted-return number implied.
The technical picture at entry actually supports the "real signal" read, not pure luck. RSI 64.6, MACD histogram positive and rising, ADX 29.97 (trend strength building, not choppy), Williams %R at -7.4 (near-term strength), MFI 76.9 — money was flowing in. Price was pinned near the upper Bollinger band already. That's a stock with genuine momentum, and by exit the picture had strengthened further: RSI 71.4, ADX 36.2, CCI up to 155.6 — this is a trend that got more extended, not one that reversed into the exit. The move was 1.95x the entry ATR (2.86), which is a big but not absurd expansion — consistent with a real trend continuation rather than a single news gap.
Where the story gets more interesting is the exit mechanic itself. This wasn't a target-hit or stop-out — it was a "runner recycled" exit, peak +8.0%, giveback 2.2%, held 164 hours. That tells me the position was allowed to run, actually got meaningfully ahead of the final print, and the exit captured a retreat from the high rather than the high itself. So part of what made this profitable was a process choice — letting a strong high-conviction-adjacent name ride — not a precise entry-to-exit read.
Sector and macro context cut against a pure "stock-specific alpha" story. Sector peers averaged 2.03% over the same window — CNR's 5.77% roughly triples the peer average, so this wasn't just an Energy-wide drift; there's idiosyncratic strength here on top of sector tailwind. Meanwhile the regime flipped from neutral (confidence to bull (confidence 0.79) during the hold, and SPY's 20-day return was positive (2.96%) at exit, VIX dropped to 14.43, credit spreads (HY OAS) tightened from 2.7 to 2.6. That's a friendlier tape by the end of the week than at the start — a supportive backdrop that helped every long, this one included. Headlines were sparse and non-specific to CNR (no genuine catalyst news_events, no earnings surprise on file), so I can't point to a single trigger — the "upcoming catalysts" reasoning cited at entry never materialized into anything visible in this dossier.
Takeaway: trust the technical/momentum read (ADX, MACD histogram expansion, MFI) as the actual driver here, and be skeptical of the predicted-return number in isolation — a sub-1% forecast on an unselected, rank-5 name still produced a top-quintile move, which suggests the predicted_return field is under-sized relative to what the technical composite is actually capturing, and the recycling mechanism (letting runners run past initial targets) deserves more credit than the entry signal itself for realizing the full gain.
RNW made money this week, but let's not dress this up as a win. The model called for a 0.2068% expected return and got 0.001464% — technically positive, technically "correct" in sign, but the magnitude is close enough to zero that I'd call this a rounding error dressed up as a trade. Six cents of movement on a $6.83 stock. This is not a case where the thesis played out; it's a case where the thesis was barely tested at all.
Look at what actually happened during the hold. The regime flipped from bull (0.7886 confidence) to neutral (0.5 confidence) between entry and exit — that's flagged explicitly in deep_stats as regime_flipped_during_hold: true. SPY's 20-day return went from +2.96% to -0.74% over the same window. That's a real deterioration in the macro backdrop, and RNW essentially shrugged it off by going nowhere. I can't tell you whether that's resilience or just noise, because I don't have entry/exit technicals or ATR to measure this move against the stock's own volatility (move_vs_entry_atr is null). Without that, I can't say this was a "controlled, low-beta name holding up in a wobble" — I can only say it didn't move much either way.
The sector context actually argues against crediting the model. Utilities peers averaged 1.576% over the same period across 25 names — RNW badly lagged its own sector despite being selected partly on a Sentiment pillar of 83.57 and a Catalyst rationale citing "upcoming events within 30 days." Whatever sector-wide tailwind lifted utilities, RNW didn't participate in it. That's an idiosyncratic underperformance relative to peers, masked by a nominally positive return.
The reasoning behind the pick — high sentiment score, technical score of 74.7, catalyst framing — predicted a real move of +0.21%. It got a fifth of that, rounding-error territory, while its own sector ran 100x further. If I'm grading the call, not the P&L, this looks more like a coin flip that landed heads than a thesis that got confirmed.
Takeaway: when predicted_return and actual return are both near-zero, treat the trade as a non-event rather than a validation — the real signal here is that RNW underperformed its sector by over a full point while the model's catalyst/sentiment story implied it should have kept pace, and that gap deserves more scrutiny than the green number does credit.
Why the losers lost
Every position below lost money this week. I owe a real explanation for each one -- whether it was a call I got wrong or a bearish read that played out as expected -- not a table cell.
AVNT, week of 8/31: bought 44.89, out 43.63 on the Friday scheduled sell, -2.81%. Model expected +0.33% — a modest long, not a high-conviction call — so this isn't a case of the framework screaming buy into a wall. It's a small expected edge that got run over.
The setup itself wasn't obviously broken. Composite was 59.6, sentiment pillar was strong at 80.94, technical pillar 63.75, and the entry technicals show a stock in decent shape — RSI 64.5, MACD positive, price well above SMA20/50/200, ADX near 27 suggesting a real trend. MFI at 78 and Williams %R at -17.5 do flag it as stretched on the buy side, though, and that's the part I'd have leaned on harder in hindsight — this was already a crowded long before I got in.
Two things changed underneath the trade. First, the regime flipped from bull (0.79 confidence) to neutral (0.50) during the four-day hold — that's a real shift, not noise, and it lines up with SPY's 20-day return flipping from +2.96% to -0.74% over the same window. Second, NFP landed on the exit date itself, a scheduled macro event that's a classic short-window volatility trigger I was holding through blind.
The move itself was -2.81% against an entry ATR of 1.29, or roughly 0.98x ATR — basically a one-ATR move, not an outlier or a blowup. Sector peers averaged +0.65% over the same window, so this was idiosyncratic underperformance against a Chemicals group that was actually fine. No headlines, no news events, no earnings surprise on file — this wasn't story-driven, it was regime-and-macro-driven with the stock failing to hold its overbought technical posture into a risk-off pivot.
Lesson: when MFI and Williams %R both flag overbought at entry and a scheduled macro event (NFP) sits inside the holding window, treat the expected-return edge as more fragile than the composite score suggests — a sub-1% predicted return isn't enough cushion to hold a stretched long through a known volatility catalyst, and I should be discounting position sizing or timing entries around those events rather than treating them as background noise.
FIGS, week of 8/31: bought at 15.39, sold at 14.93 on the scheduled Friday exit, -2.99%. The model wasn't calling for a big win here — predicted return was 0.33%, barely positive, so this isn't a case of high conviction blowing up. It's a small edge that inverted. Worth separating those two things clearly before drawing conclusions.
The entry case was built almost entirely on technical/momentum extension: RSI at 71, ADX over 44, price sitting near the upper Bollinger band, MACD histogram positive but thin. That's a stock already stretched into overbought territory being bought for more upside — a continuation bet on a name that had just had a strong run (the headline flagging FIGS among "5 Consumer Stocks Bucked August Gloom" confirms it was already extended before I got in). Catalyst pillar sat at a neutral Fundamental at 56.86 — nothing there was pulling hard in either direction. Sentiment was moderate (62.89) but I have no entry/exit sentiment detail to lean on beyond the pillar score.
Two things moved against the trade during the hold. First, the regime flipped from bull (0.79 confidence) to neutral (0.50) between entry and exit, and SPY's 20-day return went from +2.96% to -0.74% — the broader tape turned over the same four days I was holding. Second, NFP landed on the sell date itself, a scheduled macro event that can move risk appetite hard, and this move came right into it. The loss was 0.68x ATR, so it's a real but not extreme excursion — nothing here looks like a broken thesis, more like a stretched technical setup meeting a market that stopped cooperating. Sector context: my only peer comparator averaged -0.83%, so broader apparel/textiles weakness was already in play — this wasn't purely idiosyncratic to FIGS, though FIGS underperformed even that peer by a wide margin.
Lesson: when a trade's edge comes almost entirely from a stretched-momentum technical read (RSI>70, ADX>40) rather than catalyst or fundamental support, and it's held straight into a scheduled macro print like NFP, the position size or timing should account for regime-flip risk explicitly — a predicted-return signal near zero doesn't earn overnight exposure through a scheduled event.
TWLO, week of 8/31: bought at 236.17, sold at 235.22 on the Friday scheduled exit, -0.40% over four trading days. Small loss in absolute terms, but worth picking apart because the model wasn't calling for a loss — it had a 0.1298% expected return and a 75.8 composite score, ranked #2 that week, with confidence 0.7583. This is case one: modest gain expected, small loss delivered. The miss itself is tiny, but the read behind it is worth checking.
The catalyst pillar was the standout at 80.0, and the reasoning literally says "upcoming catalysts, high probability events within 30 days" — that's a forward-looking bet, not a technicals-are-screaming-buy bet. Technicals at entry were actually unremarkable: RSI 58, MACD histogram already negative (-0.099), ADX under 20 signaling a weak trend, volume ratio 0.52 — well below average. None of that supports strong momentum; the entry was priced more on anticipated catalyst than on the tape.
What actually happened over the hold: the regime flipped from bull (0.79 confidence) to neutral (0.50) by exit, and SPY's 20-day return went from +2.96% to -0.74% — the broader market turned during the four days TWLO was held. Deep stats confirm the move was tiny relative to the stock's own volatility: 0.09x of entry ATR, essentially noise. This wasn't a violent drawdown, it was a stock drifting sideways-to-down while the tape around it soured, landing right on a Friday NFP release date. Sector peers actually averaged +0.717% that week, so TWLO underperformed peers, but not disastrously — this reads as idiosyncratic underperformance in a friendly sector, compounded by macro chop. Headlines were net-positive in tone, so no negative-news smoking gun.
Lesson: when the Catalyst pillar carries the score (80) while Technical/Momentum lag (72.7/58.2) and volume is sub-average, treat it as a lower-conviction, event-anticipation trade — I should size or hold-duration-adjust for the fact that a regime flip during a short 4-day window can erase the entire expected edge, since the predicted return here (0.13%) was already inside the noise band implied by 0.09x ATR.
SRRK, entry 8/31, out 9/4 via scheduled Friday sell: -3.32% over 4 days. The model's own entry read was -0.381% expected return — essentially flat-to-slightly-negative, not a bullish call that blew up. So the frame here isn't "what did I miss," it's "I called this roughly right on direction and still ate a loss bigger than the number implied."
The composite score (66.0) selected the name anyway despite the negative predicted return, on the strength of an upcoming-catalyst flag and a Technical pillar reading 65.78 — the highest of the five populated pillars. Momentum (55.21) and Fundamental (45.69) were middling to weak, and Smart Money / ML were both null, so this was a selection built on catalyst timing and technical setup carrying a name the model didn't actually expect to make money. That's a thin basis for a "selected: true," and the outcome — a loss roughly 9x the predicted magnitude — is consistent with that thinness rather than a surprise reversal of thesis.
Context makes the miss-in-magnitude easier to explain than the miss-in-direction. The regime flipped bull→neutral during the hold (confidence dropped from 0.79 to 0.50), SPY's 20-day return flipped from +2.96% to -0.74%, and VIX ticked up 14.43→15.20 — a real deterioration in the tape over just four days. Worse, sector peers averaged +3.725% over the same window while SRRK fell — this was idiosyncratic underperformance against a rallying biotech group, not sector drag. Fundamentals were already ugly going in (operating margin -404.85%, ROE -159%, ROA -86.6%, no PE), so there was no cushion. No headlines, news events, or GDELT tone data populated — I have nothing in-file pointing to a specific catalyst or news trigger for the drop, and I won't invent one. The NFP print landed on exit day, which may have contributed to the macro wobble, but I can't isolate its effect from the broader regime flip.
Lesson: when predicted_return is already negative and the pillar mix leans entirely on Technical/Catalyst with Momentum and Fundamental both weak and Smart Money/ML null, treat "selected: true" as a red flag, not a green light — the model flagged the risk correctly at -0.381%, but selecting a name it expects to lose money on, against a strengthening peer group, is a policy question worth revisiting independent of this single result.