Personal portfolio
—
Self-reported · includes reconstructed periods
Personal investment record
Performance, in context.
A weekly record of the portfolio, its exposures and the decisions behind it. Self-reported and unaudited; parts of the history are reconstructed.
SPY
—
Total return · dividends reinvested
NSDQ
—
Nasdaq Composite · price return
Jan 6, 2025–Oct 16, 2026 · Rebased to 100. SPY is the primary total-return benchmark; NSDQ is price-return context. Definitions & provenance ↗
Performance
Fund vs S&P 500 · rebased to 100
Rebased to 100 · weekly Friday closes
| Since inceptionJan 6, 2025 – Oct 16, 2026 | Trailing 12MOct 10, 2025 – Oct 16, 2026 | YTDDec 26, 2025 – Oct 16, 2026 | |
|---|---|---|---|
| Total return | +158.7% | +64.5% | +40.9% |
| SPY (total return) | +30.7% | +18.1% | +11.4% |
| Excess vs SPY | +128.0% | +46.3% | +29.4% |
| Annualised return | +70.8% | +63.2% | — |
| Annualised volatility | 20.5% | 24.6% | 27.4% |
| Sharpe | 2.50 | 1.94 | 1.53 |
| Sortino | 4.31 | 3.30 | 2.58 |
| Max drawdown | -14.2% | -14.2% | -14.2% |
| Beta to SPY | 0.22 | 0.82 | 1.00 |
| Correlation | 0.17 | 0.40 | 0.45 |
| Up / down capture | 85% / -42% | 129% / 15% | 137% / 41% |
| Positive weeks | 71% of 93 | 66% of 53 | 57% of 42 |
| Best / worst week | +8.6% / -8.5% | +8.6% / -8.5% | +8.6% / -8.5% |
Drawdown from peak · weekly closes
Monthly returns · last weekly point of each month
| Year | Jan | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep | Oct | Nov | Dec | Year | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2025 | Fund | +4.0% | +4.5% | +7.7% | +5.1% | +17.5% | +7.0% | +3.1% | -10.3% | +4.9% | +8.2% | +8.6% | +4.2% | +83.7% |
| SPY | +1.1% | -1.3% | -6.2% | -0.9% | +7.0% | +4.6% | +3.6% | +1.2% | +2.9% | +3.1% | +0.2% | +1.3% | +17.3% | |
| 2026partial | Fund | +11.6% | -1.3% | -9.7% | +6.1% | +15.6% | +0.5% | +2.9% | +2.7% | +9.1% | -0.3% | +40.9% | ||
| SPY | +0.2% | -0.9% | -7.3% | +12.6% | +6.0% | -3.4% | +2.5% | +3.0% | +0.5% | -1.1% | +11.4% |
Holdings
Weights as of the last brokerage snapshot, as shares of total account value; cash is the remainder. Entry prices are brokerage average costs; prices marked ≈ are estimates until the next brokerage export, and a dash means the cost has not been recorded yet. Quotes load after the page.
| Ticker | Name | Dir | Entry | Last | Return | Weight | Conviction | Entered | Region |
|---|---|---|---|---|---|---|---|---|---|
| TLT | iShares 20+ Year Treasury Bond ETFFirst 38 shares bought by Sep 23 at $80.78; the bulk of the position was added in the first week of October near the lows (TLT closed $77.11 on Oct 5). Blended cost is an estimate until the next broker export. | Long | ≈$78.00 | — | — | 35.0% | High | Sep 2026 | United States |
| IEI | iShares 3–7 Year Treasury Bond ETFEntered in October 2026; the fund traded in a $113.2–113.6 range over the entry window. | Long | ≈$113.40 | — | — | 15.0% | High | Oct 2026 | United States |
| OSCR | Oscar HealthEntered in October 2026; average cost and position note to follow. | Long | — | — | — | 10.0% | Medium | Oct 2026 | United States |
| INTC | IntelSmall position entered in October 2026 after the stock fell from $127 to below $100; average cost to follow. | Long | — | — | — | 2.0% | Low | Oct 2026 | United States |
| AMAT | Applied MaterialsReduced to a 2% position in early October; the 10-share position was closed at about $540 (see the closed trade). Average cost to follow. | Long | — | — | — | 2.0% | Low | Oct 2026 | United States |
| 9992.HK | Pop MartResidual four-share position (HK$236.61 average cost) left after the 2025 trade; not part of the current sizing. | Long | ≈HK$236.61 | — | — | 0.2% | Low | Jun 2025 | China |
Exposure
As of Oct 16, 2026Book
Long64.2%
Short0.0%
Cash35.8%
Gross 64.2% · Net 64.2% · 6 positions · Top 5 = 64.0%
By region
United States64.0%
China0.2%
By sector
Treasuries50.0%
Managed care10.0%
Semis2.0%
Semicap2.0%
Consumer0.2%
Selected post-mortems
Three trades that did not work and three that did, with the lesson each one paid for. Returns are price returns on the instrument, not portfolio contributions.
What did not work
Oil recovered as the Hormuz premium deflated more slowly than the puts' time decay. The defined-risk structure capped the loss versus SCO. Closed to free capital for PBR and MTZ.
Wrong instrument. A daily-resetting 2x leveraged ETF loses to volatility decay regardless of direction; the Hormuz shock accelerated it. Replaced with defined-risk USO puts. Never hold leveraged decay products beyond intraday.
Sold on near-term fee compression and missed the depth of the private-credit tailwind: institutional capital was early in a multi-year rotation into alternatives that compounded AUM and fee-related earnings far beyond the model. The same insight now underpins the Apollo position.
What worked
Steel-cycle-bottom thesis: supply rationalisation, infrastructure supercycle and China demand recovery. Staged exits as the thesis confirmed; final exit at $70. A contrarian macro call from first principles.
IP flywheel and global collectibles expansion recognised before institutional money treated it as a real growth story. A separate, small re-entry later in 2025 at HK$236 is underwater and below the display threshold.
Built from channel work during tobacco coverage at Nomura: BAT, Smoore's largest customer, showed declining online sales and EU vape-store traffic was deteriorating months before either appeared in financials. Shorted at HK$20, covered at HK$10. The edge was reading primary data before the numbers caught up.
Closed trades
Every closed position since inception with the lesson attached. Returns are price returns on the instrument: (exit − entry) / entry for longs, (entry − exit) / entry for shorts. Prices marked ≈ are approximate exits reconciled to daily closes rather than statement fills.
Closed trades
33
Since Jan 2025
Win rate
67%
22 wins · 10 losses · 1 flat
Median trade
+12.3%
Price return on instrument
| Ticker | Name | Dir | Entry | Exit | Return | Held | Lesson |
|---|---|---|---|---|---|---|---|
| ONON | On Holding | Long | ≈$28.76 | ≈$35.00 | +21.7% | Aug 2026 – Oct 2026 | Held from late August at $28.76 through the September low of $26.76 and sold in mid-October at about $35, +22%. The one winner of the period with no semiconductor beta; the position note was never written. |
| LITE | Lumentum | Long | ≈$850.23 | ≈$1,100.00 | +29.4% | Sep 2026 – Oct 2026 | AI optical: bought in mid-September at $850 and sold at about $1,100 after the stock ran 30% in three weeks. By Oct 16 it was back below the entry price; the quarter's largest contributor was also its shortest hold. |
| FLEX | Flex | Long | ≈$109.92 | ≈$120.00 | +9.2% | Sep 2026 – Oct 2026 | Small AI-infrastructure position bought at $110 in mid-September and sold at about $120 near the Oct 6 high; the stock lost 25% in the following ten days. |
| NBIS | Nebius Group | Long | ≈$214.00 | ≈$250.00 | +16.8% | Sep 2026 – Oct 2026 | AI neocloud: bought in mid-September and sold near the Oct 6 high at about $250, +17%; an earlier partial sale in September was taken at a small loss. By Oct 16 the stock was 34% below the exit. |
| MU | Micron Technology (second position) | Long | ≈$957.63 | ≈$1,075.00 | +12.3% | Apr 2026 – Oct 2026 | Re-entered Apr 3, 2026 at $366.79, built to a third of the book and trimmed through the summer; the last 11 shares were sold in the first week of October at about $1,075, +12% on the broker's diluted cost. SOXX fell 16% from its Oct 5 high over the next two weeks. |
| AMAT | Applied Materials (second position) | Long | ≈$439.71 | ≈$540.00 | +22.8% | Aug 2026 – Oct 2026 | Entered Aug 25 at $459 and added in mid-September near $415; the 10-share position was sold in the first days of October at about $540, +23% on the broker's diluted cost. The stock fell 25% from there in two weeks; a 2% position was kept. |
| TEL | TE Connectivity | Long | ≈$202.57 | ≈$219.00 | +8.1% | Sep 2026 – Oct 2026 | Data-center connectivity, entered in mid-September at $203 and sold at about $219 in early October, +8%. It held up better than the semiconductors in the sell-off (still $212 on Oct 16); sold with the rest of the AI book rather than on its own merits. |
| LRCX | Lam Research | Long | ≈$283.69 | ≈$346.00 | +22.0% | Aug 2026 – Oct 2026 | Entered Aug 25 at $323 and nearly doubled the position at $270 in mid-September; sold at about $346 in early October. The Oct 2 close was the high; the stock lost 24% over the following two weeks. |
| MTZ | MasTec (second position) | Long | ≈$243.67 | ≈$210.00 | -13.8% | Sep 2026 – Sep 2026 | Re-entered Sep 9 at $247, the second purchase after Issue #7 had screened the name as avoid; trimmed the next day, added back, and closed at about $210 in late September for −14%. Two losses in the same name against the research's own rating. |
| AMZN | Amazon.com | Long | ≈$257.74 | ≈$250.00 | -3.0% | Sep 2026 – Sep 2026 | Entered after the Sep 2 focus-list work without a written thesis; sold in two legs (Sep 10 and late September) at about $250 for −3%. No thesis, no kill criterion, no reason to hold through a drift. |
| GS | Goldman Sachs | Long | ≈$1,037.20 | ≈$933.00 | -10.0% | Sep 2026 – Sep 2026 | A one-share position entered in early September at $1,037 with no written thesis; sold in mid-September at about $933 after the stock fell 10% in two weeks. Small in dollars, same lesson as AMZN: nothing in the book without a thesis. |
| AGRO | Adecoagro | Long | ≈$10.50 | ≈$12.50 | +19.0% | Sep 2026 – Sep 2026 | Sized before Issue #8 and sold two days after publication at about $12.50 on the post-note spike. Right vehicle for the El Niño thesis, but the hold was two weeks against a thesis whose horizon is the 2026-27 event peak; AGRO traded back to $10 by October. |
| AAOI | Applied Optoelectronics | Long | ≈$108.42 | ≈$111.00 | +2.4% | Sep 2026 – Sep 2026 | AI optical-interconnect position held for about a week and sold into the Sep 9 morning strength at about $111, a small gain. The stock was below $100 the following week and 30% lower by mid-October. |
| TCOM | Trip.com Group | Long | ≈$48.53 | ≈$41.00 | -15.5% | May 2026 – Sep 2026 | China outbound-travel position bought at the May 22 low near $46 and added above it (average cost $48.53); sold in early September at about $41, −15%, when the book was reshuffled into the September positions. The one non-technology equity of the summer, and the full write-up was never done. |
| EWJV | iShares MSCI Japan Value ETF | Long | ≈$42.24 | ≈$47.00 | +11.3% | Mar 2026 – Aug 2026 | Japan value basket held as low-beta ballast; sold at about $47 in early August (+11%) to concentrate the book. Did the job it was sized for. |
| LKNCY | Luckin Coffee (OTC) | Long | ≈$35.00 | ≈$35.00 | 0.0% | Dec 2025 – Aug 2026 | Average cost fell to about $35 after adding below the $37.69 pitch price. The stock spent the summer under $33 and the relisting catalyst produced no timeline, so it was sold on the August rebound to $35 to fund the memory position. Flat outcome: the thesis was slow rather than wrong, and capital has a cost. |
| MSFT | Microsoft | Long | ≈$371.99 | ≈$451.10 | +21.3% | Mar 2026 – Jul 2026 | Entered near the March correction lows on the 'AI is an accelerant, not a cost centre' thesis. The fiscal Q4 print on Jul 29 sent the stock up 15% in a day; sold into it at about $451 (+21%). Right thesis, taken at the event rather than held for the re-rating that followed: the stock traded above $500 two weeks later. |
| MTZ | MasTec | Long | ≈$318.34 | ≈$300.00 | -5.8% | Mar 2026 – Jul 2026 | Process failure. Issue #7 (Apr 24) screened MTZ as 'avoid' at 44x while the position was still held. The stock ran to $437 in May and gave it all back; sold at about $300 in late July for −6% against cost and roughly −30% from the peak. When your own research rates a holding 'avoid', act on it the same week. |
| CAT | Caterpillar | Long | ≈$715.64 | ≈$1,064.90 | +48.8% | Mar 2026 – Jun 2026 | Energy & Transportation as a growth segment: the data-center power thesis played out faster than modelled. Sold at the Jun 30 high above $1,000 (+49%). The stock has since retraced to about $810, so the exit captured most of the move. |
| LTM | LATAM Airlines Group ADR | Long | ≈$48.86 | ≈$57.00 | +16.7% | Mar 2026 – Jun 2026 | The oil-beneficiary long: bought after the sector contagion sell-off, sold on the late-June rebound at about $57 (+17%) as the fuel-cost tailwind was priced in. Below the $62–65 target, above the $40 stop; a trade that worked as designed. |
| BABA | Alibaba Group ADR | Long | ≈$123.51 | ≈$120.00 | -2.8% | Mar 2026 – Jun 2026 | Exited at about $120 on the June financing announcement despite the cloud re-acceleration thesis staying intact. Small loss, and the stock fell another 20% to $95 within weeks. The business stays on the watchlist; the lesson is that dilution changes the per-share case even when the operating case is right. |
| APO | Apollo Global Management | Long | ≈$111.22 | ≈$130.00 | +16.9% | Mar 2026 – May 2026 | Bought at the April litigation-headline lows, sold on the rebound to about $130 (+17%). Captured the mean reversion, not the multi-year private-credit compounding the thesis described; the stock reached $144 in August. |
| PBR | Petróleo Brasileiro (Petrobras) ADR | Long | ≈$20.70 | ≈$19.00 | -8.2% | Mar 2026 – May 2026 | The re-rating case needed the political discount to unwind while oil held; instead oil fell and the discount widened. Cut at about $19 in late May (−8%). The stock went on to $16 in July, so the exit was disciplined; the error was sizing a political-discount thesis at 17% of the book. |
| FCX | Freeport-McMoRan | Long | ≈$57.71 | ≈$60.00 | +4.0% | Mar 2026 – May 2026 | The cross-market copper thesis did not get its supply catalyst inside the holding window; took a small gain at about $60 in May. The stock has since run to about $80 on the copper move. The thesis was right and the patience was not. |
| USO | USO Jul-17 puts ($100 / $90 strikes) | Short oil | $6.50 | $4.20 | -35.4% | Mar 2026 – Apr 2026 | Oil recovered as the Hormuz premium deflated more slowly than the puts' time decay. The defined-risk structure capped the loss versus SCO. Closed to free capital for PBR and MTZ. |
| SCO | ProShares UltraShort Crude (2x inverse) | Short oil | $10.40 | $7.80 | -25.0% | Dec 2025 – Mar 2026 | Wrong instrument. A daily-resetting 2x leveraged ETF loses to volatility decay regardless of direction; the Hormuz shock accelerated it. Replaced with defined-risk USO puts. Never hold leveraged decay products beyond intraday. |
| 6969.HK | Smoore International | Short | HK$20.00 | HK$10.00 | +50.0% | Jan 2025 – Jan 2026 | Built from channel work during tobacco coverage at Nomura: BAT, Smoore's largest customer, showed declining online sales and EU vape-store traffic was deteriorating months before either appeared in financials. Shorted at HK$20, covered at HK$10. The edge was reading primary data before the numbers caught up. |
| MT | ArcelorMittal | Long | $30.00 | $70.00 | +133.3% | Jan 2025 – Jan 2026 | Steel-cycle-bottom thesis: supply rationalisation, infrastructure supercycle and China demand recovery. Staged exits as the thesis confirmed; final exit at $70. A contrarian macro call from first principles. |
| GEV | GE Vernova | Long | $590.00 | $750.00 | +27.1% | Jan 2025 – Sep 2025 | Connected the AI power buildout to grid-equipment demand months before consensus modelled it. Thesis played out ahead of schedule; exited early relative to the later run. |
| JAPAF | Japan Tobacco | Long | $100.00 | $115.00 | +15.0% | Jan 2025 – Sep 2025 | Heated-not-burn volumes were accelerating in the data while the street stayed anchored to secular decline. Defensive yield plus a structural volume shift. |
| MU | Micron Technology (first position) | Long | $110.00 | $155.00 | +40.9% | Jan 2025 – Aug 2025 | Right on direction, wrong on magnitude. Exited at $155 on a $250 'ceiling'; the stock ran past $362. HBM is bespoke, high-margin memory that scales with each GPU generation, not commodity DRAM. Re-entered April 2026 at $367. |
| BX | Blackstone | Long | $155.00 | $128.00 | -17.4% | Jan 2025 – Jun 2025 | Sold on near-term fee compression and missed the depth of the private-credit tailwind: institutional capital was early in a multi-year rotation into alternatives that compounded AUM and fee-related earnings far beyond the model. The same insight now underpins the Apollo position. |
| 9992.HK | Pop Mart International | Long | HK$70.00 | HK$140.00 | +100.0% | Jan 2025 – Jun 2025 | IP flywheel and global collectibles expansion recognised before institutional money treated it as a real growth story. A separate, small re-entry later in 2025 at HK$236 is underwater and below the display threshold. |
Reflection — the losses
Six losses, four causes. Blackstone was a thesis error: selling on near-term fee compression while institutional capital was early in a multi-year rotation into alternatives. SCO was an instrument error: a daily-resetting leveraged ETF loses to volatility decay independent of the oil call. The USO puts were a path error: the structure was right, but the Hormuz premium deflated more slowly than time decay. The 2026 losses were sizing and process errors: Petrobras sized a political-discount thesis at 17% of the book, MasTec stayed in the book for three months after the fund's own research rated it 'avoid', and Alibaba was sold on a financing at a small loss with the operating thesis intact. The lessons: match the instrument to the horizon, size political theses as options rather than core positions, and act on your own screens the week they are published.
Reflection — the wins
The common thread is that each thesis was built from primary data and first principles before the financials confirmed it. ArcelorMittal was a steel-supercycle call entered at $30 and exited at $70 for +133%; Smoore was sell-side channel work applied to a short, +50%; Pop Mart, Japan Tobacco and GE Vernova were all visible in the data before the street modelled them. The 2026 closes add Caterpillar (+49%, sold at the June high), Microsoft (+21% into the July print) and the two mean-reversion trades, Apollo and LATAM (about +17% each). The pattern in the wins is also the pattern in the missed upside: Micron, Microsoft, Apollo and Freeport all ran well past the exit. Being right on direction and early on the exit is the recurring cost of taking profits at the event.
Performance evidence
Download weekly data (CSV) ↓How each period's weekly points were obtained. Periods marked reconstructed will be re-marked as statements are reconciled; the full weekly series is in the CSV.
| Period | Basis | How the weekly points were obtained |
|---|---|---|
| Jan 6, 2025 – Mar 20, 2026 | Reconstructed | Reconstructed from the brokerage weekly account-value series. Weekly points chain-linked from the brokerage chart; not yet reconciled to monthly statements |
| Mar 27, 2026 – Apr 24, 2026 | Statement-verified | Statement-verified. Weekly returns reconciled to brokerage statements |
| May 1, 2026 – Aug 21, 2026 | Reconstructed | Reconstructed from the year-to-date account-value chart. Digitised from the brokerage chart and chain-linked from the last verified point; calibration error about ±2%; five single-day moves (Jun 2, Jul 2, Jul 17, Jul 24, Aug 7) may include cash flows and are treated as returns until statements are reconciled |
| Aug 28, 2026 – Oct 16, 2026 | Reconstructed | Broker-reported weekly returns. Weekly time-weighted returns read from the brokerage app on Oct 17, 2026 and chain-linked from the Aug 21 point; not yet reconciled to monthly statements |
Definitions on the methodology page.