Methodology
This portfolio exists to make our published theses falsifiable in public. Every position states what would prove it wrong, and closes when that happens. This page documents every design choice, assumption, and cost that goes into the published returns.
Structure
Concentrated enterprise software long/short. Seven long positions at 4,000 bps total exposure, five short positions at 4,000 bps. Gross exposure of 8,000 bps against 100% equity. The entire book is deliberate bets on enterprise software — every position is a published thesis, no index filler, no unrelated sectors. Cash funds the short side.
Universe
The portfolio is restricted to enterprise software — CRM, workflow automation, creative tools, cybersecurity, HR/finance platforms, edge infrastructure, observability, data platforms, DevOps, RPA, and AI/analytics platforms. Consumer internet, semiconductors, e-commerce, and hardware are excluded. This is a deliberate constraint: a narrow book with deep thesis work beats a broad book with shallow conviction.
Implementation
Pure active book
This is not an overlay on an index. Every basis point of exposure is an active bet on an enterprise software company. When the book holds long positions, it expects those companies to outperform. When it holds shorts, it expects those companies to underperform. Cash collateralizes the short book. There is no benchmark-based attribution — returns are absolute, not benchmark-relative.
Symmetric conviction weight
Sizing reflects conviction, not market cap or index weight. High conviction longs receive ~643 bps, medium conviction ~500 bps. Shorts are concentrated at 800 bps per name because strong short ideas are scarcer, and sizing should reflect that scarcity rather than dilute conviction into positions that are too small to matter.
Shorts are gated, not modelled
The short universe is restricted to names the broker reports as shortable and easy-to-borrow. Hard-to-borrow names are out of scope. A conservative flat accrual is applied to all short positions. This is a deliberate constraint: the crowded, expensive-to-borrow names where contrarian short alpha concentrates are excluded by construction. We accept that loss in exchange for cost honesty.
Timing Rule
A position opens at the next regular-session open following publication. Never before. The position is disclosed in the article that motivates it, not in a separate ledger. Open is resolved against the market calendar — holidays and half-days are respected.
Cost Realism
Gross and net-of-cost returns are published side by side. Net leads. One synthetic cost is accrued daily in our own ledger:
Borrow costs on shorts
The short universe is restricted to names the broker reports as shortable and easy-to-borrow. Hard-to-borrow names are out of scope. A conservative flat accrual is applied to all short positions. This is a deliberate constraint: the crowded, expensive-to-borrow names where contrarian short alpha concentrates are excluded by construction. We accept that loss in exchange for cost honesty.
A paper account does not charge borrow fees. They are real at 4,000 bps of short exposure. We accrue them anyway because omitting them would flatter the returns.
Unwind Conditions
Every position closes under one of four conditions. Every close publishes which one fired. A track record that only reports P&L is a stock-picking bot; one that reports why each position ended is auditable.
1. Falsified
The claim was checked and is wrong. The position closes, and we say so in print. This is the most important outcome — it proves the system is working.
2. Realized
The claim was checked and came true. The position closes and the thesis is booked.
3. Expired
The horizon elapsed with the claim neither confirmed nor refuted. The position closes. An unresolvable thesis is a reporting failure and reads as one.
4. Risk Stop
Mechanical, position-level. Protects the book from a thesis that is right eventually and ruinous now.
Factor Concentration Cap
A concentrated enterprise software book risks becoming a single factor bet — every position resolves on the same question. We enforce a stated cap on exposure to any one theme per side. The cap is published alongside the current reading. Readers will spot the concentration before we do, and disclosing it is cheaper than being caught by it.
Cap per side: 50% to any single theme. Current reading is published on the positions page.
Rebalance Rule
The book drifts as the market moves. Rebalancing occurs monthly, or when gross exposure deviates more than ten percent from target, whichever comes first. Rebalance brings gross back to 8,000 bps and re-sizes positions to their intended active weights.
Data Firewall
All portfolio decisions operate against point-in-time data only. Backfilled datasets (including split-adjusted research tables) are never joined into anything the portfolio treats as a decision or a resolution. The lookup path is enforced at the database role level.
Order Integrity
No order is placed that is not traceable to a published thesis with an article slug. This is an assertion, not a convention — it is enforced in code at order-construction time. On every run, broker positions are reconciled against the intended book and any divergence is alerted rather than silently corrected.
Thesis Schema
The portfolio engine consumes only structured theses. It never reads article prose. Every thesis carries:
- Ticker and direction (long / short)
- Sizing in basis points of the overlay
- Conviction rating (high / medium / low)
- Horizon — the day count after which the thesis expires if nothing else fires
- Thesis — one sentence, plain language
- Falsifiable claim — a specific, checkable assertion. Must be verifiable against a named source or a reported number, never a valuation multiple. The multiple is the entry; the reported fundamental is the falsification.
- Unwind conditions — at least one of the four kinds listed above, always non-empty
- Article slug — the piece that motivated the position
Disclosure
This portfolio is simulated and is not investment advice. Gross Retention makes no recommendation. The portfolio exists to make our published claims falsifiable in public. All positions are tracked on a live paper-trading account at Alpaca. Past simulated performance does not guarantee future results. Nothing on this page or elsewhere on this site constitutes a solicitation to buy or sell any security.