Tested against the market —
and against itself.
Beating the S&P is the easy comparison. The hard one — the one almost nobody publishes — is beating your own quality screen: does the ranking add anything beyond simply owning good businesses? VERIX ran both, kept the losses in the table, and left the caveats attached. It is a backtest, not a live record — read it as one.
A 6-leg annual study of the ranking against SPY and against its own graded universe. The second column is the one that matters: it isolates how much the ranking adds beyond simply owning the quality screen.
| Period | Top-K | Universe | SPY | vs SPY | vs Univ |
|---|---|---|---|---|---|
| 2019–20 | +29.7% | +16.3% | +7.3% | WIN | WIN |
| 2020–21 | +44.1% | +54.5% | +40.9% | WIN | LOSS |
| 2021–22 | -11.2% | -6.6% | -10.6% | LOSS | LOSS |
| 2022–23 | +35.8% | +38.1% | +19.4% | WIN | LOSS |
| 2023–24 | +36.7% | +28.8% | +24.5% | WIN | WIN |
| 2024–25 | +34.8% | +11.4% | +14.4% | WIN | WIN |
| Cumulative | 4.151× | 3.326× | 2.299× | 5/6 | 3/6 |
- This is a BACKTEST, not live performance, and not a promise of future results.
- SURVIVORSHIP BIAS, and it cannot be removed: the candidate universe is today's watchlist, which by construction excludes companies that died or delisted between 2019 and 2025. Every absolute number here -- including 'beat SPY' -- is flattered by that, and the size of the flattery is unknown.
- THEREFORE the number that means something is the SPREAD: top quintile versus the equal-weight same universe. Both carry the identical survivorship bias, so what is left when it cancels is grading skill. That spread won 3 legs of 6. The 5-of-6 against SPY is context, not evidence.
- The ranking is the Forge's CORE lens, not the full composite the app ships today. It tests the essence of the method, not the product.
- US SEC-XBRL filers only. Foreign listings have no US XBRL and are excluded entirely, so this does not describe the universe a subscriber now sees.
- Backtests carry design risk: the lens was authored with knowledge of this period, which no out-of-sample claim can undo.
- Returns are price-only and exclude trading costs, taxes and slippage.
- The graded universe itself beat SPY, so a large share of the result comes from the quality screen rather than from the ranking on top of it.
A backtest proves the idea held up in the past. The real proof is forward: VERIX logs every call at decision time with the price and the SPY baseline attached, and cannot edit them afterward. The first 90-day results land on 2026-09-23 — the same calls you can read today, wins and losses both. Simulated performance is not live performance.
Ledger as of 2026-08-26T20:01:52. VERIX is research software, not an investment adviser. Nothing here is a recommendation, a solicitation, or a promise of future results, and past or simulated performance does not predict returns.