The published backtest without option and award filings (2026-10-22)
Public summary of audit 275. No strategy rule, threshold or selection logic changed.
What changed
Some filings recorded as share purchases were in fact stock options, performance shares, warrants or employee fund units. On 22 October we excluded them from every calculation. They had flattered the backtest: the shares behind them rose about 9% over 90 days on average, and a director receiving options is not a director buying shares. The same backtest was run again on production, read-only.
The numbers, before and after
2022-01 to 2026 Q1 decisions, EUR portfolio, total returns, 0.6% round-trip cost. Sharpe and drawdown on each stock's real daily prices (audit 273).
| Figure | This morning | Now |
|---|---|---|
| Annual return of the proof basket | 13.0% | 9.9% |
| Sharpe, with 95% interval | 0.79 (0.06 to 1.73) | 0.68 (-0.15 to 1.68) |
| Worst drawdown | -20.1% | -20.9% |
| STOXX Europe 600 on the same dates | 9.3% a year | 8.0% a year |
| Basket minus the index, points a year (95% interval) | +3.6 (-4.1 to +12.7) | +1.9 (-4.5 to +9.0) |
| Since October 2023 (30 months), return | 22.1% | 19.2% |
| 2022 alone | +1.0% | -6.1% |
| Ten rows served first, annual return | 5.1% | 5.1% |
How to read it
- The basket stays slightly ahead of the index on the same dates, but the 95% interval is wide and contains zero: no demonstrated edge.
- The basket does not cap the weight of one company: at times close to half of the capital sat in a single issuer after a burst of filings. A cap was tested and not adopted (it cost about 2 points a year).
- A backtest is not a live record and does not establish an investment advantage.