Survivorship-free backtests over three periods. Each period shows only strategies with data for it. Growth of 1, log scale.
The live books on the Live page run the same frozen rules; these research tracks extend them decades back on survivorship-free universes (delisted names included), reported as the median of three rotation-grid phases.
Total × growth of 1 · CAGR annualised · Sharpe risk-adjusted return · Max DD deepest peak-to-trough · Martin return / ulcer (drawdown-adjusted) · Ulcer drawdown depth index · Dom top-1 / top-2 share of days the line ranked #1 / top-2 by cumulative growth among the lines shown (our domination coefficients). Total × is concentration-lottery noisy over long spans — read risk axes and domination first.
Pick any start year — the metrics table and the chart rebase to it ("growth of 1 invested on Jan 1 of that year"). Adds the wider metric set and the P4 research portfolio: equal-weight 25% × 4 of momAB_k8u, resAB, 7fh and resAB_sp, monthly rebalance, same 15 bps costs — a research track (in live incubation since 2026-07-22, not on the Live page).
Extra columns: Vol annualised volatility · Sortino downside Sharpe · Calmar CAGR / |MaxDD| · R² of log-equity (path smoothness, 1 = straight line) · K Kestner ratio. Click any header to sort. In the 1999 view 6f starts in 2001 (walk-forward warm-up) and the S&P-side lines are absent (no survivorship-free data).
Rank 1..7 among the strategies (grey reference lines excluded) for every start year above. Dom1 structurally punishes the P4 blend — an equal-weight portfolio is never the single leader by construction; read it for solo lines.
Share of profitable rolling 252-day windows, the 5th percentile and the worst window — "how often a year is profitable, and how bad the bad ones get". The index benchmarks are profitable more often; the strategies win by magnitude, not frequency.