The annual walk-forward out-of-sample window is 2013-2015. The system is shown against two universe benchmarks: equal-weight and cap-weighted. The system earns a positive risk-adjusted return (Sharpe 0.751, +15.36%) at roughly half the drawdown of the benchmarks; the gain is concentrated in 2013, with flat years in 2014–2015.
All four comparisons use the same 608-name Nasdaq Global Select + Global Market universe, constructed point-in-time. The PM books additionally apply a 30-day ADV > $1M screen at each rebalance; the passive benchmarks hold all 608 names.
| Benchmark | Construction | Purpose |
|---|---|---|
| Equal-weight | 1/N across all 608 names, daily rebalance | Average return of the universe; eliminates any cap-tilt |
| Cap-weighted | Market-cap weights, PIT from fundamentals data | Passive large-cap exposure for the same universe |
The strategy (Sharpe 0.751, +15.4%, MaxDD -7.8%) delivered positive risk-adjusted returns and lower drawdown than either benchmark, but trailed both on absolute return. The equal-weight and cap-weighted benchmarks ran full-net-long in a 2013-2015 bull-market window; the strategy includes a dollar-neutral sleeve and runs at lower gross exposure, which explains most of the return gap.
Returns are concentrated in 2013 (Sharpe 1.99, +17.4%); 2014 and 2015 were near-flat. The aggregate Sharpe reflects a single strong year followed by two low-conviction ones — the annual breakdown is the more informative read.
System vs. equal-weight and cap-weighted benchmarks, 2013-2015.
OOS 2013-2015
Training and selection are separated from the 2013-2015 OOS window. The page reports the locked annual walk-forward result rather than choosing the best ex-post aggregation.
Both equal-weight and cap-weighted universe benchmarks are shown. The system delivers lower absolute return but at roughly half the drawdown of either benchmark over this window.
Portfolio construction, allocation, and PnL accounting run through deterministic code. LLM output can enter only after validation gates admit a signal.
The Next.js showcase reads synced result files from public artifacts. All numbers shown trace directly to the locked result bundle.