1 · Valuation model (P/IV)
Does buying cheap-vs-our-model and avoiding expensive actually work? Approach B (Forward Test) is the real one — it uses point-in-time monthly snapshots (no lookahead). The controls: Mode = long-only vs long/short; K per sector = how many names to hold per sector; Cost/Borrow bps = trading frictions; Years = lookback; Min confidence = data-quality filter. Approach A is a quick sanity check with lookahead bias — ignore it for conclusions.
🔒 Backtest Lab
This page is private. Enter the access code.
2 · Risk signals — do they pay off? (long / short, by sector & theme)
This tests the market-risk system itself: for each signal (overall risk, the tide, conditions, valuation, smart-money + its 3 sub-inputs) it simulates acting on it — long when the signal says low-risk, short/cash when high — on the S&P, every sector, and every theme. It's newer, so it only has a couple of weeks of recorded signal history and fills in daily.
3 · Insider signals — the live event study
Every CMP-filtered opportunistic insider buy our pipeline records becomes a backtest observation (forward returns vs the S&P 500). The live table — with cluster-buy and low-density breakdowns — is published on the Insider Signals page alongside the feed itself, so the signals and their track record live side by side.