- The four features
- The same bucketed regime features the Jev Regime Screener product computes from real bars: volatility vs. its own 60-bar baseline, 10-bar trend persistence, ATR expansion vs. its 50-bar baseline, and price position vs. the 50-day SMA. Here you set them directly instead of pulling them from a real symbol.
- Deterministic baseline
- A small fixed if/else rule, written to mirror the same plain-English criteria Jev is given for each regime label. It is not ground truth — it's a transparent reference point so a disagreement means something specific: "the fixed rule says X, Jev's probabilistic read says Y," not "Jev is wrong."
- Jev's distribution
- One typed
choice question — five regime labels, one probability each — plus two noul (yes/no-as-probability) questions: is a regime change likely right now, and is a trend-following strategy compatible with this regime. Every dropdown change sends a fresh request.
- Why there's no "reveal ground truth" here
- Unlike the other three Labs plays, this isn't a scored corpus — there's no seeded label for an arbitrary feature combination. This is an evaluator/dev-mode tool for watching how one input perturbation moves a probability distribution, not a benchmark with an accuracy number.
- What's missing on purpose
- The live Jev Regime Screener also takes a real ticker symbol, computes real 1-day and 5-day returns, and runs a trade-compatibility gate that sizes a hypothetical position. All three are removed here — this Lab only asks "what regime does Jev see," nothing about a specific market or a trade.