🔬 Edge Lab HONEST
Does ANY signal actually beat the market's natural drift? Every candidate is graded against the drift base rate at 95% — and we publish the verdict whether it flatters us or not · --
Loading the edge scorecard…
Signal scorecard — return / hit vs the drift base rate
--
SIGNAL
n
SIGNAL vs DRIFT
EDGE
z
VERDICT
🧠 How to read this — and why it's the honest move
The bar is always "beat the market's natural drift," never "beat a coin flip." Stocks drift up over time, so a coin flip already "wins" >50%. The only thing that counts as edge is accuracy/return above the unconditional drift base rate, and only if it clears a one-sided 95% bound (z > 1.64). Anything short of that is noise dressed up as skill.
Right now, none of the free signals clear that bar — analyst revisions and price momentum are essentially priced in on liquid names, and the brain's own 5-day technicals are mostly drift (the honest pass-267 finding). That's not a bug; it's the efficient market. The point of this page is that we look rigorously and tell you the truth — most "AI trading" sites fake exactly the edge we refuse to claim. The day a signal genuinely clears the drift bar, this row turns green automatically.
Backtested signals (analyst revisions, momentum) are computed from real Yahoo daily history + Finnhub recommendation trends, refreshed by a scheduled job. The brain-TA and confluence rows are pulled live from the 24/7 worker. The strongest untested free signal is open-market insider BUY-clustering — it can't be backtested (no free history) so it's accruing a live forward test on the Edge Scorecard. A real, durable edge most likely needs different data (paid order/option flow), not more indicators.