Radar experiment
A watch-list, not a trade signal. The radar ranks the 97 stocks by how likely each is to move unusually much on the next session, meaning a price range bigger than 1.5× its own normal daily range (ATR14). It says nothing about up or down, and nothing about profit.
This is a public forward test. Every list is published here before the next session opens and is never edited afterwards (the git history is the proof). After 40 trading days the pre-registered pass/fail rule decides whether the radar is good enough to build on. The result will be shown here either way. Rules · Research behind it
Latest list
Loading…
Forward-test record
Loading…
Why this exists
The older "big move" score turned out to be no better than ATR, a free indicator: volatile stocks simply stay volatile (see the Track record). The radar instead targets moves that are unusual for that stock. ATR cannot predict those.
Before going public it was tested twice on data it had never seen:
- 2024 – April 2026 (walk-forward): within-day skill 0.66–0.68 (0.50 = coin flip). The best free rule, "earnings tomorrow first", scored 0.61–0.63.
- May – September 2026 (one-shot check, rules fixed in advance): 0.68 vs 0.63 for the earnings rule. About 25% of its daily top 10 moved unusually, versus 23% for the earnings rule and 11% for a random pick.
Much of the skill comes from the earnings calendar, which is free. The model adds a modest, consistent extra on top, mainly on days without earnings. Past test results do not guarantee the forward test will pass.