Public forward test in progress — a new list after every US market close. Research, not advice.

Quantitative Signal

Which stocks will swing
unusually wide next session?

Every evening the Radar lists the 10 of 97 US stocks most likely to have a trading range (high to low) more than 1.5× their own normal daily range on the next session. It is a watch-list, not a buy or sell signal: it says nothing about whether the price will go up or down. Every list is published before the market opens and never edited, and the results are scored in public against free alternatives.

Next session's watch-list How it is doing Get notified
—
Trading days scored in the public forward test (verdict after 40)
—
Share of the daily top 10 that swung unusually wide in the live test (shown after 5 days)
0.68
Before the test: ranking score (AUROC) in a one-shot check on May–Sep 2026. 0.50 = random order; the best free rule scored 0.63. This is not a win rate.

Latest watch-list

Loading…

Full list, rules and record

How it works

  1. Public data only. 5-minute and daily prices for 97 US-listed stocks, plus each company's published earnings dates. No news, no non-public information.
  2. One frozen model. The model was fixed before the test began, and its file is checksummed. The rules for passing or failing were written down and published before the first list.
  3. Published before the open. Each list is committed to a public git history before the next session starts. Late lists are not allowed, and old lists are never edited.
  4. Scored against free alternatives. Each day the list is compared with three free rules (earnings calendar, "already unusual today", and the ATR volatility indicator). If the Radar cannot beat them, it has no value, and the record will say so.

This is research, not advice. Non-personalized: everyone sees the same numbers. You make your own decisions.

Earlier research (archived)

This project started with a "+3% within 2 days" buy signal. Measured honestly, it did not make money after costs. Its skill at spotting big moves turned out to be no better than a free indicator, and it could not tell up from down. That record stays public, unchanged, because negative results are part of the evidence.

Loading…

Old signal record Scenario models Evidence

Get notified

Leave your email to hear when the Radar's 40-day verdict is out, pass or fail, and what happens next.

The form opens your email app with a pre-filled message. We only use it for Radar announcements.

Support this research

This site is free. If the research helps you, you may support it — a coffee is enough. Supporters receive nothing in return but gratitude.

Tip via Ko-fi ☕

Tips are voluntary gifts via Ko-fi. Supporters receive nothing in return but gratitude — this site stays free either way.