Every day, we show how much the scanner's Top 10 actually went on to rise, ranked by return, and
verify the scanner's real track record transparently with a running hit rate. (Both hits and misses are published)
Checking data status
Couldn't load the hit-rate data
This may be a temporary network or server issue. Please try again in a moment.
📭
No hit-rate data yet
Once daily scans accumulate, stocks get tracked and their results show up here. (At least a few days of scans need to build up before the hit rate becomes meaningful.)
Representative Hit Rate
Fixed +15% threshold · repeat picks on the same ticker within an episode count once (deduped)
As of · sample
95% confidence interval
Source live tracking (full scan_pick table) Raw (non-deduped) and volatility-matched figures are in the detail section below
There is no evaluable sample for the representative hit rate
Picks whose evaluation window is still open or under review are not included in the hit-rate denominator.
Share of results within scan-time ranges · 20 days
Middle half
· 80% range
Past-case candle coverage check · Recent 12 weeks
Period
Body ref. 50%
Upper wick ref. 90%
Lower wick ref. 90%
20-day hold return
Buy at scan-day close → hold 20 trading days
· median
SPY, same period
Benchmark, same window
Detail — raw basis (M1) & other metrics
Hit rate (raw, not deduped)
Avg. max gain (peak, unrealized)
Tracking status
A "hit" means the high touched
within
of the pick date.
Max gain is the highest return reached versus the entry price over the tracking window (up to
) — it measures whether "the move happened" (not a single day's close). Updated daily.
If it rises after the 20-day window, it's marked "Late" (not counted in the hit rate).
SAMPLE & METHOD
Sample and evaluation criteria
Eligible classification
Evaluation window in progress
Corporate-action review hold
Unverified entry basis
Legacy-basis label
Missing evaluation label
Correction coverage — of evaluated picks,
on the new basis ·
still on the old basis ·
with an unverified entry basis
·
tracked picks stopped updating because the entry basis could not be found (they may drop out of the denominator) · evaluated from stored daily bars
Hit Rate by Setup — which signal actually works
Detail — volatility-matched basis (M3)
🎯 Volatility-Matched Lift — a number that filters out volatility to isolate real skill, shown under both labels
We show results under both the fixed +15% criterion and the
volatility-matched (MFE ≥ 6×ATR) criterion together. The fixed-% criterion favors
high-volatility stocks, so when the two numbers differ, the lower one is the more conservative
estimate. The quantile boundaries and base rate use the same formula as the
base-rate proof page.
Actual hit rate
Wilson 95% CI
Volatility-matched expected rate
If it were a random stock from the same ATR distribution
Lift
Based on of
total evaluable picks
( from universe_snapshot
+ recovered from price_history).
The remaining are early picks made before
universe-snapshot collection started, so ATR can't be matched for them
— this excluded group's hit rate is , so including it would make the number above
.
The CI lower bound clears the volatility-matched expected rate under both labels — the
first real evidence, out of everything this site has published so far, that the scanner works statistically.
It does not pass both labels.
It wins under the fixed +15% criterion but not under the volatility-matched one — this could just be
residual volatility selection, so we don't claim it "works."
It wins under the volatility-matched (MFE ≥ 6×ATR) criterion but not under the fixed +15% one.
Since it fails to clear the more lenient fixed-% bar, we don't claim it "works overall" — though it
does still hold up once the volatility illusion is removed.
It fails to beat a random pick from the same volatility band under either label.
The reason we publish this number without hiding it is the methodology above — not cherry-picking a
favorable label, but an honest result, is this site's trust asset.
Volatility-matched lift: the sample isn't large enough yet to split into ATR quantiles. It'll appear once more picks accumulate.
Full Record by Scan Date
Grouped by the date it was scanned, showing the full Top 10 picked that day exactly as-is.
Where the ranking above is "sorted by the winners," this is the honest, complete record —
hits and misses alike, nothing left out.A just-picked day starts at 0% before tallying, and fills in as days pass. Click a header to expand.
Ticker
Entry
Current
Return
Max Gain (High)
Elapsed
Verdict
HitLateMissPending
🔥 Currently-Tracked Stocks — Ranked by Current Return
Every stock the scanner has picked within the tracking window (up to 60 trading days), pooled
together regardless of date, ranked from biggest gainer down. A quick way to see
which stocks are doing well right now.
(Includes both winners and losers)
Ticker
Scan Date
Entry
Current
Return
Max Gain (High)
Elapsed
Verdict
under review
HitLateMissPending
Hall of Fame — ranked by peak gain during tracking (unrealized). Not realized returns; names that fell afterwards are included.
under review
Review the data correction history
Performance figures correction completed (2026-10-01) —
We identified a calculation path where the price basis could be inconsistent across
corporate actions (splits, mergers) and dividends. We corrected the calculation code on
2026-09-24, and on 2026-10-01 we recalculated everything on the new basis, including
finished historical records and the base-rate statistics. All aggregates below now use
the same basis. The largest change is average maximum gain: the old-basis published
figure of 52.0% is 22.0% on the new basis (most of that was already reflected by the
2026-09-24 code fix). Across this full correction the hit rate rose from 27.8% to 28.7%
and the average 20-day return from 0.6% to 1.1%, a move in our favour, while the
universe base rate stayed at 25.8%. On the other side, the 'Record Volume Held' signal
lost its significance after multiple-comparison correction (31.9% to 31.1%). The
correction runs in both directions: some tickers were inflated and others understated.
Records whose entry basis could not be verified, and 4 picks with corporate actions we
could not judge, were excluded from the aggregates; two of those four would flip from
hit to miss once corrected, so that exclusion works in our favour.
How Is the Hit Rate Calculated?
We store every day's scanned Top 10 picks and track the return and max gain over the following 20 trading days,
based on the entry candle's close. We record whether the target (+15%) was reached, and we publish the full
history as-is — not just the wins, including misses and delistings.
The denominator only counts picks whose 20-trading-day window has finished (a delisting is treated
as final). A pick still being tracked isn't counted yet even if it already hit the target — if we counted it,
"other picks from the same period that haven't risen yet" would drop out of the denominator, inflating the hit
rate. Before 2026-08-11 this rule was implemented incorrectly, which showed the hit rate about 9%p higher than
reality; we corrected it as soon as we found it.
Average max gain follows the same principle — a pick whose 20 trading days haven't elapsed
yet hasn't had enough time to reach its peak, which structurally drags the average down, so it's excluded. Before
2026-08-13 this rule didn't exist, which showed the average about 2.9%p lower than reality.
We also matched the comparison window for the volatility-matched criterion to 20 trading days.
The comparison baseline (the universe base rate) is calculated from the max gain over a 20-trading-day window,
but the picks side was using the max gain over the full tracking window (up to 60 trading days). That gave the
picks up to 3x more time, which inflated the lift — in fact, 42.6% of picks hit their peak
after the 20-trading-day mark. When we matched both windows on 2026-08-13, the volatility-matched lift flipped
from +12.8%p to −1.5%p. After this correction, the scanner's edge is unconfirmed
under both criteria, and we publish that as-is.
Returns are measured on an adjusted-price basis. They are calculated from a series that
reflects price adjustments for stock splits, reverse splits and dividends — this is not a
dividend-reinvested return (we do not assume you received and reinvested the cash), and it excludes
taxes and trading costs. Before 2026-09-24 our internal documentation described this number as a
"split-only price return," but the calculation had been reflecting dividend adjustments as well. A full
scan of the ~193,000-row universe showed 12.1% of rows are affected, so instead of
changing the calculation we corrected the name to match what we actually compute
(label basis v2).
The limitations are real, too — this is a close-price-based tally, so it can differ from your actual fill
price, and we don't show a percentage until the sample is large enough. See the
methodology for the scoring system, the Learn hub for how
each individual signal works, and the base-rate proof page for measured base rates by
signal. Past hit-rate history does not guarantee future performance.
Get the daily morning brief + data alerts —
the scanner and hit-rate data on this page are Korean UI, but our
Telegram @dawnbrief channel posts a
US-market-close briefing and base-rate/fear-greed data updates you can follow either way.