Both registered flu claims, computed fresh from CDC surveillance data using the
paper’s own published code. It detects; it does not forecast. The detector is armed and public
before the season begins — every reading below is timestamped.
+8.2 wkaverage head start over the standard rule (27 seasons)
26/27seasons where it fired first
8 wkhead start last season (2025-26)
The record: every season — green fires first, orange follows weeks later
Chart: weekly flu-like illness with both detectors' onset dates marked each season. (enable JavaScript for the interactive chart)
Claim 2 — the strain call, season by season
The first strain whose own divergence crosses its trigger is the call; the season’s true dominant strain grades it. The paper’s frozen record: 19/27 correct (70%) — and only those count toward the registered score. The losses split two ways, straight from the frozen per-season data: ◐ overtaken late = the call named the strain genuinely leading at the time, and a late second wave — usually influenza B — flipped the final tally (every such season was a near coin-flip; one “winner” took just 51% of positives). ✗ missed = a true signal error: it fired on a strain that was not even leading. No clear single-strain season has ever been missed — 9/9.
Season
First to fire
When
Dominant
Verdict
2026-27
watching — the call posts here the week the first strain fires
?
come back
2024-25
A(H1N1)
5 wk after onset
A(H1N1)
✓ correct
2023-24
A(H1N1)
1 wk before onset
A(H1N1)
✓ correct
2022-23
A(H3N2)
at season onset
A(H3N2)
✓ correct
2021-22
A(H3N2)
1 wk after onset
A(H3N2)
✓ correct
2019-20
B
5 wk before onset
A(H1N1)
◐ overtaken late
2018-19
A(H1N1)
3 wk before onset
A(H1N1)
✓ correct
2017-18
A(H3N2)
at season onset
A(H3N2)
✓ correct
2016-17
A(H3N2)
at season onset
A(H3N2)
✓ correct
2015-16
A(H3N2)
1 wk after onset
A(H1N1)
◐ overtaken late
2014-15
A(H3N2)
2 wk before onset
A(H3N2)
✓ correct
2013-14
A(H1N1)
at season onset
A(H1N1)
✓ correct
2012-13
B
5 wk after onset
A(H3N2)
◐ overtaken late
2011-12
A(H3N2)
9 wk before onset
A(H3N2)
✓ correct
2010-11
A(H3N2)
2 wk after onset
A(H3N2)
✓ correct
2009-10
A(H1N1)
at season onset
A(H1N1)
✓ correct
2008-09
A(H1N1)
2 wk before onset
A(H1N1)
✓ correct
2007-08
A(H1N1)
8 wk before onset
B
◐ overtaken late
2006-07
A(H1N1)
1 wk after onset
B
✗ missed
2005-06
A(H3N2)
6 wk before onset
A(H3N2)
✓ correct
2004-05
A(H3N2)
2 wk after onset
B
◐ overtaken late
2003-04
A(H3N2)
at season onset
A(H3N2)
✓ correct
2002-03
B
9 wk before onset
B
✓ correct
2001-02
A(H3N2)
3 wk before onset
A(H3N2)
✓ correct
2000-01
A(H1N1)
4 wk before onset
B
◐ overtaken late
1999-00
A(H3N2)
at season onset
A(H3N2)
✓ correct
1998-99
A(H3N2)
10 wk before onset
B
✗ missed
1997-98
A(H3N2)
at season onset
A(H3N2)
✓ correct
Live claim-2 status: awaiting laboratory data.
The 2026-27 onset call will appear here the week it fires — subscribers hear it the same day the board updates. The signup form is at the bottom of this page. Nothing here is medical guidance.
source: CDC FluView (ILINet / WHO-NREVSS) · method: the paper · season window: MMWR wk40–wk20 · generated 2026-09-24 15:00 UTC
Flu season — onset detection ahead of the standard method
Registered inside “Cross-Domain Validation,” which carries the framework beyond finance into natural and public-health systems.
The cross-domain validation paper carries two registered claims about influenza surveillance.
First: the divergence-based onset detector — the same fast-minus-slow construction the framework uses everywhere — fires at least six weeks ahead of the standard level-based detector for flu-season onset.
Second: the flu strain that fires first in the detector is the strain that goes on to dominate that season.
Both claims resolve on ordinary public health surveillance data as flu seasons arrive, and both are checkable by anyone who downloads the same public data the paper uses.
This page reports a registered research prediction. It is education, not advice: nothing here is a forecast service, an investment signal, or medical guidance.
Follow the boards
Board updates travel through the newsletter first — stage changes, season calls, new live
boards, and every prediction grading. One essay a week, free, unsubscribe anytime.
If the confirmation email lands in spam, moving it to your inbox tells your provider to trust the rest.