Explore

The interesting relationships live off the diagonal.

Every cell is measured on the cases two models both answered, in our own ledger. Cells rest on different shared-case sets, so this is a map of relationships — not a league table.

public snapshot public-20260913140717-080dd4d9 · as of 2026-09-13 14:07 UTC

Measurementmeasurement
Horizonhorizon
Metricmetric
sea · temperature_2m · 10 models with measured shared cases · 1,474 cases over 30 days

Model matrix

each square is one pair of models on the cases they both answered — tap one to open it

paired absolute-error differencemean over shared cases of ( |f_row − y| − |f_col − y| ). Negative means the row model was closer, in °C.

row closercolumn closer– means no shared measured cases

ECMWF AIFS 0.25° single vs ECMWF IFS 0.25°

ECMWF AIFS 0.25° single

ECMWF · forecast · ecmwf_aifs025_single

ECMWF IFS 0.25°

ECMWF · forecast · ecmwf_ifs025

On the 1,414 cases both answered, ECMWF AIFS 0.25° single was closer by 0.08°C on average 0.01 over 29 day blocks).

ECMWF AIFS 0.25° single closer on 48% of shared cases · error correlation 0.71 · +24h · sea temperature_2m

Does the advantage hold over time?

Mean of |error of ECMWF AIFS 0.25° single| − |error of ECMWF IFS 0.25°| on each day. Below the centre line, ECMWF AIFS 0.25° single was closer.

+0.20.0−0.22026-08-14: −0.06°C over 28 shared cases2026-08-15: −0.06°C over 28 shared cases2026-08-16: −0.09°C over 28 shared cases2026-08-17: −0.07°C over 28 shared cases2026-08-18: −0.11°C over 28 shared cases2026-08-19: −0.11°C over 28 shared cases2026-08-20: −0.11°C over 28 shared cases2026-08-21: −0.14°C over 28 shared cases2026-08-22: +0.01°C over 28 shared cases2026-08-23: −0.15°C over 28 shared cases2026-08-24: −0.11°C over 28 shared cases2026-08-25: −0.07°C over 28 shared cases2026-08-26: +0.05°C over 28 shared cases2026-08-27: −0.07°C over 28 shared cases2026-08-28: −0.12°C over 28 shared cases2026-08-29: +0.05°C over 4 shared cases2026-08-30: +0.03°C over 4 shared cases2026-08-31: −0.03°C over 4 shared cases2026-09-01: −0.14°C over 56 shared cases2026-09-02: −0.11°C over 56 shared cases2026-09-03: −0.08°C over 48 shared cases2026-09-04: −0.17°C over 56 shared cases2026-09-05: −0.14°C over 138 shared cases2026-09-06: −0.13°C over 161 shared cases2026-09-07: −0.04°C over 133 shared cases2026-09-08: −0.03°C over 129 shared cases2026-09-09: −0.06°C over 62 shared cases2026-09-10: +0.01°C over 75 shared cases2026-09-11: −0.02°C over 68 shared cases2026-08-142026-09-11

1,414 shared cases over 29 days · values in °C · shaded band = ±1 block standard error (0.01)

The dashed trend rises across the window — the advantage is changing, not fixed.

Do their errors move together?

Horizontal: signed error of ECMWF AIFS 0.25° single. Vertical: signed error of ECMWF IFS 0.25°. Points along the rising diagonal mean the two miss in the same direction at the same time.

+3.3+3.3

1,414 shared cases (472 plotted, evenly sampled) · r = 0.71 · errors in °C

Correlation is association, not joint failure. Two strongly correlated models can both be accurate; two uncorrelated models can both be wrong.

Does it depend on how far ahead you ask?

The same paired difference at every horizon in the cohort. Left of centre, ECMWF AIFS 0.25° single was closer; right of centre, ECMWF IFS 0.25°.

+6h
−0.03 ±0.02 · n750
+24h
−0.08 ±0.01 · n1,414
+48h
−0.10 ±0.01 · n1,040
+72h
−0.07 ±0.02 · n1,180

Mean absolute error at each horizon: +6h 0.36/0.39 · +24h 0.33/0.42 · +48h 0.34/0.44 · +72h 0.41/0.48 (ECMWF AIFS 0.25° single / ECMWF IFS 0.25°).

each bar carries its own shared-case count · values in °C

One real run of cases

Site south-china-sea: what actually happened, and what each model said would happen.

28.727.325.8

observed ECMWF AIFS 0.25° single ECMWF IFS 0.25°

72 consecutive shared cases · 2026-09-08 14:00 → 2026-09-11 23:00 UTC · values in °C

This is one site's run. It shows what the aggregate looks like; it does not substantiate it.

Day by day, as numbers
2026-08-14
−0.06 · n28
2026-08-15
−0.06 · n28
2026-08-16
−0.09 · n28
2026-08-17
−0.07 · n28
2026-08-18
−0.11 · n28
2026-08-19
−0.11 · n28
2026-08-20
−0.11 · n28
2026-08-21
−0.14 · n28
2026-08-22
+0.01 · n28
2026-08-23
−0.15 · n28
2026-08-24
−0.11 · n28
2026-08-25
−0.07 · n28
2026-08-26
+0.05 · n28
2026-08-27
−0.07 · n28
2026-08-28
−0.12 · n28
2026-08-29
+0.05 · n4
2026-08-30
+0.03 · n4
2026-08-31
−0.03 · n4
2026-09-01
−0.14 · n56
2026-09-02
−0.11 · n56
2026-09-03
−0.08 · n48
2026-09-04
−0.17 · n56
2026-09-05
−0.14 · n138
2026-09-06
−0.13 · n161
2026-09-07
−0.04 · n133
2026-09-08
−0.03 · n129
2026-09-09
−0.06 · n62
2026-09-10
+0.01 · n75
2026-09-11
−0.02 · n68

Individual cases — wins and losses

gulf-of-mexico · 2026-09-06 21:00observed 30.10°C · ECMWF AIFS 0.25° single 29.70 · ECMWF IFS 0.25° 27.40
gulf-of-mexico · 2026-09-06 22:00observed 29.90°C · ECMWF AIFS 0.25° single 29.70 · ECMWF IFS 0.25° 27.40
arabian-sea · 2026-09-02 12:00observed 28.20°C · ECMWF AIFS 0.25° single 27.70 · ECMWF IFS 0.25° 25.50
southern-ocean · 2026-09-10 21:00observed -1.30°C · ECMWF AIFS 0.25° single -2.60 · ECMWF IFS 0.25° -1.60
southern-ocean · 2026-09-10 20:00observed -1.80°C · ECMWF AIFS 0.25° single -2.80 · ECMWF IFS 0.25° -1.90
bay-of-bengal · 2026-09-02 21:00observed 28.30°C · ECMWF AIFS 0.25° single 29.30 · ECMWF IFS 0.25° 28.20

These illustrate the aggregate; they do not substantiate it.

matrix-v3 · content e7e47751cba27646 · matched on task definition (domain + channel) · native unit · entity / site · target time · horizon bucket (±1h) · outcome vintage used · issue and revision policy: the receipt as stored, never re-issued

This selection is in the address bar — the link you copy opens exactly this comparison.

How to read this

  • Each cell rests on its own set of shared cases, so the matrix cannot be read as a global ranking.
  • Only cases both models answered are scored. Abstentions and non-answers are counted beside every model, because scoring answered cases alone favours selective models.
  • Uncertainty is a block standard error over shared days, not an assumption of independent cases.
  • A single illustrated case shows what the aggregate looks like; it does not substantiate it.

Coverage behind each model

ECMWF AIFS 0.25° singleecmwf_aifs025_single ecmwfanswered 1,463 · abstained 0 · no answer 3 · mae 0.34°C
ECMWF IFS 0.25°ecmwf_ifs025 ecmwfanswered 1,463 · abstained 0 · no answer 3 · mae 0.41°C
NOAA GFSgfs_seamless gfsanswered 1,463 · abstained 0 · no answer 3 · mae 0.55°C
DWD ICONicon_seamless iconanswered 1,463 · abstained 0 · no answer 3 · mae 0.44°C
JMA seamlessjma_seamless jmaanswered 1,463 · abstained 0 · no answer 3 · mae 0.52°C
Météo-France seamlessmeteofrance_seamless meteofranceanswered 1,463 · abstained 0 · no answer 3 · mae 0.54°C
Met Office seamlessukmo_seamless ukmoanswered 1,463 · abstained 0 · no answer 3 · mae 0.54°C
CMA GRAPEScma_grapes_global cmaanswered 773 · abstained 0 · no answer 0 · mae 0.54°C
ECCC GEMgem_seamless gemanswered 773 · abstained 0 · no answer 0 · mae 0.49°C
ECMWF IFS 0.25° ensembleecmwf_ifs025_ens ecmwfanswered 737 · abstained 0 · no answer 5 · mae 0.38°C

cohort · matrix-v3 · content 080dd4d94fa8c4a6 · window 2026-08-142026-09-13

Upstream evidence, alongside ours

A second body of forecasting evidence, read from a cached notflat.world capture. It is shown here for reference and coverage: it is not mixed into the matrix above, because a comparison is only meaningful when task, unit, target, horizon and outcome vintage genuinely match.

reading upstream…