History
30 days of forecast vs. what actually happened at the lake. Click a single day to see its details.
Last 30 days
Rule-based forecast (actually used)
ML-based forecast β logistic (experimental)
ML-based forecast β HistGradientBoosting (black-box model, experimental)
Actual measurements (session β₯ 1 h)
Thu Jun 25
Fri Jul 24
session (β₯ 1 h β₯ 11 kt)
marginal (β₯ 1 h β₯ 8 kt)
no wind
no data
thunderstorm (forecast rows: predicted Β· actual row: observed)
Jul 2, 2026
avg wind
gust
ignition (8 kt)
session (11 kt)
For Jul 2, 2026 Β·
as written at the time
GO
11 of 14 rules green.
π€ ML β logistic
experimental
MAYBE
GO 20% Β·
MAYBE 47% Β·
NO GO 33%
Strongest inputs β
(+) morning solar; (β) dew-point spread, FΓΆhn Ξ.
π€ ML β HGB Β· black box
experimental
NO GO
GO 4% Β·
MAYBE 43% Β·
NO GO 53%
Most distinctive inputs: morning solar, Innsbruck pressure, Bolzano pressure.
Experimental shadow models β they never drive the verdict.
How good are they? β
How the day was forecast at the time, before the rules were recalibrated β shown for comparison only.