Which forecast is most accurate?
Every forecast we track, scored the same way: what it said before the day began, against the station's official record. Pick how far ahead the forecast was made to see how skill fades with time, and where each source's edge is.
Avg miss is the mean absolute error: the typical distance between the forecast and what happened.
Bias is the average signed miss: positive means the forecast ran warm (or wet).
≤2° is how often the forecast landed within 2°F; for rain it is how often the rain/no-rain call was right.
vs NBM compares each forecast with the National Blend on exactly the same station-days, so sources with different coverage are compared fairly. vs persistence does the same against “same as the last observed day”, the baseline any forecast should beat.
Same-morning forecasts are issued by 12Z on the day; a morning low is already known by then, so lows start the day before.