Misses
On the 27 highs graded so far, our forecast was off by 1.7° on average. 33% landed outside the ± range, where about 20% (1 in 5) is expected: more than that means the range is too narrow.
On the 28 lows graded so far, our forecast was off by 1.2° on average, and it ran 0.5° too warm overall. 7% landed outside the ± range, where about 20% (1 in 5) is expected.
Each high and low is graded on the forecast made just before it began, never on a later update. One big miss is mostly bad luck; a pattern across many is something to fix, and any fix is tested in the Lab Notebook before it changes the forecast.
Recent misses outside the range
| Day or night | We said | It was | Off by | National Weather Service | Model average | What was going on |
|---|---|---|---|---|---|---|
| Monday, November 2, the high | 61.5° ± 2° | 58.8° | 2.7° too warm | 63.0° | 61.0° | cloud forecast 99%, sunshine measured 16% of a clear day, wind 9 miles per hour, the four models 1.0° apart, dew point swung 2.3° |
| Saturday, October 31, the high | 62.5° ± 4° | 66.7° | 4.2° too cold | 65.0° | 62.6° | cloud forecast 98%, sunshine measured 44% of a clear day, wind 5 miles per hour, the four models 5.1° apart, dew point swung 8.5° |
| Friday, October 30, the low that night | 49.9° ± 2° | 52.3° | 2.4° too cold | 47.0° | 50.8° | cloud forecast 91%, wind 4 miles per hour, the four models 1.2° apart, dew point swung 3.1° |
| Thursday, October 29, the low that night | 42.7° ± 3° | 39.6° | 3.1° too warm | 42.0° | 44.1° | cloud forecast 14%, wind 0 miles per hour, the four models 2.2° apart, dew point swung 20.9°, 0.01 inches of rain |
| Sunday, October 25, the high | 73.0° ± 2° | 75.6° | 2.6° too cold | 73.0° | 70.8° | cloud forecast 56%, sunshine measured 94% of a clear day, wind 1 miles per hour, the four models 0.9° apart, dew point swung 14.9° |
| Saturday, October 24, the high | 71.1° ± 2° | 67.5° | 3.6° too warm | 70.0° | 68.7° | cloud forecast 57%, sunshine measured 84% of a clear day, wind 2 miles per hour, the four models 1.1° apart, dew point swung 14.9° |
| Friday, October 23, the high | 71.9° ± 2° | 69.1° | 2.8° too warm | 71.0° | 69.0° | cloud forecast 4%, sunshine measured 98% of a clear day, wind 1 miles per hour, the four models 0.6° apart, dew point swung 17.1°, 0.01 inches of rain |
| Thursday, October 22, the high | 72.0° ± 2° | 70.0° | 2.0° too warm | 71.0° | 69.2° | cloud forecast 8%, sunshine measured 97% of a clear day, wind 2 miles per hour, the four models 0.4° apart, dew point swung 17.1° |
| Monday, October 19, the high | 68.9° ± 2° | 71.2° | 2.3° too cold | 67.0° | 66.2° | cloud forecast 0%, sunshine measured 101% of a clear day, wind 1 miles per hour, the four models 1.0° apart, dew point swung 18.0° |
| Sunday, October 18, the high | 63.3° ± 2° | 68.4° | 5.1° too cold | 63.0° | 61.0° | cloud forecast 0%, sunshine measured 102% of a clear day, wind 3 miles per hour, the four models 1.0° apart, dew point swung 13.5° |
| Friday, October 16, the high | 56.6° ± 2° | 54.0° | 2.6° too warm | 57.0° | 56.4° | cloud forecast 90%, sunshine measured 17% of a clear day, wind 7 miles per hour, the four models 0.6° apart, dew point swung 3.1°, 0.02 inches of rain |
"Too cold" means it came in warmer than we said. "The four models apart" is how far apart the four weather models were (a typical day is about 2°). The sunshine is what the station's sunlight sensor measured, as a share of a cloudless day.
Case studies
- C-005: The October 29 cold front: the high was nailed (78.9 vs 79.3 F); the night low was missed by 2 to 3 F, not the 9 F first written (corrected November 2: the night's forecasts had been read from the next date)
- C-003: Light radiation fog over the field that no airport saw (night of October 19-20, 2026)
- C-004: Two freeze-watch mornings, two opposite misses: October 17 (we said 31.2 F, it was 36.0 F) and October 18 (we said 34.9 F, it was 31.5 F) [written as C-002 by the lab assistant on October 19; renumbered C-004]
- C-002: The October 18-19 Frost Advisory night: the dew-point brake against the physics (the low was 34.2 F; our forecast 35.6 F, live-updated 34.4 F; the physics 32.1 F)
- C-001: The October 16-17 Freeze Warning night: what we said, and how it went (the low was 36.0 F; our forecast 31.2 F, 4.8 F too cold; corrected October 19)
When a day or night goes badly, or the weather does something notable, the lab writes it up: what the station measured, what each forecast said and when, and why we missed. A case study never changes the forecast by itself; the ideas it raises are tested on other days first.
Highs and lows, by kind of day
| Graded | Off by, on average | Ran, on average | Outside the ± range | Cloudy or wet ones outside the range | Misses of 5° or more | |
|---|---|---|---|---|---|---|
| Highs | 27 | 1.7° | 0.1° too cold | 33% | 33% of 9 | 1 |
| Lows | 28 | 1.2° | 0.5° too warm | 7% | 14% of 7 | 0 |
The ± is the range the real temperature lands inside 8 times in 10, so about 1 in 5 should land outside it. Looking back over 2½ years of forecasts made the same way (research note R-001): the ± for the high is honest on ordinary days but too narrow on cloudy and wet days (about 1 in 3 outside instead of 1 in 5), and most misses of 5° or more were misses all four weather models made too, usually when the clouds came or went at a different time than forecast.