Sunday was a warm late-July day, topping out at 90° at about 4pm and holding there for about two and a half hours. So what was AccuWeather's forecast for 7/31?
81 - 85 - 81 - 89 - 83 - 94 - 92 - 95 - 96
Early on, they forecast extremely cool temperatures, as is AccuWeather's "tease the customer" norm; as the date got closer, their forecast temperatures soared. As it turned out, their lowest high forecast was 9 degrees below the actual temperature; their highest was 6 degrees too high. Again, they never actually forecast the temperature we had, instead just offering scattershot guesses so that no matter what the weather was, they could say that they were pretty close on one day or another (when you forecast highs have a 15° range, it's pretty easy to find at least one day that's close to the mark).
This is telling me that I need to create a subscription weather-forecast site. Apparently there's lots of money to be made in being wrong.