To be honest weather products are pretty good for temps, cloud cover, and precip predictions for 48 to almost 72 hours. Highly localized conditions like lake-effect precipitation are more difficult to predict exactly for location, that should be expected. The granularity of weather models is in the many km range, not at the km level. The last few times I did flight planning, the weather predictions for ceilings and cloud tops were reasonably accurate en route. It's been pretty rare that tops en route are say, forecast to be 7000 and they turn out to be 14,000. Beyond 48-72 hours the granularity grows exponentially and the timing is increasingly uncertain, but the general sequence of weather features, like frontal passages and temperature changes, are generally correct. Just not certain enough in location and timing to necessarily be useful for specific route planning far into the future. But maybe useful enough to know that in the next week the weather will be likely to be frequently unfavorable for VFR flight, let's say. Or that a stretch of clear weather is likely to take hold.
