Yea, but that’s why the confusion matrix exists. In machine learning, having a 90% accuracy is fruitless if you’re missing all the true positives (leading to 0% precision and 0% recall). Likewise, if you say it’s gonna snow all the time, you’ll catch all the true positives (snowstorms), but even though your recall will be 100%, your accuracy and precision will be 10%. F1 score is a key metric in weather forecast scoring, I would imagine.