Most machine learning courses race past the statistics and assume you either already get it or are happy to take it on faith, and that is exactly the gap StatQuest fills better than anything else I have found. Josh Starmer takes one idea at a time, things like p-values, logistic regression, ROC curves, random forests, gradient descent, and backpropagation, and builds them up from the ground with simple drawings and worked examples until you actually understand what is going on rather than just memorising a formula. The videos are short and ruthlessly focused, so when you hit a wall in another course you can search for the exact topic, watch a few minutes, and go back with the thing finally making sense. His style is famously goofy, with little songs and a triple BAM when a concept lands, and while I know that grates on some people, I think it does real work in keeping you relaxed through material that usually makes beginners tense up.
The honest limitation is that this is a library and not a curriculum, so there is no set order, no projects, and no one telling you what to learn next, which means StatQuest works best as a companion to a structured course rather than as your only source. There is a paid side too, with PDF study guides and the illustrated machine learning book, and they are nicely done if you like having something to print and annotate, but you can get enormous value without ever spending a penny. If you have ever felt like you can run the code but do not really know why it works, this is the resource that fixes that, and I recommend it to almost everyone learning this stuff.