$ sign wasn't even necessary.

I used a solution similar to this one at first and passed but when I came back and tried an optimized version the optimized code does pass but this old version won't anymore. I think sometimes you can get lucky with this approach if the randomized tests are smaller numbers, but in general they expect the solution to be more optimized. This should probably be higher than kyu 6 imo because 7's generally don't care so much about optimization and this unoptomized approach is the more obvious answer.

Same here, I couldn't figure out how to implement regex like this one then I gave up xD.

You're misreading the logs, your code fails the 4th test, not the 1st.

But shouldn't that be enough to return "true" in the the 1rst test case?

And that's exactly what is happening. Just like the message indicates, your code incorrectly returns true when it should return false.

The name of the function should be count, not count_occur. You can confirm this by checking how your function is being called in the sample tests:

function test(string, expected) {
  assert.deepEqual(count(string), expected, `Test failed for string = '${string}'`);
}

The solution setup (the code you see when you enter the trainer for the first time) also includes a function named count.

Finally, please use appropriate formatting (code blocks) when posting code.