At first I thought this is about adding binary digits, it is the literal sum of 3 values

('3y36^2+2y36', 2.6); // is 10296 and not '25.48'
('-2000+0.5x2^3', 5.5); // is -1978 and not '-1916.81'
If you have yo proove that YOUR answer is right so your description is, gently speaking, bad.

Thanks!

At first I didn't understand either. You need to sum all the arguments (they are in decimal) and convert the result to a binary as an array.

better to return empty list in case of incorrect input, avoiding multiple output types

So, what are we supposed to do?

Well no, I think I got it... We are supposed to remove all numbers trailing the variable... What the point of that??? And not a word in the description about that weirdness?

Many tests are wrong:
solvePolynomial('-y198854887455565', 1) should return '-1.00'

solvePolynomial('-y 19 885488 74 555 65', 1) should return '-1.00'

etc.

There are hundreds of katas where half of the work is distinguish True/False from 0/1.
Why fight against language (in python True == 1) in every little kata?

In js how this:
-2000+0.5x2^3', 5.5 => '-1916.81'
If I try to calculate all I'm getting is -1978. In order to get -1916.81 the x = 20.7975.

Fixed.

This killed almost all of the solutions.

Javascript defines both 0 and -0 values, but we do not distinguish these two. (see Examples)

It seems that tests do distinguish between -0.00 and 0.00:

solvePolynomial('-0.00003', 0) should return '0.00': expected '-0.00' to equal '0.00'

For simplicity, the results must be formated using fixed-point notation of 2 digits.

For simplicity of what, exactly? Do we really need all this rounding-based comparison in so many katas? Why, oh why?

Resolved.