This puzzle is an implementation of Huffman Coding which is a data compression algorithm. Normally you need 8 bits to store each character of text, for example A is 01000001 B is 01000010 etc.
However Huffman Coding uses the frequency which each letter is used and creates a tree by using each letter as a “leaf” or node. The two least used letters are combined to create a new node with a frequency of the sum of the two frequencies of the letters that went into it. That node then gets added back to the available nodes, then you select the next 2 lowest and combine them.
This continues until you only have one left. Once the tree is built you can assign each letter a unique code by following the tree down and appending the bits, when you reach the box with just the single letter that is the end and you have found that letters code. For the most used letters the code will be fewer than 8 bits. What is unique is that no code for any letter is the prefix for any other letter, for example if the letter A was code 01, no other letter would start with 01 as its code, and this works all the way down which means your most used letters will have short codes and save bits, there may be a few longer than 8 bits but there are so few of them the overall length of the file will still be smaller.
I have compressed the coordinates using Huffman Coding, this was the Huffman Tree:
This was my final Compressed message
l9XsvV1f558TNGdB+5vjq/xTPLO3UW5r40qGuGcHrXDe4jejcnCL/XDPPiIJnbqLFM8tcNKg