I'll take the Numworks calc (Sagittarius).
As for my strategy, I basically found a path through the maze by hand that had as few points as possible, without worrying about my score too much. I made sure to take a very shallow angle when exiting so that the path would be long enough to get the maximum score.
I then tested every single combination of angles and distances that would put me within some distance (if I remember correctly, about 1.5 tiles, though I changed it several times) of my original points, while still allowing me to get the maximum score. This was feasible because you get points off for each digit after the first decimal place, or before the units place, which means that there are only 100 angles and 100 distances to try for each point. The exception for this is the last point, which has to be more than 10 units away in order to meet the length requirement. This took a few hours to run in its entirety, and I was left with a few thousand paths. Prior to the scoring change, all of these paths would have left me with a score of 72, but after the scoring change it was now possible to get a fractional score.
A "heatmap" of all the paths I found. The color of each point represents the last "target" point it was able to get to, with purple being the final point, outside of the maze.The fractional part of the score was determined based on the sine of the sum of the angles and distances, which means that the input to sin() could be no more precise than to the tenths place. For the range of numbers that this sum could possibly be without adding a full point to my score, the tenth whose sine is closest to -1 is 11.0, and sin(11) is = -0.999990207.
Unfortunately, none of the few thousand paths that I had found added up to 11.0. However, I realized that I was only calling a_droite() in my path, and that if I replaced a_droite(x) with a_gauche(-x), x would be subtracted from the sum rather than added to it, without actually changing the validity of the path. I tried out every combination of a_droite and a_gauche calls for each of my few thousand paths, and found one that added up to 11.0, giving me a score of 71.0000097934493.
Here's the code that I used to find this path.
You'll notice that this is still very slightly off from my actual score - I realized that I could slightly decrease sin(sum) by slightly decreasing sum, and that I could slightly decrease sum by subtracting a tiny amount from each distance and angle. Normally, this would cause my score to increase by several full points because of the decimal places after the tenths place, but because the score is rounded to 5 decimal places before the length of the number is calculated, if the difference is less than 0.000005, you aren't penalized. So, I subtracted 0.0000049999... from each distance, to get a sum of 10.99991, a sine of -0.99999060081, and a final score of 71.00000939918644. I believe that this is within a floating-point error of the ideal solution.
Here's the solution that I submitted, for reference.
I actually tested all of this on a computer, as I didn't have my EP with me. I ported polycalc to use cv2 for drawing stuff.
I also tried to submit a few, uh, let's call them
gimmicky, solutions. My first instinct was to try an integer overflow, but to my disappointment, Python doesn't actually have those. The numbers, they just keep
going.
After that, I tried
subclassing the float class so that it would return an empty string. Technically, this wasn't modifying the internal state of laby.py

. I could have also returned a custom string class with a modified __len__ method, but in CPython __len__ must return a positive integer, which also gets converted into a regular integer, not one that I could subclass. Interestingly, MicroPython allows you to return a negative length for __len__, so I could have probably gotten a score of negative infinity if the admins didn't shut that idea down

.
Finally, I also found a solution that's far better than the others that I decided not to submit since I was already in first.
Here's the link if you want to give it a try for yourself.