I have increased the number of potential spawn point candidates that are considered, in the spawn point formula currently existing, in v4.1.
The main problem with the formula as is is that it is a "good enough" solution. It gets the job done 80% of the time or more which is fine by me considering you already get 3 seconds of invuln after you spawn anyway and you just need to suck it up sometimes.
Anyway, so the original system considers 4 potential, randomly picked, spawn points and then ranks them based on how many enemies are within a 512 unit distance (~40 ft). The lower the number the better, obviously. Well if all 4 spawn points picked to evaluate are near enemies then you are shit out of luck.
I increased the number of spawn points considered to 8 in v4.1 which should bring my "good enough" percentage up to around 95-98% of the time except for really crowded servers on basement, complex, stack, etc.
----
On the contrary, VC's proposed method looks at EVERY spawn point on the map and evaluates the number of players in a weighted normal distribution based on their distance from a threshold to the spawn point. Then the best candidates are thrown into a weighted randomization picker which prevents the occurance of "the same point" being picked over and over.
This method would increase the amount of calculation proportional to the number of spawn points on the map (which can be upwards of up to 30-40!) not to mention increasing the base amount of calculation with the normal distribution and randomization picker. All in all, it is more work for little gain, which is why I rejected it. Plus I am not about to code and debug something that already is implemented and works.