Yesterday I arrived at the solution to step 1 of the problem of predicting rotation decisions given variability of rage gains. Step 1 is finding the probability distribution of rage gains on a given time interval. Due to same-speed weaponry, there are multiple ways to arrive at the same rage income amount via multiple combinations of hits/misses of melee swings. This proved a kind of organizational challenge because the solution had to work for both fury styles, and a wide range of attack rates and time intervals.
In lay terms, probability distribution basically is the answer to this question: what is the probability to get each individual rage amount such that all probabilities add up to 100%. Here's a picture of a scenario with average swing speeds with dual 2.6, 13% miss rate, and 4x a length of a standard dual-GCD combo
As the number of swings checked increases, since the experiment follows either multinomial or binomial distribution and due to increased observations of misses, a phenomenon occurs via the Central Limit Theorem which allows the shape of the distribution to mostly follow the familiar bell-shape.
I can use this dynamic solution to solve step 2, which will be determining the probability distribution of rage-based rotation decisions per given sections in the rotation. The goal is to be able to predict where Heroic Strikes are in a rotation, how Battle Trance is utilized, and if and how abilities are skipped due to rage lulls. The original motivation behind this system was to have a system that can be used to more properly estimate Execute damage. Execute damage has always been an unanswered problem in my spreadsheets, and hopefully I will be able to deliver on it this time.
The sheet is so much more powerful now than it used to be when it comes to compartmentalization; It's so much more de-vacuumed, heterogeneous, salad-bowl, whatever you want to call it, than before. I'm actually having some fun though utilizing the training I am getting at school for actual projects.
