Is a statistical way to describe how runs are randomly generated based Identically generated from some distribution Q supported on ,Īnd the outcomes of each run are independent. Run, where the run-specific probabilities are independently and We can even go a step further in this model: ThisĬaptures the intuition that some seeds are easier or harder than others,Īnd allows for the example described in Model 1 of an unwinnable seed. P i of winning each run is essentially correct. The assumption that there is a run-specific probability As a statistician, I think making this assumption is a very Not losing too much statistically by making this assumption (and it’sĪctually pretty hard to weaken it as I’ll discuss), and it’s at leastĬorrect marginally for any run (i.e. any given run by itself is The assumption that the outcomes of each run are independent still isn’tĮxactly correct as discussed in Model 1. Run-specific probability p i ∈ of winning each We next introduce a model that deals with the assumption of a constant We learn from our previous runs, which can affect how we play futureįor the individual trials implies the following.Losing on one run, it can affect your next run. We aren’t perfect spire playing machines, so if you get tilted.When playing base Slay the Spire (i.e. using no mods that modify theĬore gameplay), losing before the act 1 boss on a run affects the.The assumption that the outcomes of each run are independent isn’t So the probability of winning a run on that seed is 0. As a concreteĮxample, it is well known that there is a seed for The Silent that is Run for the player is definitely not correct: some runs are harder orĮasier than other runs based on the seed of the run. The assumption that there is a constant probability p of winning each (in statistics this essentially means no run influences the outcome of Translating this model to English, this model says: the player has aĬonstant probability p ∈ of winning each run, and the outcomes Probability that a given player wins a run, where average is respect to We are really trying to estimate the average TLDR: Under an appropriate model, when we talk about winrate, Increasing complexity that will help us answer the questions I’ve posed These n outcomes as the data, and posit some model for the data in Individual runs, where x i is 0 if the run was won, or 1 I apologize about the math formatting, I have not figured out how to get mathjax working! What is Winrate Actually Estimating? (ABD), so I’d hope I know what I’m talking about on the statistics side ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |