An Analysis of Offensive Move Types in Pokemon Uranium *OLD*

[TwistedLogic]

NOTE: This thread is now outdated. For the more up-to-date version, go here.

Considering the large number of new players to the competitive scene in the Pokemon Uranium, I figured it might be fun to do a little exercise in type coverage from a numerical perspective to give new players an idea of what types are good for general coverage, and which ones have more specialized uses.


Methodology:

I have compiled a list of 118 Pokemon that might appear in your battles that includes the following:

1. All fully-evolved Pokemon and Mega Pokemon
   
2. Pokemon that aren’t fully evolved, but can serve a different purpose, or fill a particular role better than its evolution.
   
3. If a Pokemon has more than one ability, and at least one of them affects the Pokemon’s resistances, then the different abilities are counted as separate Pokemon.
   

Unnatural nuclear Pokemon were not included as I felt it would skew the results too high without actually adding anything of note to the data. Pokemon not currently available without hacking are also not included for the obvious reason that the opponent can’t use what they can’t get a hold of.

All Pokemon of interest were compiled into a table with their type resistances and the total number of Pokemon for each “resistance tier” (e.g. “Immune”, “Doubly resistant”, “Weak”, etc.) for each type. These data were then processed in various manners to come up with several possible ranking lists for the types based on different criteria.

The first method  is a ratio of the number of “Positive” (i.e. super effective, marked as Pos) matchups to “Negative” (i.e. immune and not very effective, marked as Neg). Thus, the type’s score is determined by the formula,

.

This, however, does not take neutral coverage (Neu) into account, which can be an important part of the viability of a type for dealing with a wide variety of threats. Thus, for the second method I take this into account, leading to

.

This method proves to too heavily favor a type’s neutral coverage, leading to a set of ranks that would make any competitive player worth their salt laugh. So, a simple ratio doesn’t seem to be the answer. The method I settled on is a weighted average, producing the average effectiveness multiplier you can expect given a random defending Pokemon. This gives us a bit of a more complicated formula:

.


Data:

Table 1 contains a list of all Pokemon of interest, their types, and all resistance values. It is too large to be included in this thread. It is in this doc if you're interested.

Table 2

Table 2 contains the number of Pokemon of interest in each resistance tier of each type, save Nuclear.

Table 3

Table 3 contains the processed data after the formulas detailed in the Methodology section have been applied.

Table 4

Table 4 contains each type’s ranking for each scoring method, as well as an average rank for fun.


Results:

The tier list for each method is as follows:

Rankings

A ranking of each type based on each system.


Problems:

The main problem with this methodology is that it gives each possible Pokemon equal weight as if you’re just as likely to run into a Trawpint as you are an Inflagetah. It also counts some Pokemon multiple times, which could skew results. This could all be rectified by weighing each Pokemon by its usage statistics, but that will require more tournaments and record-keeping in the future in order to get those statistics. I plan on making a follow-up analysis once the metagame has developed enough to have meaningful usage statistics.

*EDIT*
Upon further review of the raw data, I appear to have forgotten to include Stenowatt without Motor Drive. That said, all the additional data would do is slightly change the scores of Ground (with the addition of yet another Pokemon weak to it), and Electric (by adding an additional Pokemon that resists it) by a fraction of a percent. None of these should change the rankings, but I couldn't continue in good conscience without providing some degree of transparency.



Conclusions:

Looking at all three methods, the main thing that’s clear is that the best overall offensive type, aside from Nuclear, is Ground; an unsurprising revelation to any seasoned competitive Pokemon player.

Focusing on the weighted average, which is, in my opinion, the most accurate of the scoring systems, there is more old news, but also some interesting data points worth noting, like that Ice and Fairy are numerically identical when their effectivenesses are averaged out, or that Steel — generally considered a poor attacking type — deals slightly more than normal effectiveness on average, likely due to the number of Fairy-types in the Tandor dex.