Given the vast complexity of competitive Pokémon, there exist numerous possibilities for abstractions which can simplify the problem space. Outside of more general modifications that can be made to almost any game to decrease the difficulty (e.g., increasing the amount of time available per decision), these relaxations can be grouped into three broad categories — those which reduce the scope of the game as a whole, those which increase the amount of information revealed, and those which pare down the state space through arbitrary restrictions.
Game
As detailed in the glossary, the game of competitive Pokémon is made up of a couple mostly orthogonal components — i.e., team-building and piloting (and optionally drafting). Most often these components are treated separately, as attempting to solve competitive Pokémon as a whole approaches general-game playing in terms of difficulty.
Information
Revealing information which is usually hidden or unknown can prove quite effective at making solving competitive Pokémon tractable, though care must be taken as doing so can dramatically impact the nature of the game:
- knowledge of available choices (e.g., whether the player’s active Pokémon is trapped or disabled) is one of the most minimal modifications that can be made to the game, and mostly exists as a development convenience (allows you to be fully aware of the available choices without having to deal with uncertainty) and has relatively minor implications to actual playing strength or realism
- knowledge of an opponent’s HP (either exact HP or increased accuracy via the
HP Percentage Clause Mod
) is helpful for being able to perform reverse damage calculation and more quickly infer the opponent’s sets, but is still minor enough to not have dramatic gameplay implication (most standard play in Smogon University’s metagames between human players already leverages this) - knowledge of an opponent’s team makeup and configuration is significantly more artificial as it removes the aspect of information gathering/hiding and provides a significant advantage to the player, but as the battle progresses team prediction eventually determines this information anyway. This is the most common simplication and many different options exist in this area
- knowledge of further hidden information known only to the engine (e.g., remaining sleep turns) beyond what’s covered above but while still retaining some stochastic aspects without revealing the RNG seed is difficult to pin down, as exactly what information should be revealed is rather arbitrary
- knowledge of the RNG seed (which both provides full information on existing unknown information such as duration of various statuses and limited knowledge of immediate future outcomes) helps determinize the game tree by effectively removing stochasticity, though the amount of unknowns revealed means that game play starts to get too artificial
- knowledge of the opponent’s upcoming move during the current turn is interesting in that it sequentializes the game (thus removing the complications of dealing with simultaneity), but obviously dramatically changes how the game is played
Note that the preceding options aren’t necessarily always mutually exclusive — modifications can be combined to produce different potentially interesting gameplay variants.
Restrictions
Some of the biggest reductions in information set size can be obtained through restrictions on the rules involved with team-building:
- game options can be restricted to a predefined subset of Pokémon Species/Moves/Abilities/Items. Almost all competitive Pokémon formats restrict options to some extent (e.g., from banning “mythical” Pokémon in Nintendo formats or Pokémon which are highly used in Smogon University’s to a combination of limitations such as those found in Little Cup), though this can be taken to extremes (e.g., a conmeta like the “Shanai Cup”)
- players can be forced build teams only from a pool of specific Pokémon species that were previously selected in a draft phase by the player with knowledge of their opponent’s pool of allowed Pokémon species (e.g., Smogon University’s Draft Leagues). This change reduces the number of possible movesets for both players, but also changes team-building due to being able to cater one’s team through exploiting knowledge of an opponent’s options (it also introduces a “drafting” component to “team-building” and “piloting”)
- allowing players to only choose teams from a fixed pool of sample teams as opposed to building ones from scratch. This simplifies the team-building component (rendering it a “team-selection” component) and also greatly reduces information set size during play provided the pool only consists of a limited number of teams
- requiring Pokémon on teams be drawn from a specific pool of possible sets, similar to how Battle Factory rental Pokémon work on the cartridge games
- allowing for Pokémon sets to vary, but restricting set variation by limiting the pool of possible choices for each field. e.g., Pokémon Showdown’s Random Battle formats
- providing a preview of the opponent’s team before the battle and introducing a phase before the battle takes place where players can select their team order (e.g., most formats post Generation V).
- reducing a team’s maximum size from the standard default of 6 vs.6 to either a strict vs. before the battle, or through a bring choose mechanic where players must select a subset of their original team to actually battle with (e.g., Battle Stadium Singles or Smogon University’s 1v1 format)
- requiring players to share information about their team with their opponent before the match via open team sheets (e.g., Nintendo’s Pokémon Video Game Championship ruleset in 2023)
As with information, some of these restrictions can be combined to arbitrarily bound the size of the information set (e.g., obi’s 1v1 thought experiment which combines maximum team size restrictions and aspects of drafting).