Move Scoring & Switch Conditions

Move Scoring & Switch Conditions

Move scoring rules modify the score of individual moves during evaluation. Switch conditions control when and how the AI considers switching Pokemon. Together, these two systems handle the fine-grained decision logic that runs before action priorities apply their broader strategic rules.

Move Scoring Rules

Unlike action priorities (which apply once per action type), move scoring rules are evaluated for each move individually. They are the primary way to teach the AI about type effectiveness, immunities, accuracy, and other per-move factors.

{
  "move_scoring_rules": [
    {
      "id": "stab_bonus",
      "condition": "q.move.is_stab",
      "score_modifier": {
        "type": "multiply",
        "value": "1.5"
      },
      "priority": 20
    }
  ]
}

Move Scoring Fields

Field	JSON Key	Type	Default	Description
ID	id	string	Required	Unique name.
Condition	condition	Molang	Required	When this evaluates to >0, the modifier is applied. Use "1" to mean “always.”
Move Condition	move_condition	Molang (optional)	null	Additional condition with access to q.move.* queries.
Score Modifier	score_modifier	object	Required	How to modify the score. See Score Modifier Types below.
Priority	priority	int	0	Reserved for future sorting. Currently, rules are applied in definition order.

Score Modifier Types

Each rule’s score_modifier has a type and a value. The value field is a Molang expression, so it can reference queries dynamically.

Type	Behavior	Example
multiply	new_score = old_score x value	\{"type": "multiply", "value": "1.5"\}
add	new_score = old_score + value	\{"type": "add", "value": "50"\}
set	new_score = value (ignores old score)	\{"type": "set", "value": "0"\}
script	new_score = value (Molang expression returns the final value)	\{"type": "script", "value": "q.move.power * q.move.type_effectiveness"\}

Dynamic Values

Because value is a Molang expression, you can write rules like:

{
  "type": "multiply",
  "value": "q.move.type_effectiveness"
}

This multiplies the move’s score by its type effectiveness (2.0 for super effective, 0.5 for not very effective, etc.).

Common Move Scoring Rules

Type Effectiveness — Always factor in type matchups:

{
  "id": "type_effectiveness",
  "condition": "1",
  "score_modifier": { "type": "multiply", "value": "q.move.type_effectiveness" },
  "priority": 10
}

Immunity Check — Zero out moves that the target is immune to:

{
  "id": "immunity_check",
  "condition": "q.move.type_effectiveness == 0",
  "score_modifier": { "type": "set", "value": "0" },
  "priority": 100
}

Predicted KO — Heavily boost moves that can knock out the target:

{
  "id": "predicted_ko",
  "condition": "q.move.predicted_damage_percent >= q.target.current_hp_percent",
  "score_modifier": { "type": "multiply", "value": "3.0" },
  "priority": 50
}

Accuracy Penalty — Penalize inaccurate moves proportionally:

{
  "id": "accuracy_penalty",
  "condition": "q.move.accuracy < 100 && q.move.accuracy > 0",
  "score_modifier": { "type": "multiply", "value": "q.move.accuracy / 100" },
  "priority": 5
}

Status Redundant — Do not use status moves on already-statused targets:

{
  "id": "status_redundant",
  "condition": "q.move.is_status && q.target.has_status",
  "score_modifier": { "type": "multiply", "value": "0.05" },
  "priority": 60
}

Ability Immunity — Respect ability-based immunities (e.g., Flash Fire):

{
  "id": "flash_fire_check",
  "condition": "q.target.revealed_ability == 'flashfire' && q.move.type == 'fire'",
  "score_modifier": { "type": "set", "value": "0" },
  "priority": 95
}

Switch Conditions

Switch conditions define when and how the AI considers switching Pokemon. They are evaluated during switch scoring — each condition checks the currently active Pokemon and optionally filters which bench Pokemon are valid switch targets.

{
  "switch_conditions": [
    {
      "id": "bad_matchup",
      "condition": "q.pokemon.under_threat > 0.4 && q.pokemon.current_hp_percent > 0.4",
      "target_condition": "q.switch_candidate.matchup_score > 1",
      "priority": 60,
      "bias_multiplier": 1.5
    }
  ]
}

Switch Condition Fields

Field	JSON Key	Type	Default	Description
ID	id	string	Required	Unique name.
Condition	condition	Molang	Required	Evaluated for the currently active Pokemon. Uses q.pokemon.* queries.
Target Condition	target_condition	Molang (optional)	null	Evaluated for each switch candidate. Only candidates where this passes are considered. Uses q.switch_candidate.* queries.
Priority	priority	int	0	Reserved for future sorting. Currently, conditions are applied in definition order.
Bias Multiplier	bias_multiplier	double	1.2	Multiplied with the switch candidate’s score. Use <1.0 to discourage switching, 0.0 to prevent it entirely.

Common Switch Conditions

Never Switch When Trapped — Respect trapping effects:

{
  "id": "no_switch_trapped",
  "condition": "q.pokemon.is_trapped",
  "priority": 200,
  "bias_multiplier": 0.0
}

Do Not Abandon Boosted Pokemon — Discourage switching away from Pokemon with significant stat boosts:

{
  "id": "no_switch_boosted",
  "condition": "q.pokemon.stat_boosts.atk >= 2 || q.pokemon.stat_boosts.spa >= 2",
  "priority": 90,
  "bias_multiplier": 0.2
}

Preserve Low HP Pokemon — Encourage switching out a weakened Pokemon that is still under threat, preserving it for later:

{
  "id": "preserve_low_hp",
  "condition": "q.pokemon.current_hp_percent < 0.3 && q.pokemon.under_threat > 0.25",
  "priority": 75,
  "bias_multiplier": 1.8
}

Switches vs. Action Priorities

Switch conditions modify the score of individual switch candidates (which bench Pokemon to switch to). Action priorities with action_type: "switch" modify the score of the switch action itself (whether to switch at all). Use switch conditions for “switch to the right Pokemon” logic and action priorities for “should I switch?” logic.