Damage roll selection in AI_CalcDamage

[Pawkkie]

Description

As is, AI_CalcDamage always assumes the lowest possible damage roll when calculating move damage. This PR adds a parameter to AI_CalcDamage that can specify using an alternate approach based on context, and adds an enum that includes highest, lowest, and average options to start with.

This PR also adds a very simple AI_FLAG_CONSERVATIVE to act as an inverse of AI_FLAG_RISKY that can be used to have trainers always use the lowest damage rolls when scoring their moves.

For all cases where this function is called, I've specified the rolls as follows. Happy to change any of these if we think it's a better idea:

• HasBadOdds: Assume player gets max roll (favour conservative switching decisions)
• Smart mon choices: Assume player gets max roll (favour conservative switching decisions)
• Best mon damage (a default case for vanilla switching): Assume AI gets an average roll
• Should use Z move: Assume AI gets an average roll
• Simulated damage: Use average rather than lowest by default for pre-calculating AI damage. Risky AI will instead use highest rolls, and conservative AI will instead use lowest rolls.

I should also note that the HighestRollDmg function currently accepts dmg and as input and immediately returns it as is. This is really just for future proofing, or very easily letting users change how they want bounds of the AI's damage rolls to behave without needing to change any of the other functions.

People who collaborated with me in this PR

Just me writing it, but @iriv24 gave me the idea!

Discord contact info

@Pawkkie

[DizzyEggg]

Imo it's a good idea, and I'm all for this change.

Why is the PR failing though?

[AlexOn1ine]

Imo it's a good idea, and I'm all for this change.

Why is the PR failing though?

probably failing because a lot of tests assume lowest roll for their damage calcs.

[Pawkkie]

It's failing on exactly one test, which is one I wrote for the switch AI overhaul last fall, "AI_FLAG_SMART_MON_CHOICES: Mid-battle switches prioritize defensive options". In this test the AI should switch out their lead Ponyta to a defensive Aron that can tank the player's attacks. With this PR, the AI doesn't switch at all, and leaves Ponyta in to get KO'd.

I genuinely do not know why this test is failing. My project has deviated from expansion a bit and I'm pulling new stuff individually so it's hard to directly compare, but this is directly copied from the implementation I'm using which passes everything on my end. I was very surprised to see this fail after being submitted.

I'll try to take some time to investigate further, but I'm very open to suggestions, especially regarding major AI stuff that's changed the last few months that might have introduced whatever AI logic changes have caused this deviation and that I might have missed so aren't in my project to cause the same problem.

[Pawkkie]

Okay this should be fixed and finished, the test in question relies on an Aron not being 3HKO'd by a Swellow, and it turns out Boomburst's high roll is literally 34% of its HP with zero IVs and EVs. I have inflated poor Aron's SpDefense in the test as though it had 31 IVs in the stat, which allows it to tank the Boomburst and the test behaves as intended.

Figuring out why it doesn't 3HKO in my project will be a separate problem lol

[Bassoonian]

Conflicts

[Pawkkie]

Conflicts

Updated :)

[Pawkkie]

Spacing fixed :)

raised issues resolved