Fix speed ties

[mrgriffin]

See #4772, this is Alex's suggestion of caching the speed tie-breaks at the start of the turn.

[mrgriffin]

The doubles test fails, and I think this is because each battler shouldn't have a 50-50 chance of being faster than another. As an example, if battler1's bit says it's faster than battler2 and slower than battler3, but battler2's bit says it's faster than batlter3 then we have 3 > 1 > 2 from the first and 2 > 3 from the second, and that's obviously inconsistent.

I think the correct approach would be to generate a number 0..23 which corresponds to the number of permutations of 4 (MAX_BATTLERS_COUNT) elements.

[mrgriffin]

The doubles test fails

Pushed a fix for this, which is exactly as described in my message—I generate a number 0..23 and interpret that as the relative order of all four battlers.

[mrgriffin]

I've un-drafted this. I would like to make the doubles test faster, but I don't think it's possible to do in the test system as it exists rn.

My idea is to replace the 24 PARAMETRIZEs with just counting the order that each battler goes in, and then in a FINALLY (does FINALLY even work for PASSES_RANDOMLY?) we can check that turn order "slot" contains the same count for each battler. But as I say, I'm not sure how to do that. If anybody has suggestions then I'm all ears :)

[hedara90]

Could 3 different fixed damage moves and a healing move work?
Nature's Madness deals 50%, Crush Grip BP depends on HP of target, Dragon Rage 40 and Recover heals 50% of total.
I haven't done the math, but all the orders should result in a different final HP for the target if its max HP is less than 80, I think.

[mrgriffin]

Oh, that's clever! My gut feeling is that should work 🤔

[hedara90]

Attackers
A level 22 Wobbuffet with Crush Grip
Something using Nature's Madness or equivalent
Something at level 25 using Seismic Toss
Target
A level 13 Wobbuffet using Life Dew

I think this could work, but I didn't calculate all permutations.

[AlexOn1ine]

Any reason we can't merge this? This solves the issue.

Also conflicts. I made a new function for ai speed calcs and removed the randomness during a tie but now that there is consistency across the board we should bring it back. Not the function but the speed tie check. I think having the separate function has other benefits.

[hedara90]

Oh, I was working on trying to set up a test that didn't have the massive array of turn orders.
But I hit some problems with it, I might have a setup that works though.
Life Dew, Crush Grip, Endeavor and Super Fang might result in unique final HP values depending on the speed tie.

[AlexOn1ine]

Life Dew, Crush Grip, Endeavor and Super Fang might result in unique final HP values depending on the speed tie.

hm, I feel like we shouldn't hold up a merge because of this since it solves a crucial problem.

[hedara90]

Yea, the test can wait.
I'll try to calculate it soon, but Crush Grip makes everything so much more complicated.

[AlexOn1ine]

Any reason we can't merge this? This solves the issue.

Also conflicts. I made a new function for ai speed calcs and removed the randomness during a tie but now that there is consistency across the board we should bring it back. Not the function but the speed tie check. I think having the separate function has other benefits.

@mrgriffin conflicts 👀

@tertu-m I'll merge it once mgriffin solves conflicts if that's okay

[mrgriffin]

conflicts

Resolved! (Hopefully—I only ran the "Turn order" tests locally because it's bedtime 😅)

[AlexOn1ine]

conflicts

Resolved! (Hopefully—I only ran the "Turn order" tests locally because it's bedtime 😅)

For some reason test/battle/ability/comatose.c:48: Unmatched ANIMATION fails. Which is weird.

[mrgriffin]

For some reason test/battle/ability/comatose.c:48: Unmatched ANIMATION fails. Which is weird.

I watched the replay and it's very weird—the Ditto uses Toxic on itself (that's probably a difference in the speed ties in the old and new code)... But as far as Toxic goes I would guess what's happening is that Ditto tries to go for Celebrate which would be the first move in its moveset... but the first move is now Toxic due to it transforming.

[hedara90]

Test setup:
Player
Wobbuffet, 480 MaxHP, 360 HP, 100 Def
Anything
Opponent
Wobbuffet, 490 HP (default), 100 Atk
Anything

Turn
Player Wobbuffet, Endeavor -> Opponent Wobbuffet
Player Anything, Life Dew
Opponent Wobbuffet, Crush Grip -> Player Wobbuffet
Opponent Anything, Super Fang -> Player Wobbuffet

Possible Outcomes

Player Wobb HP	Foe Wobb HP
188	360
189	360
261	360
235	360
262	360
202	360
189	378
189	189
189	480
188	480
188	240
188	188
262	262
262	142
202	403
202	202
262	283
202	283
235	180
261	180
235	235
235	300
261	141
261	261