Data Pokemon Sun/Moon Battle Mechanics Research

[Cresselia92]

For such a major mechanic, I think the lack of real confirmation is enough to de-confirm it. Throw it in the same box as the Unaware nerf and the Eviolite nerf for "changes that would be kind of nice but don't actually exist."

Personally, I think the main problem is that there isn't a confirmation from either side. Nobody confirmed the nerf, but nobody confirmed the lack of nerf either. Unaware and Eviolite had been tested before getting the confirmation, at least.

All in all, confusion is a quite tricky matter, since it is very luck-based. There is that guy/gal who tested the confusion hit rate many times, but due to the nature of confusion itself, that could have been more a matter of incredible/terrible luck rather than a conclusive test.

If someone had to confirm this, it would be better if the testers actually checked the mechanics themselves inside the games, rather than relying on (unreliable) calculations.

So, unless there is some programmer that manages to crack down the mechanisms themselves, I don't think we can give a final verdict yet. After all, we're talking about the company who suddenly gave to Dark-types an immunity to Prankster-affected moves, so anything is possible.

 

[UltiMario]

I haven't done any personal testing, but I've seen multiple people post sample sizes of 50-100 and every last one of them has been ~50%. I'm pretty sure Confusion wasn't nerfed.

 

[Edu]

When a Pokémon with Trace get in against Battle Bound, Trace will not active, but if the Grenin is defeated, than Trace will activate against the replacement poce.

 

[SeraviEdalborez]

When a Pokémon with Trace get in against Battle Bound, Trace will not active, but if the Grenin is defeated, than Trace will activate against the replacement poce.

Same goes for basically all un-Trace-able abilities, such as Schooling or Stance Change.

 

[LazyDex]

Personally, I think the main problem is that there isn't a confirmation from either side. Nobody confirmed the nerf, but nobody confirmed the lack of nerf either. Unaware and Eviolite had been tested before getting the confirmation, at least.

All in all, confusion is a quite tricky matter, since it is very luck-based. There is that guy/gal who tested the confusion hit rate many times, but due to the nature of confusion itself, that could have been more a matter of incredible/terrible luck rather than a conclusive test.

If someone had to confirm this, it would be better if the testers actually checked the mechanics themselves inside the games, rather than relying on (unreliable) calculations.

So, unless there is some programmer that manages to crack down the mechanisms themselves, I don't think we can give a final verdict yet. After all, we're talking about the company who suddenly gave to Dark-types an immunity to Prankster-affected moves, so anything is possible.

Actually, I remember that the reason that all the talk about confusion getting nerfed is a thing, is because of what the dataminers found concerning a change in the in-game probability calculations for hitting yourself in confusion. If I remember correctly, the overall probability based on these calculations was then determined to be an estimated 1/3, thus leading to the conclusion that confusion had been nerfed. Just to make sure, is anything like this, found by the dataminers (the same guys who discovered things like the Leech Life buff and the Thunder Wave nerf), considered to be credible? (No offense to them. Those guys gave us all that wonderful info, after all. This is just my paranoia/OCD wanting to make absolutely sure.)

 

[UltiMario]

Actually, I remember that the reason that all the talk about confusion getting nerfed is a thing, is because of what the dataminers found concerning a change in the in-game probability calculations for hitting yourself in confusion. If I remember correctly, the overall probability based on these calculations was then determined to be an estimated 1/3, thus leading to the conclusion that confusion had been nerfed. Just to make sure, is anything like this, found by the dataminers (the same guys who discovered things like the Leech Life buff and the Thunder Wave nerf), considered to be credible? (No offense to them. Those guys gave us all that wonderful info, after all. This is just my paranoia/OCD wanting to make absolutely sure.)

The 1/3 Confusion came from some a single Japanese user on twitter during the Sun and Moon Demo, not one of the dataminers. Everything you just said was entirely made up.

 

[Siggu]

The only strange thing I noticed about confusion in my game is that it almost always lasts the minimum amount of turns, but it's just luck... I guess.

 

[Talpr0ne]

The 1/3 Confusion came from some a single Japanese user on twitter during the Sun and Moon Demo, not one of the dataminers. Everything you just said was entirely made up.

Is the self hit ratio info that can't be datamined or what? Why not simply ask to a dataminer instead of sitting here doing pointless and unreliable math?

 

[B. Katarn]

Is the self hit ratio info that can't be datamined or what? Why not simply ask to a dataminer instead of sitting here doing pointless and unreliable math?

I believe the problem is it is not so easily tracked down. The data is not in readable computer code like Perl and many of the discoveries were done using known pointers from older games, such as what the new TM list was but not knowing the new moves' identities. Finding the part that deals with confusion and then interpreting that into a value is hard... not impossible, but hard.

Also, consider the fact that RBY is a miniscule file size (11MB) but we only learned of the Body Slam / Normal types effect in Gen1 months ago.

 

[Marty]

What happens if Instruct is used on a Pokémon that flinched? There are a few cases here:

A Pokémon and its opponent switch in on the same turn. Next turn, opponent uses Fake Out, making the Pokémon flinch. It is then Instructed.

Since there's no last used move, it fails.

A Pokémon uses a move while its opponent switches. Next turn, it chooses a different move, but its opponent uses Fake Out, making the Pokémon flinch. It is then Instructed.

The Pokemon flinches again.

A Pokémon uses a move while its opponent switches. Next turn, its ally tries to Instruct it, but its opponent used Fake Out. It then tries to use a different move.

The Pokemon flinches, and then flinches again.

Dumping the rest of the Instruct tests I did some time ago here:

Instruct fails to work on

rampage moves (Thrash)
charge moves (Dig)
recharge moves (Hyper Beam)
Mirror Move
Metronome
Me First
Mimic
Focus Punch and Beak Blast and Shell Trap (fails DURING the "charge" too)
Copycat

I also checked your Reflect Type VS Burn Up question from the bug reports thread. Although Reflect Type fails against purely "typeless" previously-Fire-type Pokemon, by using Forest's Curse on it, then using Burn Up, and then using Reflect Type on it, the Reflect Type user becomes Normal/Grass. I'll be interested to know how Moltres works, if having a secondary type to fall back to is different from an added type in this case.

 

[CPU]

Perhaps worth noting for confusion tests is that my in-game team very seldom hit themselves in confusion. Considering that playing with Pokémon in Pokémon Refresh allows them to cure themselves of status conditions, I wouldn't be surprised if they also have lower confusion self-hit rates. Not applicable for competitive play, obviously, but perhaps pertinent for testing.

Or I was just lucky.

 

[UltiMario]

Marty If you use Helping Hand (or whatever 4+ prio move), Fake Out the Helping Hand Pokemon, then Instruct it, does it Flinch? Do moves like Iron Head check for flinch on Instruct? I'm guessing the answer to both is yes, and as a sidenote, Dancer should work similarly (flinchcheck on dancer procs).

 

[Marty]

Marty If you use Helping Hand (or whatever 4+ prio move), Fake Out the Helping Hand Pokemon, then Instruct it, does it Flinch? Do moves like Iron Head check for flinch on Instruct? I'm guessing the answer to both is yes, and as a sidenote, Dancer should work similarly (flinchcheck on dancer procs).

Indeed, the flinch flag can be set even if the Pokemon's place in turn order passed already, resulting in a flinch during moves executed by Instruct or Dancer later that turn.

 

[Charlotte]

Looking into the confusion stuff a bit more because the doubt annoys me (I would love evidence from the code too but what we have is sufficient)
I did 3 battle in the battle tree (I don't use refresh anyway, but to make sure the extra effects aren't active) in which I clicked confuse ray numerous times, totaling in 203 confusion rolls, 68 of which were self-hit. data here http://pastebin.com/4ucE7uyM
Of course, if it still were 1/2 on self-hit then the expected number of self-hits would have been 101.5, a lot higher than 68. In fact signifcantly so with a z-score of about 3.37 and a p-value < 0.01
The Japanese large sample test from the demo (which was streamed) came to the same conclusion when evaluated, confusion self-hit is extremely likely (can't stress how much so) to have been reduced.

 

[LazyDex]

UltiMario, I didn't make it up (at least intentionally), I honestly thought it was data from the datamine. Looks like it's yet another instance of me getting my details screwed up. Again.

New question I just thought of, though: Do spread moves used in a battle royal do 50% damage or 30% damage (or some other lowered value)?

 

[Hesh Kadesh]

30% I believe is the text used in game. Do not know if mechanics supports it though.

 

[UltiMario]

Looking into the confusion stuff a bit more because the doubt annoys me (I would love evidence from the code too but what we have is sufficient)
I did 3 battle in the battle tree (I don't use refresh anyway, but to make sure the extra effects aren't active) in which I clicked confuse ray numerous times, totaling in 203 confusion rolls, 68 of which were self-hit. data here http://pastebin.com/4ucE7uyM
Of course, if it still were 1/2 on self-hit then the expected number of self-hits would have been 101.5, a lot higher than 68. In fact signifcantly so with a z-score of about 3.37 and a p-value < 0.01
The Japanese large sample test from the demo (which was streamed) came to the same conclusion when evaluated, confusion self-hit is extremely likely (can't stress how much so) to have been reduced.

Further digging into what Japanese tests went for (it looks like we've got tests in the thousands that favor 33%) it certainly seems that way, but there's things that irk me. It's because of fairly frequent posts like this.

testing of confusion self-hit: 47 of 93 uses hit self in confusion. looks to still be 50%, though there's of course margin for error with this few tests. don't have the patience to do more trials right now however.

I know I saw someone else post 54/100 as a sample they got somewhere here (I swore it was in this thread but I can't find it) and some other smaller samples too. Obviously these are tiny sample sizes, but the frequency of them irks me heavily, especially with how difficult its been to pin down confusion's chances.

I don't think it's an impossibility that there's something else manipulating confusion RNG in SM. I wouldn't doubt that something being completely irrelevant to competitive since it'd likely be an in-game mechanic, so I feel like I shouldn't worry about it, but it bothers me that we have some small sample sizes that push closer to 1/2 but bigger ones pushing towards 1/3. It could just be a statistical anomaly, but it bothers me enough to the point that I feel like it's probably worth investigating.

 

[SupremeDirt]

Further digging into what Japanese tests went for (it looks like we've got tests in the thousands that favor 33%) it certainly seems that way, but there's things that irk me. It's because of fairly frequent posts like this.



I know I saw someone else post 54/100 as a sample they got somewhere here (I swore it was in this thread but I can't find it) and some other smaller samples too. Obviously these are tiny sample sizes, but the frequency of them irks me heavily, especially with how difficult its been to pin down confusion's chances.

I don't think it's an impossibility that there's something else manipulating confusion RNG in SM. I wouldn't doubt that something being completely irrelevant to competitive since it'd likely be an in-game mechanic, so I feel like I shouldn't worry about it, but it bothers me that we have some small sample sizes that push closer to 1/2 but bigger ones pushing towards 1/3. It could just be a statistical anomaly, but it bothers me enough to the point that I feel like it's probably worth investigating.

would like to point out it's entirely possible that those rolls for me were influenced by pokemon refresh, i'm honestly not sure. I'd forgotten at the time that that could affect that sort of thing so I hadn't checked. I'd like to say that it wasn't, but I can't confidently say as much.

 

[Aqqq]

Further digging into what Japanese tests went for (it looks like we've got tests in the thousands that favor 33%) it certainly seems that way, but there's things that irk me. It's because of fairly frequent posts like this.



I know I saw someone else post 54/100 as a sample they got somewhere here (I swore it was in this thread but I can't find it) and some other smaller samples too. Obviously these are tiny sample sizes, but the frequency of them irks me heavily, especially with how difficult its been to pin down confusion's chances.

I don't think it's an impossibility that there's something else manipulating confusion RNG in SM. I wouldn't doubt that something being completely irrelevant to competitive since it'd likely be an in-game mechanic, so I feel like I shouldn't worry about it, but it bothers me that we have some small sample sizes that push closer to 1/2 but bigger ones pushing towards 1/3. It could just be a statistical anomaly, but it bothers me enough to the point that I feel like it's probably worth investigating.

Those sample sizes aren't even particularly small. They're obviously not as conclusive as a sample size of 1000, but they're enough to make it extremely unlikely those samples were obtained if the actual chance was 33%. For instance, the probability of getting at least 47 self-hits out of 93 with a 33% chance each time is less than 0.1%.

So it is very likely either there is some other mechanic involved, or some people have been doing something wrong in gathering their data.

 

[SunDriedRainbow]

Sorry if this is duplicate info, but Encore cannot prevent a Z-Move from being called.

Turn 1: Pokemon A uses move 1, Pokemon B uses move 2.
Turn 2: Pokemon A uses Encore, Pokemon B uses Z-Move 3. It succeeds.

 

[TwilightMystery]

UltiMario, I didn't make it up (at least intentionally), I honestly thought it was data from the datamine. Looks like it's yet another instance of me getting my details screwed up. Again.

New question I just thought of, though: Do spread moves used in a battle royal do 50% damage or 30% damage (or some other lowered value)?

I have the player's guide, and it says all moves that hit multiple targets deal 50% in battle royale and 75% normal value in double battles.

 

[Siggu]

I'm not sure if this is how Torment was already working (does not help that Bulbapedia, as unreliable as it can be, has the effect description worded really badly) but this happened:
- Mimikyu used Play Rough and OHKOed Manectric. Opponent sends out Noivern.
- Noivern used Torment.
- Mimikyu used Play Rough successfully (and OHKOed Noivern but that's not the point).

I'm convinced that Torment could prevent a move used in the same turn...

Proof: NHTG-WWWW-WWW3-Q4CE

 

[Accel]

I'm not sure if this is how Torment was already working (does not help that Bulbapedia, as unreliable as it can be, has the effect description worded really badly) but this happened:
- Mimikyu used Play Rough and OHKOed Manectric. Opponent sends out Noivern.
- Noivern used Torment.
- Mimikyu used Play Rough successfully (and OHKOed Noivern but that's not the point).

I'm convinced that Torment could prevent a move used in the same turn...

Proof: NHTG-WWWW-WWW3-Q4CE

Torment prevents your opponent from using the same move twice in a row after you use the move itself, and its effect continues even after the user faints.

Hypothetical situation: Let's say I'm in with a Hydreigon vs a Clefable which has moonblast as its only attacking move. I click Torment and die to the moonblast, and now I'm allowed to bring in a setup sweeper of my choice and set up while the Clefable is forced to use any other of its 3 non-attacking moves or switch out.

 

[CatKawaii]

What would happen if I use Revelation Dance with an -ate/Galvanize Ability?

 

[pyuk]

Since this whole confusion chance thing seems to still be under debate, I've decided to start collecting my own data. These preliminary results unfortunately don't include data on the duration of confusion, but I'll start counting that from now on. I tested in the wild in an SOS chain of Oricorios while spamming Teeter Dance with a Tangled Feet Spinda, and in 212 attempts at making moves by confused Pokémon (including attempts to copy Teeter Dance by the Oricorio), 135 of the moves succeeded, while 77 resulted in a self-hit. This comes out to about a 36.32% chance to hit one's self in confusion, disregarding the turn a Pokémon snaps out. It's not a terribly large sample size, and again, I neglected to count the number of times a Pokémon snapped out of confusion, but as I was testing, the ratio seemed to stay around 40-60 of self-hits to successful moves through Confusion.

Also, I noticed some interesting mechanics with Dancer while collecting this data, though these may already be known.

• Attempting to use a move through Dancer while confused has to pass through the roll, same as any other move, and this also counts towards Confusion's duration counter.
• Dancer didn't copy Teeter Dance unless it successfully confused at least one Pokémon.
• For some reason, when there are two Pokémon on the field with Dancer, and a third Pokémon uses a dance move, the slower Pokémon with Dancer will copy the move first. At first, I thought it might've just been speed ties, but this was consistently the case, and I never noticed any two Oricorio actually tying. On a somewhat related note, Teeter Dance applied confusion in the same exact order every time, regardless of speed: ally first, then (the user's) right to left on the opponent's side.