The dumbest thing the AI has ever done?

Who were you using? If Rapidash didn't have any non-resisted moves his AI may have decided it was better to heal (remember the AI takes the type of status moves into account, which is why you can beat Lance's Dragonite with a low level Tentacool since he'll keep on using Barrier).
A Rhyhorn. (Fun fact: After that battle it evolved into Rhydon)
 

Pikachu315111

Ranting & Raving!
is a Community Contributoris a Top Smogon Media Contributor
A Rhyhorn. (Fun fact: After that battle it evolved into Rhydon)
As I thought. Rhydon is part Rock-type which resists Fire and Normal, all what Blaine's Rapidash knew. Blaine's AI must have concluded that since all its moves are resisted and it had an item, using the item was the "best" choice.

So I remember a very distinct case of AI stupidity for my first playthrough. It was against Lusamine round 2, and her Clefable against my Kommo-o. I had forgotten to swap out of it to Magnezone, so I decided, "eh, fuck it. Might as well sac it so something can come in for free," and went with Work Up.

It used Moonlight, apparently figuring I was going to Poison Jab it or something.

I just took it, and kept buffing up, while it kept spamming Moonlight until I was at +5. Then I nuked Clefable with Poison Jab, and murdered the rest of her team, with only Mismagius landing a hit on me with Shadow Ball

Easiest boss fight ever :P
And even weirder it has Moonblast so if it was going for type advantage it should have used that. Only thing I can think is the AI derped and kept using the wrong Fairy-type move. If Bulbapedia's ordering of moves it to be believed, in the first fight her Clefable's Moonblast was its third move (it didn't have Moonlight nor a fourth move) and during her second it was moves to the first slot with Moonlight being put in the fourth. Maybe the AI got the placement confused between the fight and so when it looked for a super effective move it either starting on the third slot (and went to Moonlight one slot over) or on the last slot. I would say the AI programming should be better nowadays to not do that, but hey, programming hiccups happen.
 
A spin-off case: in Pokemon Super Mystery Dungeon, the AI thinks it has infinite PP.

Have you allowed one of your partners to use a room-clearing move? Behold how he wastes all that move's low PP on a single enemy in a tiny room.
Have you allowed one of your partners to use a ranged move? No problem! He will endlessly spam the move until the enemy is at melee range, even if he has melee moves that deal more damage.
Is one of your partners set to "Let's go together!", your bag is full and there is some dropped Poké? Nah, the bag is full, your partner is not going to pick it up.

Huh? What is that? Is there an enemy Sawk in the floor?

Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!
Sawk used Rock Smash!

And suddenly it wasted all its PP on just making a room bigger... and finding nothing of relevance.

At least I gotta give Chunsoft some merit in that the AI is much more intelligent than in Explorers, where partners could get ridiculously stupid the moment they were two tiles away from you.

While they do have that atrocious lack of PP-preservation intelligence, at least they know how to use ranged moves well.

UPDATE: Adding another one from the Battle Tree.

Opponents using Thunder Wave when Trick Room was active.
 
Last edited:
Okay, I already posted this in the Battle Tree thread, but it DESERVES to be posted here, so...

Super Multi Battles. I'm paired up with Cynthia and her Weakness Policy Lucario and Scarfed Garchomp. We're at match 49, and Lucario's already downed the enemy's Empoleon, while I decide to ignore the Eelektross it was paired up with to help out Lucario since it struggles to get kills at times. Regirock comes in, and I Grass Knot it, so Lucario's Aura Sphere can kill it off....

It Aura Spheres the EELEKTROSS instead. FOR SOME RETARDED REASON. Cue Regirock EQing both Lucario and Nihilego to oblivion.

Me: Okay, we can still salvage this. I just need Cynthia's Garchomp to Outrage, and that should break down the eel.

It EQ's instead, and Celesteela can't break through Eelektross before the eel roasts her alive. The horrible gong show then ends as Garchomp struggles itself to death against an eel that was slowly burning it.

I'm beyond pissed right now. This was the stupidest thing I've ever seen in ANY Battle Facility.
 
Voltorb in Leafgreen

Screech, Screech, Screech, Screech, Sonicboom (Then I killed it)
Uh, is that a trainer's Voltorb or a wild one? It's pretty much assumed that the AI for wild Pokemon will just use its moves randomly, while trainers, especially gym leaders and battle facility trainers, will be more strategic with their moves. (except when they do something incredibly dumb like using Trick Room the turn after it already used it or alternating between weather it sets up)
 
Uh, is that a trainer's Voltorb or a wild one? It's pretty much assumed that the AI for wild Pokemon will just use its moves randomly, while trainers, especially gym leaders and battle facility trainers, will be more strategic with their moves. (except when they do something incredibly dumb like using Trick Room the turn after it already used it or alternating between weather it sets up)
Wild. I don't really know how the AI works, just that it can be stupid at times. I don't think the fact that it was wild takes away from the fact that it was not a smart move at all, though.

Edit: In case you don't realize, screech lowers physical defense, and sonicboom is a ranged move that always does 20 HP
 

Pikachu315111

Ranting & Raving!
is a Community Contributoris a Top Smogon Media Contributor
Wild. I don't really know how the AI works, just that it can be stupid at times. I don't think the fact that it was wild takes away from the fact that it was not a smart move at all, though.

Edit: In case you don't realize, screech lowers physical defense, and sonicboom is a ranged move that always does 20 HP
Wild Pokemon uses one of their moves at random, they don't use any strategy. Not sure if this is also true for Legendaries.
 
Wild. I don't really know how the AI works, just that it can be stupid at times. I don't think the fact that it was wild takes away from the fact that it was not a smart move at all, though.

Edit: In case you don't realize, screech lowers physical defense, and sonicboom is a ranged move that always does 20 HP
To be fair, wild AI is completely random. The only thing that probably affects it is nature as in GenIII's mechanic (note: probably. Not sure as of now)
 
Wild Pokemon uses one of their moves at random, they don't use any strategy. Not sure if this is also true for Legendaries.
To be fair, wild AI is completely random. The only thing that probably affects it is nature as in GenIII's mechanic (note: probably. Not sure as of now)
Interesting. I suppose the chance that I would have gotten four screeches in a row is fairly low then, assuming it has four moves. (I don't know, I didn't catch it.)
 
Interesting. I suppose the chance that I would have gotten four screeches in a row is fairly low then, assuming it has four moves. (I don't know, I didn't catch it.)
Probability-wise it's no less likely than any other permutation of moves, it just appears significant because it has more of a pattern to the human eye than, say, Screech-Sonicboom-Sonicboom-Screech-Sonicboom.
 
Well I mean like, if it had four moves, it would have a 25% chance for each. If screech was obtained the first time it would go on to a 12.5% chance of getting it twice in a row, 6.25% for three times in a row, and 3.125% for four times in a row, if my math is correct. (It probably isn't, I suck at math, but you get the point)
 
The odds of the same move being chosen over four consecutive turns, assuming equal chance of selecting each of four moves every turn, is .25^4, or roughly .004% (1 in 256). It's just no more likely than, say, Screech-Sonic Boom-Screech-Whatever Other Move it Has.
 
The odds of the same move being chosen over four consecutive turns, assuming equal chance of selecting each of four moves every turn, is .25^4, or roughly .004% (1 in 256). It's just no more likely than, say, Screech-Sonic Boom-Screech-Whatever Other Move it Has.
Haha I was way off. I suppose the point of this is that the chance it would have chosen such a bad sequence of moves is quite low.

Edit: More specifically, sonicboom four times is just as likely and the pokemon I was using only had like 60hp, so it would have taken it out in four turns rather than just die in 5, doing only 20hp damage. (sonicboom always does 20)
 
To give a better mathematical explanation:

(Assuming a wild Pokemon has 4 moves and all of them have enough PP to not skew chances)

- At any given turn, the wild Pokemon "AI" has a 25% chance of picking any move. Each chance of picking a move is independent from the previous pick.
- Now, if you want to know the chance of picking the same move move N times in a row, then the chance is not 25% because you are forcing a result. The chance is 0.25^N. (When you want the chance of multiple things happening all together, you multiply their probabilities)

The chance of picking the same move 4 turns in a row is 0.25^4 = 0,00390625. This is, a 0.39% chance of it happening.

(This is the explanation behind the gambler's fallacy. Just because each move has the same chance to be picked each turn, it does not mean they will be picked the same amount of times in a given amount of turns)
 
To give a better mathematical explanation:

(Assuming a wild Pokemon has 4 moves and all of them have enough PP to not skew chances)

- At any given turn, the wild Pokemon "AI" has a 25% chance of picking any move. Each chance of picking a move is independent from the previous pick.
- Now, if you want to know the chance of picking the same move move N times in a row, then the chance is not 25% because you are forcing a result. The chance is 0.25^N. (When you want the chance of multiple things happening all together, you multiply their probabilities)

The chance of picking the same move 4 turns in a row is 0.25^4 = 0,00390625. This is, a 0.39% chance of it happening.

(This is the explanation behind the gambler's fallacy. Just because each move has the same chance to be picked each turn, it does not mean they will be picked the same amount of times in a given amount of turns)
Interesting. (Why is everyone better at math than me? TT)
 
Sheer Force Hariyama uses Fake Out. At the end of the turn, its Flame Orb Burns it.

Battle Maison is so silly.
Granted, the Maison Pokemon's abilities are random... which makes me wonder why they can't lock certain abilities to certain Pokemon sets. If you're going to program in a Hariyama holding a Flame Orb, can't you also ensure its ability is Guts to take advantage of the burn?
 
Granted, the Maison Pokemon's abilities are random... which makes me wonder why they can't lock certain abilities to certain Pokemon sets. If you're going to program in a Hariyama holding a Flame Orb, can't you also ensure its ability is Guts to take advantage of the burn?
Yeah I guess it would take extra memory to lock the ability. Although, all regular trainers always have the first ability on their Pokémon, so they could have just built all the sets around that.
 
An instance in Super Singles Battle Tree:

I faced Veteran Priya carrying a Raikou-1 when my Tapu Lele was paralyzed. Having taken damage earlier, I was planning to weaken the Raikou so my Mega Salamence could come in for the kill. So Raikou uses Shadow Ball and it looks to be a 2HKO, and I just decide to leave my Lele to die to it.

Then it uses Z-Thunderbolt. A move which would've screwed me over because:
252 SpA Raikou Gigavolt Havoc (175 BP) vs. 0 HP / 0 SpD Salamence-Mega: 286-337 (86.4 - 101.8%) -- 12.5% chance to OHKO
Key word: would've.

Needless to say, I won.
 
Last edited:

cityscapes

Take care of yourself.
is a Tiering Contributoris a Community Contributor Alumnus
UPDATE: Adding another one from the Battle Tree.

Opponents using Thunder Wave when Trick Room was active.
This happens to me too. I have this Trick Room FEAR Mimikyu team and for some reason, on the turn after I use Trick Room, some opponents will choose to go for Thunder Wave. Against a Level 1 Mimikyu. They could OHKO the thing with Gust.

I...I don't get it.
 
Desolate Land is supposed to evaporate any Water type attack. Apparently, nobody told the NPCs in Alpha Sapphire, because every time I switch in my Groudon, the opponent's Kyogre will keep on using Origin Pulse. I predicted the NPC would continue to use it, so I used Rock Polish to outspeed.
 
Against Totem Wishiwashi, I used my Psyduck, because it resisted Water and had HP Electric. Psyduck had Cloud Nine. Totem Wishiwashi used Rain Dance every single turn for several turns while I slowly killed it. I guess this is because the AI acted like there was no weather (which makes sense unless it's deciding whether to use a weather-causing move).
 

Users Who Are Viewing This Thread (Users: 1, Guests: 0)

Top