Unpopular opinions

[War Incarnate]

I feel like a lot of the discourse surrounding Pokemon boss design (with regards to the human trainers) would be solved if the AI was smart enough to figure out how to utilize switching.

Come on, we get so many fuckin advantages over the boss it's not funny, why can we switch at will while they can't? It's not fair, and I feel like it just kneecaps boss design to such an unfair extent because it presents such a large advantage that no matter what you do with the bosses themselves, their difficulty is eventually gonna be nullified.

This is, in my opinion, the fundamental flaw that brings down the in-game trainer fights, from the gym leaders to villainous bosses and the champions. That, and the lack of held items, good movesets, or (in the case of gym leaders) the over-emphasis on Monotype teams (if you're gonna do type specialists, be like Corbeau and have off-type mons that cover type weaknesses), but those are made much worse by the in-game trainers' complete inability to actually utilize switching as a mechanic.

 

[chroma_2903]

this is why we should return to totems #alolasupremacy

 

[Bakugames]

i agree npcs should switch more, though its good to be able to tune it correctly — ai will ALWAYS be exploitable, but overly switching can make them easy to trap, similar to how an ai that tries to use setup moves can become weak because they prioritize it over attacking.

 

[pika pal]

I feel like a lot of the discourse surrounding Pokemon boss design (with regards to the human trainers) would be solved if the AI was smart enough to figure out how to utilize switching.

Come on, we get so many fuckin advantages over the boss it's not funny, why can we switch at will while they can't? It's not fair, and I feel like it just kneecaps boss design to such an unfair extent because it presents such a large advantage that no matter what you do with the bosses themselves, their difficulty is eventually gonna be nullified.

This is, in my opinion, the fundamental flaw that brings down the in-game trainer fights, from the gym leaders to villainous bosses and the champions. That, and the lack of held items, good movesets, or (in the case of gym leaders) the over-emphasis on Monotype teams (if you're gonna do type specialists, be like Corbeau and have off-type mons that cover type weaknesses), but those are made much worse by the in-game trainers' complete inability to actually utilize switching as a mechanic.

This makes me think of how in the anime it's made an explicit rule that in Gyms, only the challenger can switch/substitute a Pokemon, and the write around is simply... the Gym Leader is that strong.

Like if Gym Leaders can't be smart, then just make them outlevel you (Edit: Or at least be way above the "local" level curve you'd grind to) so the player needs to rub braincells together to keep up. Type Match Ups should make it easier, not win the fight on its own. A Final Fantasy game, even an easy one, doesn't decide the fight because you used Thunder ona Water Enemy; often that's just expected and such bosses can hit harder to incentivize you to exploit such weaknesses.

 

[War Incarnate]

Like if Gym Leaders can't be smart, then just make them outlevel you so the player needs to rub braincells together to keep up.

This is not really plausible in-game without level caps, because otherwise nothing is stopping the player from grinding to the Gym Leader's new level.

At that point, it's not really a solution, because the same problem I described in my original post pops up to haunt once more.

...I guess you could do "[level of your highest leveled pokemon] + x" kind of level scaling (with a minimum level in case your guys are somehow below the intended level for the boss), but that might pose a different set of problems.

 

[Krockodile]

I agree that switch mode would make it more challenging; part of Legends ZA's difficulty is that it doesn't give us time to switch mons mid-fight, so we have to plan how to pivot safely.

If I'm offering my cents on why the AI wouldn't prioritise switching, maybe it's because it might prove more difficult than expected from GF's perspective. Later trainers at the end of the game will have higher IVs across all stats than the player's, and their potential hindering nature, not to mention the player might have a team of not-great mons they want to try out. The average casual player may not have all the information needed to understand the Pokémon battling system, such as Type Matchups and immunities, since the game emphasizes experimentation. I also think that having Switch mode on was fine because enemy trainers back then were higher levelled and had potions to heal their mons (you can too, but your mons were generally weaker in terms of stats). I imagine bosses like Cynthia's Garchomp, Ghetsis' Hydregion, and Professor Sada/Turo (latter two have phenomenal coverage) were already difficult in Switch mode; now imagine them in Set mode. These bosses force prediction and risk management, which Set mode amplifies.

That being said, if going above the level cap and the CPU's inability to switch mid-turn can be annoying, why don't we just limit ourselves by disabling our own ability to switch out (i.e., set mode)? Going over the level cap can be hard to avoid in newer games due to the forced exp share, but I do think having switch mode would teach us more about pivots and make it more challenging.

I get that it doesn't fix the issues with the games, and we don't get rewards for making self-imposed challenges, but I think, in the end, we players can try to make it enjoyable for ourselves, and sometimes the satisfaction is what we want.

 

[Ironmage]

It's always a bit funny to me when this comes up, because the AI can switch. The strong AI has been able to switch with a relatively sensible condition since at least gen 3. This is well-known to battle facility players. The reasons the AI doesn't switch under casual circumstances are because it often has low type diversity (so it doesn't think anything can safely switch in on the expected attack) and keeps getting OHKOd (the AI doesn't predict movesets, it needs to be attacked before it can attempt to switch).

At the moment, I'd settle for no longer forcing the ace to be last. An opportunity for a quip isn't worth giving the player the entire fight to prepare for what's supposed to be the biggest threat. Heck, to go back to the Cynthia discussion, Milotic's bulky enough that Garchomp switching into a second Thunderbolt aimed at it should be a real possibility in BDSP, but she's dumbed down in order to aura farm.

 

[Hugin]

(the AI doesn't predict movesets, it needs to be attacked before it can attempt to switch).

Honestly, that might be the easier fix. Just set whatever flag the AI uses to remember moves to default assume the player always has STAB. It would take some care to handle things like temporary typings(probably ignore them completely) and situations where the player has revealed 4 moves and none of them are STAB, but those are so niche that even if the AI doesn't handle them well, 99% of players will never notice. And it would let clever coverage(even if that coverage is only as clever as Ice Beam on your water starter) still feel useful.

Whereas everyone does notice when Crasher Wake leaves Gyarados in against Luxray, loses it, then sends in Quagsire and leaves it in when the player switches to Roserade.

 

[Volt-Ikazuchi]

AI is a mess. Now a hotter topic than ever!

Abandon all hope, ye who read this post, for it's actually getting worse over the generations.

Remember the useless raid allies in SwSh? The AI is all but fixated on any status moves it can use, especially healing.
It is possible for them to switch, as Ironmage explained, but that way is absolutely impractical.

However, they can and will use pivot moves to gtfo if they're available, so that's the best compromise we can get with the current system.

It is unfortunate though, Stadium 2's AI was very good compared to this, and it could switch unprompted.

 

[Bakugames]

my hot take is they should allow the ai to cheat

not to the degree of the rby ai, i think that was kinda stupid and punished the player for trying to use switching and smart positioning, but i think that giving the ai ways to either guess or know things it shouldnt would actually make the playfield more fair
even if we ignore the fact a player can just see what moves/pokemon an npc has, we have the upperhand for being able to deduce things and bait the ai or even just cheese it. i think letting the ai cheat would be a fun way to give it an ability to "predict". maybe if you use switching cheese to bait immunity, itll know your switch and use a move based on that. maybe the ai will roll a 50/50 and if it lands on it, itll Learn about your coverage moves etc.

 

[DaddyBuzzwole]

my hot take is they should allow the ai to cheat

not to the degree of the rby ai, i think that was kinda stupid and punished the player for trying to use switching and smart positioning, but i think that giving the ai ways to either guess or know things it shouldnt would actually make the playfield more fair
even if we ignore the fact a player can just see what moves/pokemon an npc has, we have the upperhand for being able to deduce things and bait the ai or even just cheese it. i think letting the ai cheat would be a fun way to give it an ability to "predict". maybe if you use switching cheese to bait immunity, itll know your switch and use a move based on that. maybe the ai will roll a 50/50 and if it lands on it, itll Learn about your coverage moves etc.

i also think that there should be one character every game that can just straight-up rby cheat. i'd reserve it for the final post-game superboss, like gsc red, emerald steven, bw cynthia, etc. a single fight each game where they do more than just turn the levels up and add good evs/ivs/items, target demographic be damned, just crank the difficulty level all the way up to "fuck you" and make me break a sweat for once

 

[Azery]

i also think that there should be one character every game that can just straight-up rby cheat. i'd reserve it for the final post-game superboss, like gsc red, emerald steven, bw cynthia, etc. a single fight each game where they do more than just turn the levels up and add good evs/ivs/items, target demographic be damned, just crank the difficulty level all the way up to "fuck you" and make me break a sweat for once

lowkey this is why volo is so memorable because, in a game/series that is almost to a fault fair to the player(which is kinda unusual for a jrpg lol), he just fucking cheats like crazy

 

[Volt-Ikazuchi]

my hot take is they should allow the ai to cheat

not to the degree of the rby ai, i think that was kinda stupid and punished the player for trying to use switching and smart positioning, but i think that giving the ai ways to either guess or know things it shouldnt would actually make the playfield more fair
even if we ignore the fact a player can just see what moves/pokemon an npc has, we have the upperhand for being able to deduce things and bait the ai or even just cheese it. i think letting the ai cheat would be a fun way to give it an ability to "predict". maybe if you use switching cheese to bait immunity, itll know your switch and use a move based on that. maybe the ai will roll a 50/50 and if it lands on it, itll Learn about your coverage moves etc.

Just having the game assume mons have STAB would do so much

 

[Bakugames]

Just having the game assume mons have STAB would do so much

itd be cool if the lower ai only just assumed stab, but higher ai automatically knew what stab moves you had and did calcs around that, because it can be easy to explain it as "high level trainers know pokemon movesets and what youd have available at your level/journey period"

 

[DaddyBuzzwole]

itd be cool if the lower ai only just assumed stab, but higher ai automatically knew what stab moves you had and did calcs around that, because it can be easy to explain it as "high level trainers know pokemon movesets and what youd have available at your level/journey period"

considering that this is basically how low-level vs high-level play works in competitive it would make perfect sense to translate it to in-universe ai. it would also mean you'd be able to legitimately use lures in-game, which would be really funny

 

[Lontra Assassina]

I feel like a lot of the discourse surrounding Pokemon boss design (with regards to the human trainers) would be solved if the AI was smart enough to figure out how to utilize switching.

Come on, we get so many fuckin advantages over the boss it's not funny, why can we switch at will while they can't? It's not fair, and I feel like it just kneecaps boss design to such an unfair extent because it presents such a large advantage that no matter what you do with the bosses themselves, their difficulty is eventually gonna be nullified.

This is, in my opinion, the fundamental flaw that brings down the in-game trainer fights, from the gym leaders to villainous bosses and the champions. That, and the lack of held items, good movesets, or (in the case of gym leaders) the over-emphasis on Monotype teams (if you're gonna do type specialists, be like Corbeau and have off-type mons that cover type weaknesses), but those are made much worse by the in-game trainers' complete inability to actually utilize switching as a mechanic.

While I agree that switching is underutilized as a mechanic, I believe the idea that the in-game bosses should be on equal footing to the player is a bit misguided. I wish I could find the quote proper, but I believe Aonuma said in an interview that Zelda puzzle design is supposed to make the player feel smart for solving it, not make the player smarter in general. Similarly, trainer battles in the main story aren't meant to make you a better player: they're teaching you the logic the game. Gym Leaders and the like don't switch out very often because that provides a more consistent experience so players can figure out the mechanics as they play. They're much more like boss phases in a more traditional RPG.

 

[Ravenson]

An interesting thing to add to this better AI discussion is: Let the gym leaders and Elite 4 members remember any Pokemon and moves you used against their underlings/earlier E4. Obviously a nightmare to code, but it hardly seems unreasonable to imagine Brock seeing what you use against that dweeb - and suddenly skipping gym trainers instead of pathologically fighting each and every one for that sweet, sweet EXP has the tiniest of reasons to recommend it (though of course the truly pathological will just leave the gym before fighting the leader and swap in an all-new team).

 

[Volt-Ikazuchi]

I wish I could find the quote proper, but I believe Aonuma said in an interview that Zelda puzzle design is supposed to make the player feel smart for solving it, not make the player smarter in general.

Aonuma also said the only reason people want Zelda to get closer to its robust dungeon design roots is because of nostalgia and he straight up doesn't understand why we'd want to go back to a non-fully open Zelda game.

Bro is washed beyond recognition.

And this kind of kiddie glove design is honestly too lame to entertain. I just can't cosign this outside of a potential easy mode.

 

[Bakugames]

i mean honestly id be ok w an easy mode in pokemon. if a kid struggles then give them an easier time so they can have fun. but i also want something more engaging too. note that ive never been talking about some insane hard mode, im ok with pokemon not being a hard series, rather i just want fights to be interesting and Fun and have some bite to them. Tate and liza werent insanely hard or anything, the totems arent souls bosses, but theyre all interesting design that got me to have fun anyway

 

[DaddyBuzzwole]

i mean honestly id be ok w an easy mode in pokemon. if a kid struggles then give them an easier time so they can have fun. but i also want something more engaging too. note that ive never been talking about some insane hard mode, im ok with pokemon not being a hard series, rather i just want fights to be interesting and Fun and have some bite to them. Tate and liza werent insanely hard or anything, the totems arent souls bosses, but theyre all interesting design that got me to have fun anyway

i say there should be an easy/medium/hard select screen at the beginning of the game. easy mode being in line with what we have now, medium being more like bw2 challenge mode/bdsp e4 rematches, and hard having all of these ai improvements we're talking about, good evs for boss trainers, maybe level caps and bans/limits on using items in important battles, etc. obviously this would require actual work so of course it will never happen, but a guy can dream

 

[sg pokemon]

(Especially considering the thread title) The highest level of AI knowledge should just let it know your pokemon's every detail tbh: all their exact raw stats (including multipliers from choice items/AV), actual ability, moveset and anything else of relevance; if you as the player can have (eventual) knowledge of all this stuff about the opponents' mons, why can't they do the same for you at their strongest point (especially if they're fighting under imperfect conditions, i.e. no full pkmn teams of 6/not full moveslots)? (I'd almost suggest also having them know your entire actual team in advance as well, but it feels bizarre and off to have them be aware of the pokemon that you haven't even revealed yet).
Also note that I suggest for them to have knowledge of your mons' entire data, not act in accordance to other unrelated things; i.e. have them make their decision after the player's input, aka input reading which is what the RBY AI does. Because I agree that's not only too "unnatural" but also just too much in general; if the AI can respond and counter perfectly to your every play, then one runs to risk of needing to counter-team the opponent in order to be able to make winning plays no matter what the AI tries to counter with; this is ofc, not particularly engaging or fun. However, there should still be some measure to prevent the player from just endlessly switch stalling around for the purpose of pp stalling the opponent(s) out of their most critical move(s) without much counterplay, as that too is not only long and tedious but also frankly, uninteractive and lame....yet you can get away with it just fine, if you know the AI's exact playing patterns (however I do in general far prefer an AI that makes actually consistent move decisions to maximize engagement/challenge level).

 

[DrPumpkinz]

(Especially considering the thread title) The highest level of AI knowledge should just let it know your pokemon's every detail tbh: all their exact raw stats (including multipliers from choice items/AV), actual ability, moveset and anything else of relevance; if you as the player can have (eventual) knowledge of all this stuff about the opponents' mons, why can't they do the same for you at their strongest point (especially if they're fighting under imperfect conditions, i.e. no full pkmn teams of 6/not full moveslots)? (I'd almost suggest also having them know your entire actual team in advance as well, but it feels bizarre and off to have them be aware of the pokemon that you haven't even revealed yet).
Also note that I suggest for them to have knowledge of your mons' entire data, not act in accordance to other unrelated things; i.e. have them make their decision after the player's input, aka input reading which is what the RBY AI does. Because I agree that's not only too "unnatural" but also just too much in general; if the AI can respond and counter perfectly to your every play, then one runs to risk of needing to counter-team the opponent in order to be able to make winning plays no matter what the AI tries to counter with; this is ofc, not particularly engaging or fun. However, there should still be some measure to prevent the player from just endlessly switch stalling around for the purpose of pp stalling the opponent(s) out of their most critical move(s) without much counterplay, as that too is not only long and tedious but also frankly, uninteractive and lame....yet you can get away with it just fine, if you know the AI's exact playing patterns (however I do in general far prefer an AI that makes actually consistent move decisions to maximize engagement/challenge level).

I'm pretty sure most games' Smart AI does actually read your Pokemon's actual stats. Like, the more basic versions will literally perform mock damage calculations of each of their moves on your Pokemon's real stats and will choose which attack to use accordingly.

 

[Bakugames]

yeah the ai is super good at calculating exact damage ranges of all their moves on your pokemon, to the point that you can exploit that to your advantage (though imo thats not something that necessarily needs fixing. its good to give openings to the players instead of making the ai play optimally 100% of the time).

i do think the actual team building and lack of switching and reactivity is the main problem. the player should not be punished for pivoting, but they should be punished for trying to cheese via swapping immunities. the player shouldnt be punished for finding an opening to boost, but they should be for spamming five billion swords dances agilities calm minds double dances and attempting to sweep.

pokemon bosses arent necessarily easier. put cynthia as she is in sv and shes five times easier to deal with because you can make extremely strong, fast teams with insane moves and her actual team lacks any synergy or scare factor beyond raw stats, making her very easy to overwhelm. theres not much different from her gastrodon or milotic to geetas gogoat and avalugg, but this isnt 2008 where moves were limited, the dex sucked and stats were small. gamefreak can make better teams that match the tools a modern player will have access to, without making a shitty hardmode fanhack esque team

 

[AztecCroc]

yeah the ai is super good at calculating exact damage ranges of all their moves on your pokemon, to the point that you can exploit that to your advantage (though imo thats not something that necessarily needs fixing. its good to give openings to the players instead of making the ai play optimally 100% of the time).

i do think the actual team building and lack of switching and reactivity is the main problem. the player should not be punished for pivoting, but they should be punished for trying to cheese via swapping immunities. the player shouldnt be punished for finding an opening to boost, but they should be for spamming five billion swords dances agilities calm minds double dances and attempting to sweep.

pokemon bosses arent necessarily easier. put cynthia as she is in sv and shes five times easier to deal with because you can make extremely strong, fast teams with insane moves and her actual team lacks any synergy or scare factor beyond raw stats, making her very easy to overwhelm. theres not much different from her gastrodon or milotic to geetas gogoat and avalugg, but this isnt 2008 where moves were limited, the dex sucked and stats were small. gamefreak can make better teams that match the tools a modern player will have access to, without making a shitty hardmode fanhack esque team

Milotic is significantly harder to come by than Gogoat or Avalugg. And much stronger in terms of BST.

 

[CPU]

Milotic is significantly harder to come by than Gogoat or Avalugg. And much stronger in terms of BST.

Milotic only edges out Gogoat's BST by 9 points (and Avalugg by 26), that's a fairly small gap all things considered. Definitely harder to obtain though.