- Greetings, Pokefans, Michael here, and today, I'm going to
be reacting to a video that definitively determines which trainer in Pokemon Red is the best. This video is the Pokemon
Red slash Blue AI tournament, pitting all the NPC trainers
up against each other to figure out which one is the best. The video is by Piman Rules, and a huge thank you to them for giving me permission to
make this reaction video, make sure to subscribe to their channel and watch the original video, both linked in the description below. And of course, don't forget to subscribe to my
channel, that'd be great. But anyways, that covers
all the intro stuff, let's dive in. All right, here we go. - [Piman] The trainers in
Pokemon Red are so weak that they can be easily
bested by a seven year old. - I don't know about easily. I personally remember Pokemon
Ruby took me some tries, not like restarting, but like I really had to over-level
to beat Brawly, anyways. - [Piman] Or an army of Twitch users. - There we go.
- But maybe it's unfair to judge them by human standards.
- Twitch Plays Pokemon. - [Piman] What would
happen if we would were pit all the AI trainers in
Pokemon Red against each other in a giant tournament? - I can't wait to see who wins, I feel like the champion rival is the most likely to win. Side note, this visual effect, very cool. No idea how they did it,
but it looks awesome. - [Piman] Who would come
out as the strongest of all and who would be the surprise underdogs? That was the question I set out to answer in my newest project, Scientifically Ranking
the Pokemon Red Trainers. - Ooh.
- By the way, this video is broken up into several parts, if some of them don't interest you.
- Well, I'm gonna watch all of 'em, it's fine.
- Feel free to skip around, there are convenient
timestamps in the description. First, let's talk briefly about how the AI in the Gen I Pokemon games actually works. - I'm interested to hear about this 'cause I don't know crap. (laughing) I may have gotten an engineering degree, but I didn't do much coding, except for that one class, CHEM 210, just no Sundays, anyways. - [Piman] There are 391
trainers coded in the game, and they all follow the same basic AI. Each turn, they start by picking a move, each of the up to four
moves they can choose from starts with the priority value of 10. - Oh.
- The lower the number, the higher the priority. - Oh.
- Then based on the trainer's class, up
to three modifications are applied to these priority values. The first modification
strongly deprioritizes moves that only inflict a status element if the opposing Pokemon
already has a status. Almost all trainers use this modification. - That would make a lot of sense, this is really cool actually, wait, I'm learning so much already, it's only been a minute-19. - [Piman] The second modification slightly prioritizes moves
that have certain effects, such as healing or raising stats. This modification is supposed to apply only on the second turn a Pokemon is out. - Oh.
- But, well, we'll get back to that. - Foreshadowing.
- Only a select few trainers follow this modification. - Oh, wait, go back,
what are the trainers? - We'll get back to that, only a select. - Pokemaniacs, Super
Nerds, Jugglers, Psychics, Chiefs, Scientists, Gentlemen, and Lorelei like their secondary effects, interesting. Lorelei makes sense, making
her go for the ice moves, maybe to get a freeze, potentially. - [Piman] Few trainers
follow this modification. The third and final modification slightly prioritizes moves
that are super effective. However, this calculation
is a bit primitive, it doesn't care whether the
move actually does damage. - I know about this, I know about this. That's how the ATV, All Terrain Venomoth, in Twitch Plays Pokemon was
able to beat Lance's Dragonite, 'cause it new Barrier. It's like, oh, psychic's super effective on Venomoth's poison, but Barrier doesn't do damage, so it just spams Barrier
instead of attacking. - [Piman] And it doesn't
bother checking both types of a dual-type Pokemon. This leads to interesting quirks, like the trainers in Celadon Gym that won't try to hurt you if you come at them with a Bulbasaur because they just spam poison. - Oh.
- The third modification is used by a larger handful of trainers. Once the appropriate
modifications have been applied, the AI picks randomly between the moves with the highest priority. Of course, using moves isn't
the only thing trainers can do, a handful of trainers have
additional AI routines that allow them to perform actions like using items or
switching out their Pokemon. - Looks like it's listing mainly
the gym leaders. (laughing) - [Piman] These also
correlate to trainer class, and they usually follow logic like if health is below 1/5, then with 50% probability, use a Super Potion, otherwise, use the selected move.
- Ah. Only 50% probability, that's fascinating. I feel like later generations, after it reaches a certain point, using it's guaranteed, maybe. Can I just say, Piman is
doing an excellent job of explaining this. - [Piman] These actions can be done a limited number of times, which resets when switching out. So then, we know how the AI works, but how are we going to
pit them against each? - This visual effect is so cool. - [Piman] Other, well, since the AI is so simple, it would be easy enough to
just implement in Python. Then we could wire. - Easy enough in Python, you know, that software that I've
never once used. (laughing) - [Piman] It up to a locally running instance of Pokemon Showdown,
set to the Gen I rule set. But this doesn't feel entirely
in the spirit of things, I want to see how strong
these trainers are in their home turf, and that means actually
running Pokemon Red. So here's how the system works, the basic setup is that
we have two emulators that are kept in sync, but
mirrored from one another. - (gasping) The drama.
- So say we're pitting Blaine. - [Blaine] When you're hot, you're hot. - [Piman] Against Misty. - Excuse me, (laughing) "When you're hot, you're hot." Is that a line from the anime? That caught me off guard, play. - [Misty] Misty, Misty. - Oh, I know that reference, that's from the old Pokemon spoof video, where it's like, you know, with like badly drawn cartoons, you know, "They're walking faster
than us, he's a clever one," you know, "I want Agumon," "That's a Digimon, Ash." Yeah, it's Trash Animation, I think. (upbeat digital music) - [Piman] When one emulator
will set up a save state where the opposing trainer is Misty and the player has a party
identical to Blaine's. Then whenever the player
needs to make a choice, we take a save state, swap the data for the
player and the enemy, then load that save state
in the second emulator. From there, we let the battle play out, and see what action Blaine, who's now controlled by
the game's AI, takes. - That's so cool.
- Then it's a simple matter of pressing the right
buttons in the original game to perform that action. This does unfortunately
still leave a few quirks because of the way the game
treats the player trainer versus enemy trainers. For example, player Pokemon
gain XP if they defeat an enemy and use up PP when using moves. These are easy enough to disable though by just hard coding memory values. But even with these accounted for, there still exists two subtle differences. The first has to do with priority, the player is always able to switch out or use an item before the enemy moves, while enemy trainer actions get priority based off the speed of their Pokemon and the priority of the move
that they would have used had they not instead taken an action. This means the enemy trainer
has to win the speed check or select a move like Quick Attack in order to take actions
before the player moves. Second, some AI trainers. - That's wild. (laughing) So that means sometimes they can like heal before or after you attack? I don't understand this, that's crazy. - [Piman] Will try healing their Pokemon even when they're at full health, thus wasting a turn. However, the game won't let the player do this to their own Pokemon. These quirks hopefully won't
have too much of an impact on the final results. - Aha, it is I, Grunty Boy. - No, no, I am not
dealing with this today. - As if you could escape me, what are you doing? - Drowning you out using
my earbuds from Raycon, the sponsor of today's video. - Oh yeah, Raycon, they
make fabulous earbuds. Their premium wireless
audio for half the price, available in a range of colors,
patterns, and fit options. They have six hours of playtime, seamless Bluetooth pairing, more bass and more compact design for a comfortable noise-isolating fit. And I know all this because I have my own, you know, that I stole from you. (upbeat music) Yeah, I'm really enjoying using them. I love listening to my
music and podcasts with them when I'm out and about, you know, committing crimes, since the lack of dangling wires, lets them not get in my way and keeps me from having
to look at a screen. I am obsessed with them, you know, just like other celebrities
like Snoop Dogg or whatever, and I must say, I'm
enjoying them even more knowing that I stole them from you. (upbeat music) You're not the hottest PokeTuber. - Mc-xcuse me? - Aha, gotcha, now you
have to listen to me taunt you about having
stolen your Raycons. - Oh, is that all you were saying? Well, that doesn't bother me, Raycon prioritizes customer
experience start to finish and has a 45-day return policy, so my buying experience was very pleasant. Plus anyone can get 15% off by going to buyraycon.com/mandjtv, linked in the description below. - Oh, well, darn. I guess we'll both just
enjoy our Raycon earbuds now. - I will certainly enjoy mine, but anyways, why did you come here? - Just to annoy you, I wasn't able to get you
with the theft taunting, but I gotta say, I did get you by telling you you're not
the hottest PokeTuber. - Eh, not really, I was just shocked at such an outlandishly false statement. - Fair, well, I guess I'll have
to get you next time, ta-ta. - Well, all right, anyways, thanks so much to Raycon for sponsoring, but now back to the video - [Piman] With everything wired up and all the major bugs ironed out, we're ready to run the tournament. I pitted trainer against trainer until everyone had a chance to fight everyone else.
- This looks so cool. - [Piman] In a true round-robin style, plus a few battles on top
of that for good measure. In total, I ran over 87,000 battles, that's about 4,000 hours worth of battles, and collected statistics about
each trainer's performance. - I really hope you sped
up the game. (laughing) - [Piman] Which we'll take
a look at in just a bit. The most basic statistic is the win-loss-draw ratio. You might be wondering
how draws can happen in non-competitive Pokemon, even if both Pokemon
faint at the same time, the game will still declare a winner. Well, because the AI is so primitive and is unhindered by PP constraints, there are lots of scenarios where two AI trainers will get themselves locked
in an endless battle. Because of this, I instituted two draw scenarios.
- Oh. - [Piman] The 75-turn rule, inspired by the similar rule in chess, counts the battle as a draw if 75 turns go by without
either side taking damage. - That's a good, wait, taking damage? - [Piman] The 1,000-turn rule
counts the battle as a draw if 1,000 turns go by
without either side winning. - (laughing) Oh God, that sounds miserable. (laughing) Oh wow, that's wild. Man, the amount of coding that went into this is really impressive. - [Piman] I also count the match as a draw if the emulator freezes or if some other
unaccounted-for bug occurs, but fortunately, that didn't
happen too many times. - Okay, that's nice. - [Piman] The second statistic
I calculated is an Elo score. This should be familiar to
anyone who plays chess or. - No.
- Pokemon Showdown, as well as a host of other games. It's very similar to the MMR systems used in a variety of competitive games. Basically, every player has a score which goes up when they win and down when they lose.
- Oh, okay. - [Piman] The clever bit is that the amount the score changes varies based on the score differential between the two competitors. So for example, you gain more points for beating a significantly
stronger opponent, but likewise, lose more
for losing to an underdog. - That makes sense, okay.
- This makes Elo a better overall indicator of
skill than the win ratio because it accounts for the
skill of your opponents. It's not necessarily the most well suited for this type of tournament though, because the combatants are static AI. Elo is designed to capture the change in a person's performance over time, and thus can potentially
change dramatically based off the order of matchups. - Oh yeah, that's a good point, how does he account for this? 'Cause it seemingly,
like the order of battles is like not really tracked, it's just the total performance. - [Piman] Despite this downside, I chose to use it for this project. - Oh.
- Because I hoped it would be familiar to a lot of people and provide an easily sortable
score for each trainer. - I also feel like just the
sheer number of battles, like he was saying that it would change depending on the order you did, but I feel like the sheer number
probably accounts for that, like it's 87,000 total battles, like I feel like that would
make things average out. Well, I hope. (laughing) - [Piman] Mainly though,
I just wanted to steal Tom7's excellent name, Elo World. (Michael laughing)
Of course, kids these days don't care about stinky math and numbers, they like tier lists. - I do like tier lists.
- So I also categorized each trainer into a tier, from F to S. And because I like to pretend
to be a data scientist. (Michael laughing)
I decided to try a mathematical approach to tiering. Since we already have an
Elo score for each trainer, dividing them into tiers is basically just a data science problem of dividing up a sorted list of numbers. A kernel density estimation is a good way of doing that, or so. - All right, this is starting to get out of my familiarity range. Ugh, who am I kidding, it's been there. - [Piman] I'm told, which
results in a graph like this. The local minima of this graph represents places in the Elo scale that have a low density of trainers, in other words, natural
boundaries between tiers. Of course, that's way too many divisions, so we can tweak the
bandwidth of the kernel until we find a number of divisions that feels appropriate for a tier list. Then, like any good data scientist, we can look at the beautiful
mathematically perfect numbers our calculation produced, decide they don't feel right, and just kind of fudge
it until we like it. (Michael laughing)
The result of that fudging is the tier you'll see as we scroll through the list of trainers. - (laughing) I love that, (laughing) do all this stuff, eh. (laughing) Fun fact, I was recruited
to like apply and interview for a data science company like
two years after graduating, and I'm like, "I am a YouTuber, go away." - [Piman] Finally, here you'll see each trainer's greatest
victory and defeat. - Nice.
- That is the highest Elo trainer they won against and the lowest Elo
trainer they lost against. So without further ado,
let's start the ranking. - Love this music.
- In very last place, squarely in F tier, we have the first rival
battle against Squirtle. - Obviously, (laughing) that is the lowest trainer
in the game, isn't it, maybe? Ah, actually no, now
that I think about it, like a Youngster with a Pidgey or Rattata, that's interesting. Also again, the visual effects, I don't know how he did
it, these are incredible. Is this done in like Blender or something? I don't know. - [Piman] You fight
the rival several times throughout the game, of course. And the rival's team varies based off what starter you picked, so the rival ends up accounting for 24 of the 391 trainers in the list. Anyway, with a team consisting of a single level five
Pokemon, it's not surprising that the rival finds
himself in last place. What's mildly surprising however, is the fact that the
rival with the Bulbasaur so outperformed the other two. - Bulbasaur is the best Kanto starter, that's not surprising to me, there's a lot of rock types
and a lot of water type. - [Piman] Earning himself
a spot in D-minus tier. This is probably because of
Bulbasaur's unique dual typing. - Mm, I see.
- Making our way through D-minus tier and up to D tier, we find a plethora of Bug Catchers and the second rival
encounter from Viridian City. This time, the Squirtle rival is able to join the Bulbasaur in D tier, thanks to learning Bubble, but Charmander continues to lag behind. D tier is pretty big
compared to the first two and is home to a number
of unremarkable trainers. - This is so cool. These visuals are just,
how long did this take? I'm just blown away by this editing. - [Piman] However, it is also home to the biggest upset victory in the game. - Oh.
- Rocker Number Two, with an Elo of just 707, managed to beat the B-plus
tier's Swimmer Number 13, with an Elo of 1867. - Well, I mean, that makes sense, he's got two electric
types up against a Starmie that's less than 10 levels higher, that doesn't surprise me
now that I see what it was. Cool to know about the upset though, like good for you, Rocker guy,
you suck most of the time, maybe 'cause you keep
blowing up your Pokemon. - [Piman] Looking at their teams, it should be clear why, the Rocker has a major type
advantage over the Swimmer. - Yep.
- Obviously, he just overwhelmed the poorest Swimmer with a torrent of super
effective electric-type, oh. - Oh.
- Well, that does technically count as a loss for the player trainer. - Oh, oh, he just blew up. (laughing) - [Piman] Well done, Rocker Number Two. D tier is also home to
the first gym leader, Pewter City's own, Brock. Looking at the trainers
around him in the ranking, it seems he's only slightly more powerful than the other trainers near Pewter City. - Oh, yeah, wait, weird that this Hiker right above him has more Pokemon, yeah, but they're weaker and
somehow did a bit better. That's interesting. (upbeat digital music) You can see the green bars kinda grow, well, sort of.
- In C-minus tier, we find the Green fight
from Cerulean City. As you can see, Bulbasaur
is maintaining its lead in the rival starter rankings. - Makes sense. - [Piman] Near the top of C-plus tier, we find a trainer with a
rather unique strategy, the Juggler. The Juggler is one of the few classes that will actually
switch out their Pokemon. Female Cool trainers, for example, will do this with some forethought, they'll have a chance of switching out if their Pokemon's HP is low. The Juggler is much simpler, they'll just switch out
every turn with a 25% chance. I believe this is what the.
- (laughing) Oh, oh, oh, 'cause it's a
juggler, that's genius. - [Piman] Kids call a big brain strategy. Unfortunately for Juggler
Number Five though, this didn't quite pay off, and we find a Victory
Road trainer down here with the SS Anne scrubs. Except, no, that analysis
is not entirely correct, this particular Juggler's failing had nothing to do with
switching out Pokemon too often, as you can see, he doesn't even
have a Pokemon to switch to. - Right.
- His problem is that his lone Mr. Mime has
only a single attacking move, Double Slap.
- Oh. - [Piman] Not the most
effective strategy either. - That makes sense. - [Piman] Anyway, just barely squeaking by in the B-minus tier, we find
the second gym leader, Misty. It's curious to see that
she's so much higher in the ranking than Brock, and substantially higher than the trainers around Cerulean City, around the middle of B-minus tier.
- Huh. - [Piman] We find Beauty Number 10, who is notable for partaking in the longest battle I caught on video, one of only four battles
in the entire tournament to end in a draw by the 1,000-turn limit. - Oh, my God, that's so many. (laughing) - [Piman] Her opponent
was Lass Number Five from back in D tier. With her level 29 Bulbasaur and Ivysaur, you'd think the Beauty
would be sure to win, but alas, Oddish is a grass-poison type.
(Michael laughing) So the Beauty just
spends the entire battle trying to use Poison
Powder on it to no avail. (Michael laughing)
Meanwhile, the Oddish is spamming Absorb, which seems to be missing pretty often for a 100% accuracy move. In fact, it's not missing, but instead dealing zero damage, which the Gen I games register as a miss. With neither side doing any damage, the battle would normally
end by the 75-turn rule, but once in awhile, Oddish is able to pull off a critical hit. (Michael laughing)
Dealing one damage at a time to Ivysaur and resetting
the 75-turn counter. - Oh my God, that's hilarious, that Oddish is just so bad. (laughing) I didn't know about the zero damage thing, I genuinely didn't know that. I guess in later gens, they made it so it will do one damage, but that's so funny. - [Piman] This battle went on
for an hour and 20 minutes, before being cut short
by the 1,000-turn rule. My condolences to the Lass, she surely would've won
if it weren't for that, eventually anyway.
(Michael laughing) In B-minus tier, we also
find Lieutenant Surge. - There he is.
- Perhaps surprisingly not dramatically higher
in the ranking than Misty. - Yeah, Pokemon are just a bit stronger. I love the music choice. Oh, we're really going, we are going. - [Piman] In B-plus tier, we find the Karate Master
of the Fighting Dojo. - Ooh.
- You know, B plus seems a bit low for someone vying for the title of Saffron City gym leader.
- Oh yeah, he's lower than the Scientist.
- Based on the ranking, he falls somewhere between
Lieutenant Surge in B minus, and Erika in A minus. His competitor, Sabrina,
is nowhere to be seen. Well, we'll get back to Sabrina. - Ooh, Sabrina's gonna be good. - [Piman] Speaking of
Erika, here she is now. A minus is a pretty respectable rank for the Celadon City gym leader.
- Yeah, look at that win ratio. - [Piman] But what's this? Oh no, poor Lorelei, who would've?
- Wait, look at those draws, why are there so many draws for Lorelei? Why is she only an A minus despite being an Elite Four
member, what is going on? She's only one spot above
the rival with 20 Pokemon, level 20-something Pokemon, what? - [Piman] Thought we'd see
a member of the Elite Four all the way down here? Battles against Lorelei
follow a predictable pattern, she opens with Dewgong, who then just spams Rest
for the entire battle. This strategy naturally leads to Lorelei having by far the highest
draw count of all trainers, over 1/3 of all draws in the tournament are attributable to Lorelei. But why does she spam Rest so much at the expense of anything else? - Yeah.
- Well, remember what I said earlier about the AI prioritizing healing moves on the second turn a Pokemon is in battle, well, funny story, I
realized at the last minute while writing this script that there was actually a bug in my code, it never updates the enemy turn counter. - Oh.
- So presumably, Lorelei is just stuck prioritizing Rest for the entire battle, when actually, she should only do it for the second turn, (laughing) oops. Or so I thought, or so I thought, but no, actually, if you look at the base save state I'm using, the enemy turn counter
isn't stuck on turn two, it's stuck on turn one. So actually, Lorelei isn't
over prioritizing Rest, she's under prioritizing it. And if you look at the
teams she drew against, you can actually see why
she used Rest so much, almost all of them have a
fighting or poison-type Pokemon. - Oh.
- And Rest is a psychic-type move. So while my code did have a bug in it, it was really to her benefit and Lorelei's poor
performance in this tournament is entirely on her, anyway. - That was a roller coaster, my God, I loved that. (laughing) I was like, "Oh, this crazy thing, oh, it's 'cause he made a mistake, okay, I mean, that's fair. Wait, no, it's because the Gen I AI is dumb with super effective moves. - [Piman] Unsurprisingly,
among the Pokemon Tower rivals, Ivysaur leads the pack. - It's so funny that like the trainers that have a psychic-type status move, like Rest or Barrier, can really be screwed over in this because of just the endless amount of poison types in Gen I, 'cause they wanna keep
using that psychic move that doesn't do damage. (laughing) Also, I should've mentioned this earlier, 'cause it just kicked in, but like I didn't know
that the AI trainers can never run out of PP in Gen I, they definitely can in later games 'cause you can stall it out, but I didn't know they
couldn't run out in Gen I. (upbeat digital music) - [Piman] Moving on to A tier, we predictably find
Koga, followed by Blaine, and the Viridian City, Giovanni, mixed in with a scattering
of Victory Road trainers and Silph Co. Rivals.
- Oh, Bruno. - [Piman] Once again, the
Venusaur rival maintains its lead. - Makes sense.
- Then kicking off A-plus tier, finally we find Sabrina. - So good.
- Sitting at a cool 2370 Elo, well ahead of all the other gym leaders. - Makes perfect sense.
- And even Bruno of the Elite Four. - I mean she'd crush Bruno, her psychic types are
so powerful, (laughing) so powerful, psychic's broken. - [Piman] Just a few spots shy of Agatha. I think Bulbapedia puts
it best when it says, "Psychic-type Pokemon
had virtually no match in the Gen I games."
- Yep, there you go. - [Piman] My condolences this time go out to the Saffron City
Karate Master, he really didn't stand a chance.
(Michael laughing) (Karate Master speaking
in foreign language) (Michael laughing)
Rounding out the Elite Four, we find Lance in the number eight slot. - Number eight, okay.
- At least the Elite Four has the decency to sort
themselves in the proper order. (Michael laughing)
Between him and the champion though, we find the rival fight from Route 22, once again, it's the Venusaur rival that pulls ahead of the pack to earn himself a spot in fifth place. - Interesting, he included
the unused trainer data for Professor Oak, and
that it lost to the rival, despite having Pokemon
of a far higher level. - [Piman] Then finally, we reach
the champion rival himself. In fourth place, with 2516 Elo, it's the Venusaur champion.
- Maybe he's not gonna talk about it.
- Who in a shocking upset was unable to maintain his otherwise consistent
lead in the very end. In third, with 2527, the Blastoise champion.
- Wait, only third. - [Piman] And finally, in second place, with an impressive 2534 Elo.
- Wait, who's first, is it the third?
- It's the Charizard champion. - Oh, it's the last Oak one, okay. - [Piman] Wait, second place? Yes, my friends, there is a trainer that beats out even the champion rival, standing alone in S tier.
- There it is. - [Piman] With 2559 Elo, the
undisputed strongest trainer in all of Pokemon Red is
Professor Oak with a Blastoise. - Interesting.
- Now, I won't blame you if you don't remember
the Professor Oak fight in Pokemon Red, it never actually appears in the game, but in the game's data
is a full set of rosters for three Professor Oak fights. In each case, using the
fully evolved starter left behind by you and your rival. This fight was probably meant to be a post-game true final boss that got cut at some point in development. It's a shame. - Agreed.
- Because the data shows Professor Oak would've been
a truly formidable opponent. Before we wrap up, there's
just one more fun graph I want to show you. - Oh, wait, but why were
the other two Professor Oaks weaker than the champions? I still have questions, how did the rival, it's Professor Oak, Blastoise, the three champion rivals and
then the two Professor Oaks, why were the two Professor
Oaks weaker than the champions? I'm so confused. - [Piman] So far, we've
been looking at trainers in order of their Elo, but we could also look at them in the order you fight
them in the actual game. - Oh.
- The x-axis in this graph shows the order each trainer is fought based off Bulbapedia's
walkthrough of Pokemon Red, and the Y-axis shows Elo.
- That is interesting. - [Piman] This graph is
a ton of fun to stare at, and there's lots of
interesting things to see. First off, it trends upwards, which is a pretty good sign that we're doing something
right with our methodology. (Michael laughing)
Moreover, all the gym leaders and rival fights stand out
above the rest of the pack, clearly indicating
their superior strength. - Yep.
- It's also easy to spot the trainers who hugely underperform, like Scientist Number Five from Silph Co., whose party consists of a single Electrode with Self-Destruct, or
Fishers Number Nine and 11, who have teams of just Magikarp. (Michael laughing)
Perhaps the most interesting thing about this graph is the way it levels off
towards the middle, this. - That makes sense to me, it makes sense that it levels off, because the progression of the Kanto games is just like, you reach a certain point, and then it's like, "All right,
you can go everywhere now." Like as soon as you get the Poke Flute, you've got half the fricking region, like that entire route
from Celadon to Fuchsia and from Lavender to Fuchsia, that's just so many trainers that they have to make
around the same strength because you can fight them at any, like you can go this way,
you could go this way, you can go this way and then this way, and it's just like, that makes sense, that's the leveling-off point. With like Fuchsia, look, Fuchsia's right in the
middle of that flat point, so that makes perfect sense to me. - [Piman] Corresponds
with the branching path in the middle of the game, where you can choose to fight
either Koga or Sabrina first. - Sorry, I stole his thunder. (laughing) - [Piman] People generally consider Koga to be the easier fight, and we can clearly see in the graph that choosing to fight Sabrina first results in a large difficulty spike. - Yep.
- It's interesting that the Bulbapedia editors chose this order for their walkthrough, perhaps someone should suggest a revision. - That was a fascinating video, I'm very impressed with the visual effects and the coding that was involved, and I still wanna know like what was the reason behind
the other two Professor Oaks being weaker than the champion? I don't know, I feel like another interesting
sequel that could be done could be taking the AIs of
the various trainer classes and giving them the same teams. So you're ranking it based, 'cause this ranking system ranks them based on the AI and the team, I think it'd be really cool to see like which trainer class, like which trainer
specifically has the best AI. It's probably still the champion, but you could just give all the trainers, like it'd be a lot easier 'cause there'd be a lot
fewer trainers to do, 'cause you'd like, just
one instance of each class. I think that would be really cool to see, but this was a fantastic
video, I loved it. Thanks again to Piman Rules for giving me permission
to make this video, make sure to go check out their channel linked in the description below, and thank you so much for watching, with an extra special thanks
to my patrons over on Patreon for helping support my channel independent of fluctuating
YouTube ad rates. If you wanna help support
me in the same way, the link is in the description below. Also, if you wanna check out
more of my fun Pokemon content, I recommend these videos here. All right, that's all I have for now, so it until next time, Pokefans, gotta catch them all.
