How a Programming Error turned a Gym Trainer into one of Pokémon Speedrunning's Worst Fights 

You are actually underselling the difficulty of this fight. Hideki has semi-random AI; it can randomly choose any of the moves, unless they fail, then he will not pick the move. That being said, if it picks Foresight, it cannot pick Foresight again, and same for Focus Energy. The ideal fight is Foresight into Focus Energy into get range and kill. Machop will be picking Low Sweep way more often than you led on.

Doesn't seem like an error to me. Hideaki was just prepared to kick the player's ass.

We were all warned about the threat of a top percentage Rattata, but we didn't see the top percentage Machop coming.

it's hilarious to me how this trainer, in all normal circumstances, is nothing to worry about. just in speedruns.

Funny how a level 13 Machop gives Speed runners a bad time, but casual gamers just go to town with their Ralts and Tailow or their newly evolved Dustox/ Beautifly.

Let me shed some light on how this programming error came about. The developers assigned a "difficulty value" to each trainer. When read in 8-bit binary, it was intended that the LEFT 5 bits be read to determine the IVs, but instead the RIGHT 5 bits are read. We know this is a mistake, because it makes much more sense for IVs to get higher when you face more prominent trainers, but instead they seem pretty random. Here are some examples of how the error manifests: Black Belt Hideki, difficulty value 120 (01111000) - intended left 5 bits: 01111 (15 IVs), actual right 5 bits: 11000 (24 IVs) Leader Brawly, difficulty value 160 (10100000) - intended left 5 bits: 10100 (20 IVs), actual right 5 bits: 00000 (0 IVs) Elite Four Drake, difficulty value 200 (11001000) - intended left 5 bits: 11001 (25 IVs), actual right 5 bits: 01000 (8 IVs) Champion Steven, difficulty value 240 (11110000) - intended left 5 bits: 11110 (30 IVs), actual right 5 bits: 10000 (16 IVs)

Honestly single mon trainers having stronger stuff makes sense.

I know it's annoying to the speedrunners, but I love stuff like this. Just one random guy who didn't play by the "it's a videogame" rules and went out and busted his 28 across the board IVs Machop because he loved it and wanted to train it up. He truly believes that might is right!

"My Machop's in the top 1% of Machop!"

I've never been exposed to a Pokemon speedrun before, the idea of the optimal early-game strat being a Mudkip screaming endlessly is so funny to me

The guy was training his Machop all his life, pay respects.

Finally, a trainer who lives up to their intro dialogue. "Might is right!"

Might as well snap the player characters neck after they lose cause that save is absolutely getting deleted lmao

I love how a single random trainer just becomes a problem like this, all due to a funny error

Interesting enough, there’s that hiker spinner just outside of azalea town, with a machop, that WILL end your speed run if challenges you.

This kinda reminds me of one if the very silly questions about pokemons gameplay systems you can have. Pokemon breeding produces a lot of mons you dont use and will probably just dismiss from the box. If those pokemon get released into the wild, theres surely some where a kid gets lucky and picks up the equivalent of a god machop, thats only slight blemish is that they have the wrong nature or something. I honestly wouldnt be suprised if this was a legitimate reason that villain teams have so many of the same pokemon. Someone out there wanted the optimal version of that pokemon and left a horde of super bidoof in their wake, and all they had to do was bribe the guy releasing them to just give them to the team.

Hideki is truly the great equaliser, with both OR and AS runners sweating. He also keeps the runners humble, as casual players normally just sweep him due to extra grinding, having more 'mon and better team coverage, etc.

It's a mercy that this happens early on in the game. So ending a run isn't as devastating as it would be later.

Not an error, Machop just locked in

“i’m not mathematically literate” “sqRt (-1)” manz str8 imagining things

Man decided to train his Pokémon against human opponents. They aim for the balls, always. Sounds like it’s intentional.

99.99999999% of players, just beats him with no trouble and doesn't remember him. Speedrunners: ANYTHING BUT A SLIGHTLY BUFFED MACHOP

I’m obsessed with speedrun ending trainers because in any other kind of playthrough you blow past chumps like this and the second rival battle

Says "Might is right!" & has insane IVs. Lol. Anyway, neat analysis video! Thanks for uploading!

"Pokémon is for kids, it's sooooo easy" Hideki's Machop:

IVs are in 5-bit chunks within the pokemon's data.14 in binary is 01110. If Machop was supposed to have straight 15s, then its IV line would look like "011100111001110011100111001110" If this data was put in wrong so that the game 'clipped' the first bit off, you would instead have "11100111001110011100111001110?" that is, it would read five bits at a time and get 11100 as each IV except possibly the last (which I believe is Speed). As such, its IVs would come out as 28 across the board, with the possible exception of Speed which could be 28 or 29, depending on if the next bit after falling off the end of the data was 1 or 0. The mistake of putting the data in a single bit short, or having the data be read somehow a single bit off from where it was supposed to, suddenly makes the Machop extremely beefy. While I am not sure if these are the exact internal values being misread, the example does seem to suggest that such a mistake is entirely reasonable - if I had to guess, I would suggest that the pointer to the memory address that this specific Machop uses was typoed to be 1 more than it should be, so it starts reading memory one bit 'late'. If this were an older game that used bytecodes like the old NES and such, it's the kind of thing that could probably be fixed with a single Game Genie code, being the result of a single byte of ROM being altered - as it's a 3DS game, no idea.

From a wordbuildig perspective I like the idea of same "difficulty" trainers having some seemingly randomized variability between their values by, like, a couple points in either direction, and even of the occasional random trainer mon being a very well IV'd mon, just because, hey, that'd logically happen every now and again, other people can also get lucky. Heck, I's even accept IVs on trainer mons being variable by save file, personally. But, all that said, 100% see how what I've said above is increasing hell for speedrun doers.

For half of the video I wasn't paying much attention and thought he was saying Garchomp. Nah, this is just a Machop

Due to the law of large numbers, 50 fights is more than sufficient for you to get a statistically significant sample (30 or so would have done). That’s a brutal right nonetheless. 8:43

seeing mudkip so much is making me happy. I love him

Ya know, I swept him like nothing on a casual run. I did not realize he was crouching tiger, hidden menace like this.

For what it's worth, 50 is a perfectly acceptable sample size for this! Damn you, jacked Chop.

Steven's Cradily anc this single Machop were the bane of my Swampert only run. Before you explained the Machop, I knew what was coming...

I thought the title was implying that the programming error turns the gym trainer into one of the Pokémon 🐸

Seems "ain't nothing but a funky beat-down" is at play with this Hideki..

This is amazing timing. I just started a semi blind run of Omega Ruby. Haven't played this in ages, relying on vague memory and general pokemon knowledge. I don't know wtf I'm walking into. This dude almost wiped me, and took down two of my team out of nowhere.

I like to think Hideaki was intentionally designed that way by a dev who just hates speedrunners. I doubt that's true, but that's a very funny conspiracy in my head.

I like the topic and your presentation but I felt like overall the video was missing some impact without an in depth explanation of either 1) the mechanics of why this Machop is so jacked (I know you said you don't understand them but I still want to know) or 2) what time concessions runners have had to make to deal with the machop; conversely, how the run could be faster or more consistent without the Machop Looking forward to the next one!

I appreciate how engaged you are with your community. Probably something that won't last as you grow, but it's still really nice to see since even a lot of the smaller youtubers I've seen barely interact with their viewers. Also you're doing a superb job with your content, can't wait to see more.

I don’t know much about ORAS speedrunning, but are any items used during the run? I was thinking either an Oran berry to increase survivability against Low Sweep, or grabbing the Silk Scarf in Dewford to give a 20% boost to Echoed Voice, decreasing the risk of a 4HKO requirement.

You really upsell that 28 IV, but it only raises its stats by 1-2 points. Brawleys level 14 Machop is more powerful stat-wise.

If there's ever an Emerald remake, Hideki would 100% be the new gym leader.

Just found your channel! Can't believe you only have less than 5k subs your content is so good! Definitely subbing and looking forward to the future of your channel!

Along similar lines of stuff with the wrong IVs, Volkner's Electivire in Platinum *would* have had 303 IVs in each stat because of a similar typo, but the game's smart enough to check for values outside the 0-31 range and rerolls, resulting in a weird spread of 4/3/29/18/15/19.

It's almost as though this fight wasn't meant to be fought with just a low level mudkip.

50 is a totally reasonable sample size. The lowest sample size to get an expectable margin of error is 30. In most statistical tests, this is sufficient enough to make a conclusion with 95% confidence, or 5% error. Of course in the world of Pokemon, this may as well be 50/50 (Looking at you, Heat Wave).

Great video!

I don't speedrun games, but this was interesting to hear about. Thanks for the video. I have done a couple Nuzlockes, including Alpha Sapphire, but I don't recall Hideki being too bad. Guess it's different when you have time to train and prepare, unlike a speedrun.

I would love more trainers (preferably optional ones) with optimized pokemon. Just for that little extra challenge.

You got blessed by the algorithm!

Outstanding! 🔥🔥🔥

Love this 💯💯💯

I got an idea for a rom hack of ORAS: Make the Machop have 0 IVs in all stats to get revenge

its a gym trainer being fought on-level without a type advantage or a type that negates fighting altogether and people are shocked that the gym trainer can ko a pokemon in 3-5 turns? there's a reason most speedruns are using the type system to completely negate danger or finishing gym pokemon off in 1-2 turns if you go in to an even match without any advantages on these gym trainers that's what your experience will be its just a coincidence this is the one in or/as which sticks out early enough that players don't have their mudkip 1shot engine online

Wow i never knew that thanks mr tortoise

You know what, I did actually get messed up by this guy in a casual run. I didn't really think much of it at the time, but it makes a ton of sense hearing it like this.

I think this video really needed a comparison to an “intended” machop to really sell the problem. Because to my understanding, the problem really isn’t because of the programming error, but more so the moveset, and the error only makes it worse. But the video doesn’t spend enough time on the error itself or how those stat changes percolate to making the fight truly an abnormal RNG check because of the error. The framing of the narrative is just a little off and I feel like you give too much credit to the IVs without explaining the impact the IVs have on the damage rolls. Still a fun video, learned something new from both the videos and the comments.

Great video!! Just wonna put it outt here that I think putting "a singular machop" instead of "gym trainer" in the title is a much bigger draw to the video

something so funny about a trainer with such a powerful pokémon that looks like an alien baby

neat video. subscribed

To say that this was a mistake implies that speedruns are of any interest to the developers, which I don't think is true

me with my main ace ralts in oras against hideki: hold my beer

You looking at the machop and then to your mudkip alright "objective, survive!"

Clearly in the next pokemon game Hideki should show up as a newly minted gym leader with an overpowered Machamp.

I always just walled the second gym in ORAS with a quad resist to fighting types like zubat, or the wurmple evolutions, so that machop is virtually nothing to me. It's Brawly's meditite with STAB confusion that gets me since it's such good coverage

great video

What if GameFreak is just trolling speedrunners by this point? The ""option to skip cut-scenes"" in Let's Go already made me think that.

this sounds like an intended wall for an actual challenge during any run.

I mean, I assume you have to know what you’re doing to become a gym trainer. Good to see someone taking this job seriously.

This was a neat video

Haha no if you bring Zubat or Ralts during a casual playthrough Hideki is just like butter.

I don’t recall ever having any trouble with this particular Machop. I typically used a Combusken that knew Peck and Power Up Punch and had Speed Boost. I didn’t even really need to use Peck as three turns was plenty of time to make the Attack stat go higher via Power Up Punch and Speed Boost allowed you to outspeed most Pokemon.

given the 30/50 failure rate, the standard deviation is 0.07, and as such you can be 95% confident that the true failure rate is between 0.46 and 0.74 Generally 30 samples is the minimum you’d want for statistics, 50 is fine, more is always better “statistically significant” is also usually used when the 95% confidence range does not include the expected case (ie the results are statistically significant and indicate a deviation from normal), there’s not really a usual case (null hypothesis) here tho

You make good music and good videos. Glad the algorhythm picked you up judging from your highly varying view counts

I'd love to know how other gym fights stack up to their casual run counterparts. Whitney's Miltank comes to mind.

Just my personal feedback… it would’ve been really great to get even a short/simple 10 second explanation as to what happened that affected ONLY this trainer’s IVs. That’s super interesting to me and would’ve been beneficial Also, it’s one thing to say, “these IVs make the fight so much harder”, but another to explain why. Even a few quick damage calcs to show the difference would’ve been really great It was a good video and I enjoyed it for sure!!! Just wanna give feedback because I care about it and want to look forward to seeing more in the future :)

I’m glad the tradition of coding errors making Pokémon crazy difficult has continued into the modern age! Great video.

I actually kinda love this. Running into a random badass with a single freakishly strong Pokémon who wrecks you before you even fight the actual boss is the sort of thing you see in an anime, meant to teach the protagonist how big the world they've entered really is or something like that. They'd probably go on to fight the gym leader and be like: "This isn't so bad compared to that other guy." The fact that it was caused accidentally by a coding error somehow just makes it better: It's like how MissingNo was just a glitch but became memed as some sort of eldritch, reality-corrupting entity.

ORAS is still my favorite. It’s everything I imagined Sapphire to be like if it had 3D models and all the cool immersion stuff i dreamed of as a kid, like petting your pokemon and your bond with it MATTERING.

Funnily enough I remembered this guy from a challenge run long ago that had a distinct lack of fighting counters, good to know i wasn't going insane and bros actually stacked

Sounds like a skill issue. Chad Machop kicking ass.

Cynthia to her Garchomp: Look what they need to mimic a fraction of our power

Hideki isn't bugged, he just wakes up and chooses violence!

Hey the linked videos explaining runners feeling about omega ruby and alpha sapphire are appreciated.

Good throwback about breathiing

I wonder if there is a world where catching/evolving exactly a shedninja is correct to help get past hideki?

Then again, the second (mandatory) rival fight in Emerald destroys me even when I'm casually playing through the game.

I thought my Machop was in the top percentage of Machop

I don't remember having any trouble with Hideki when I played through the Fighting-Type Gym in OR/AS

In the original pokemon blue of my youth - I recall the macho - guy being on of the few specific foes I had to grind in order to beat. That would predate this bug. Honestly though, I recall him because the pokemon was not (AFAIK) one you could get in the wild - so I can’t really say if he was hard, easy, or just countered my team.

I haven't played this game in years, but i still remember how Hideki sent me back the town over (i forgot to pokecenter) my first go at the gym. I stopped playing for two weeks.

I love the genre of "devs made an error in some coding value and players have suffered from it ever since". Another good one is how in Morrowind, the pickpocket skill check is accidentally stacked severely against you to the point that you need high level magic buffs to even have a semi-solid shot at success.

You will think this would get patched when wifi has been commonplace for years when the game were released 🙄

great video man

In my "casual" run I'musing dexnav to gather the buffest team of my favorites that there ever was. So I kinda just rolled up to this gym with a 3 star ralts and swept.

RS Haidiki can’t hurt you he doesn’t exist: This video

******** EDIT: Oops. I did all that analysis with Karate Chop instead of Low Sweep. Also didn't take EVs into account, lol ******* I ran the stats through an IV calculator, and going from 15 to 29 IVs in all stats gives him plus 1 in everything except for Attack and Defense (+2 each). Running it through Smogon's damage calculator, it's a guaranteed 2HKO on a -Def 0 IV Mudkip with crits either way (15 or 29 IVs). Without crits, it changes the chance of a 2HKO from 25% to 88.7% with 29 IVs, so that's pretty significant. But a speedrunner likely wouldn't settle for a 0IV Mudkip, and I can't imagine a -Def nature being useful. With a neutral nature and 24+ def IVs on Mudkip, even with the 29 IVs on the Machop, it becomes a 3HKO without crits. But it's still a guaranteed 2HKO with crits. If we lower the Machop down to 15 IVs, this means the Mudkip could survive 2 non-crits with only 8 def IVs. (I originally wrote 20, so maybe I made a mistake somewhere, lol) So really, it seems like this battle is influenced by the Machop's IVs, but only if you don't get a good defense stat on the Mudkip...or if you get unlucky RNG, which could happen either way. For a +def Nature Mudkip, it's actually only a 25% chance of 2HKO on crits from a 15 IV Machop...if it also has 31 IVs in defense (even 30 IVs is not enough to save it from a guaranteed 2HKO). As for non-crits against the +def Mudkip, against a 15 Atk Machop, it's a 42.5% chance of 3HKO, and a guaranteed 3HKO from the 29 Atk IV Machop. (actually, the Machop only needs 17 Atk IVs to achieve that) So lots of research and calculations that I probably explained poorly, but basically, I'd argue that the IVs are a smaller factor than plain and simple RNG here, mostly from crits and move choice! The exception is if Mudkip has a neutral defense nature, and between 8~23 defense IVs. It looks like you can soft-reset for a 20+ IV Mudkip (11 def on neutral nature instead of 10), so if you do SR for that def of 11, only an IV of 20~23 would make a difference and not be caught. That's a chance of 4/12 or 1/3, meaning that the programming error only has a 33.3% chance of influencing this battle at most! (it's actually less because this only comes into play when the AI only chooses Karate Chop and never crits). I'm sure speedrunners have dealt with random chances of more than 1/3 in their runs, haha If I've made any mistakes, I welcome any corrections from any other number nerds like myself! haha

DAN HIDEKI! BACK AFTER THE ULTIMATE IMPROVEMENT!

you've heard of takeshi's castle, but dewford gym? that's hideki's castle

Youngster Joey WISHES his Ratatta was as top percentage as this Machop