How FNaF RNG Broke... 3 times

Alternate Title: Everyone is stupid except me
TL;DW: The max value in the function used for random numbers in the game engine, Clickteam Fusion, overflows after 65,535, meaning Golden Freddy is actually running Random(34464) and Shadow Bonnie is running Random(16960). Also, because Clickteam Fusion only generates 16-bit integers, the Random Number Generator can only represent probabilities out of 65,536, which can be pretty inaccurate. Taking this into account means that Golden Freddy is actually 1/32,768 and Shadow Bonnie is 1/16,384. There was also a glitch with the game engine on Android that caused the Random Number Generator to output incorrect numbers, which made Golden Freddy impossible to encounter normally (on Android only).
Huge thanks to Folmic, Yunivers, and others for helping make this video possible; I appreciate it a ton ❤️
Become a Taken+ member for only $0.99 (at participating restaurants):
/ @itztaken
Additional Notes and context:
Note 1 (0:26): While I sort of imply that the issue with Golden Freddy's odds is a glitch, it seems to be an intentional (or at least known) limit of the game engine. It's an odd limit, and isn't immediately obvious, so I don't fault anyone for not knowing.
Note 2 (2:55): You can calculate exactly what number it overflows to by doing modulo 65536.
Note 3 (3:40): In order for Enemy Snowball attack damage to overflow, your health needs to exceed 65,535, which is only possible at level 13,087 or above. Because counters (including exp) are limited to 999,999,999, natural character levels are limited to below 10,000.
Note 4 (4:32): We're only interested in errors with Random(x) being out of the normal range, and Jack-O-Chica is the only one that both directly uses Random(AI) and functions below 1 AI. There are other animatronics that function below 1 AI, but they aren't affected by this limitation.
Note 5 (5:50): This bias is equivalent to modulo bias, but because the implementation actually uses a bitshift, the specific numbers that are affected are different/spread out, and the average is barely impacted.
Note 6 (6:09): In this case, I obviously wouldn't actually consider implementing rejection sampling. Changing the RNG to use 32-bit (or just using the built-in rand functions) would alleviate any significant bias.
Note 7 (6:40): My phrasing of "simulating the RNG" might be concerning, but I just mean calculating the number of RNG seeds that would cause Golden Freddy/Shadow Bonnie to spawn, not simulating the RNG millions of times, which would be inaccurate.
Note 8 (6:57): This was actually found before the RNG bias, but it flows best when it's placed after it {:}
Note 9 (7:43): Woo, the actual implementation! This is an implementation of a Linear Congruential Generator (LCG) with the parameters a = 31415, c = 1, and mod 65536. I can generally consider all seeds to be equally likely because the chosen starting seed is based on your computer time in milliseconds, and there's nothing too wrong with these particular parameters.
One thing to notice, though, is that the period for all seeds is m/4 or 16,384. This means that all seeds can be grouped into 4 long loops. For anything as/more common than 1/16,384, there will be at least 1 chance in every group, but Golden Freddy is 1/32,768 and groups 1 and 4 do not include him. This should not impact his odds much, as he only ever checks his RNG 535 times, and the starting seed is initialized every time you enter a new frame/new night.
This is, however, actually an issue in the modern Clickteam Fusion Windows runtime, as some error is causing it to only generate starting seeds in groups 1 and 4, and Golden Freddy wouldn't be possible if the game was updated. This is not a problem in any other runtime, so the remasters are not affected by this, not even the Microsoft Store version, as that uses the UWP runtime.
Music Used (in order):
Mini-Golf Intro - Five Nights at Freddy's: Security Breach (SiIvaGunner) ( • Mini-Golf Intro - Five... )
Super Mario 64 - File Select (fakeout)
Just Add Water - Freddy Fazbear's Pizzeria Simulator
Thank You For Your Patience - Freddy Fazbear's Pizzeria Simulator
Title Theme (Beta Mix) - Freddy Fazbear's Pizzeria Simulator (SiIvaGunner) ( • Title Theme (Beta Mix)... )
Title Theme - FNaF 1 - 7
Last Breath - Ultimate Custom Night
Shop - Five Nights at Freddy's AR: Special Delivery
Credits - Freddy Fazbear's Pizzeria Simulator
In-use throughout the video, “VHS Gothic Condensed” (fontstruct.com/fontstructions...) by Spottie Leonard is licensed under a Creative Commons Attribution-Share Alike license (creativecommons.org/licenses/b....
Chapters:
0:00 Intro
0:24 FFPS Stat Glitch
2:31 What This Changes
5:11 Let's Be Precise
6:53 But Wait... There's More!
8:35 Outro

[@itztaken]

*Check the description for extra notes and details, and for the list of background music throughout the video.* There are subtitles provided, and also one (telegraphed and silent) jumpscare at 4:41. Any questions are welcome, but this video is the best general explanation I could create. *FAQ:* *Q:* _What's MB2/Mystery Box 2 Jack-O-Bonnie?_ *A:* Mystery Box 2 (MB2) is a move in FNaF World that swaps out your party members with 4 random ones. Scott made a slight miscalculation, and the maximum character number was 41, which didn't even exist when the game launched. It being so rare meant it never got patched, and after update v1.2 dropped, Jack-O-Bonnie became character #41. *Q:* _Why are the probabilities powers of 2?_ *A:* They're simplified fractions of 4/65536 and 2/65536; as the way it's setup, the probabilities of all the random stuff in Clickteam Fusion are out of 65536. They just happen to be powers of 2 though, they could've been 3/65536 or 10/65536. *Q:* _Did the Android glitch also affect iOS?_ *A:* No, it was exclusive to Android. You can actually verify this by trying to find old videos of Golden Freddy on Android vs. iOS. It's not that hard to find people finding it on iPhone, while every video on Android just shows the 1/9/8/7 Easter egg. *Q:* _Is that The Cure - Close To Me playing?_ *A:* Yeah? Kinda, it's a SiIvaGunner rip that uses the FFPS soundfont: kzbin.info/www/bejne/qGeQg2V9dtiejdE

[@mattybrunolucaszeneresalas9072]

Telegraphed?

[@juanlilg2868]

Time travel comment

[@crystaljiang5976]

@@mattybrunolucaszeneresalas9072 I think he means that he warns you about it happening ahead of time

[@mattybrunolucaszeneresalas9072]

@@crystaljiang5976 I see

[@nonameTimeFile]

what jumpscare? without any sounds its looks like goofy fan animation, that what i called scary.

[@BackseatStreams]

Hell yeah brother. Keep those misinformers (me) in check! So wild that the odds people have been working with for so long were sometimes so dramatically off!

[@sotoastit]

hey man, love your videos!

[@michaele8444]

How dare you misinform the foolish viewers (me)!

[@Itz.MangoVR]

Yo backseat, whats good

[@Foodiefe2]

backseat didnt you figure out golden freddy on the fnaf 1 playthrough

[@MlilMbutPepeP]

Backseat, if you have watched astral spiff's FNAF world hard mode but 10 times harder, then you should send this video immediately and show him the chance to unscrew a boss

[@TechRules]

I'm sorry, it only generates 16-bit integers??? That's incredible. I remember seeing the unreasonably high odds of some of the easter eggs and frankly being shocked that we still saw them in content creator playthroughs as often as we did. Guess that explains a lot. The Android bug stands out to me significantly, as well. When I documented all the old mobile ports, I did so using iOS figuring there wouldn't be any substantial differences other than the microtransactions. I did notice that it took me significantly longer to get Golden Freddy to show up in FNaF 1's old port, but I just attributed that to possibly being because of faster nights, or just from bad luck. Because of that, I never believed any of the claims that Golden Freddy didn't exist on Android for a single second, since, y'know...everyone was really eager to guess randomly back then. Thank goodness we don't do that now, huh? Imagine if everyone guessed at something big like, for example, Fire Escape Ending requirements! Great video as always, and thanks for setting the record straight! (Also, using the negative overflow bug in FFPS to help explain the concept was honestly really slick. I need to start taking notes from your content it seems, lol)

[@itztaken]

You would have been right to use either port for any of the games after FNaF1, as they use the exact same MFA, just using different Clickteam exporters. But hold on, when did you document the old mobile ports 😅 Anyway, I tried to make the video mirror my own process of figuring out the why and how of it all. When Folmic mentioned a potential error with the RNG, that FFPS glitch was what I immediately went to, and I'm glad it translated well {:}

[@TechRules]

It was a while back, nothing public or interesting haha. I was mainly interested in seeing if it really was impossible to deter Foxy in the FNaF3 port, and that spiraled into checking a whole bunch of things. That was back when the tools to fully do that for the later games weren't public, so I felt a sort of obligation to notate all those things. Now that others can do it and there are great content creators that can uncover that information for everyone, I'm just always excited to see what the experts find lol. Looking forward to the next videos!

[@mjdevlog]

man i miss your content. Glad that you're still active as a viewer. Hope you're okay and healthy✨

[@RawrKatz]

I never expected to see tech rules here I love your channel

[@citriosis]

omg tech rules moment?? love your stuff!

[@lpfan4491]

It's amazing how(roughly) 33% of Fnaf jank in these early games is "either a debug-thing scott forgot to disable or a poorly thought out hotfix for something else", 33% is "Scott made a typo/forgot to set upper bounds" and the other 33% are "Clickteam just kinda sucks enginewise". Perfectly balanced, as all things should be. Lol. At least it didn't impact anything actually important this time.

[@gregorymirabella1423]

What other things have been impacted?

[@philipp04]

Huh. Explains how RWQFSFASXC actually appeared this frequently (there are quite a few people who saw it multiple times, wouldn't expect that if it was actually a 1/1000000)

[@itztaken]

Yep. I was happening to discuss with Folmic a case where someone got him twice, and the crazy odds I had calculated (~1 in 107501). He mentioned that Tru3P1ay3r alleged that there was some limit of ~40k with Clickteam RNG, but we quickly figured out it was actually an overflow, which adjusted the odds to ~1 in 36.

[@shrudigan]

Got them two nights in a row once and I KNEW the code was wrong

[@samloredescarregado1755]

without code on base and comparing, the odds is that you flip your monitor on and off more often on fnaf 2 than you check cam 2b on fnaf if i explained bad: in fnaf 2 by default you flip over you cam on and off several times, more than in fnaf 1 average night, due the freedom of no energy usage, and as the gamr forces you because puppet everytime otherwise: clickteam fusion buggy messy uggy

[@AstralSpiff]

THANKS FOR GAME RARITY SCREENSHOTS IM ON MY WAY TO BECOME THE 2ND TAKEN++ MEMBER

[@AstralSpiff]

I’m so plus’d up rn

[@Bradleybus]

What

[@amongsussyballs]

@@Bradleybus he plus'd up

[@black_m1n825]

what is even taken++????

[@itztaken]

is this an ad segue

[@oldamar9994]

my headcanon for the FFPS stat glitch is that this restaurant was actually a money laundering scheme, with the 20 visitors paying thousands in tips each. Henry destroys the restaurant at the end because the IRS found out about the scheme, burning all of the money and killing both him and Mike in the process. Can't believe henry would do such a thing 😔

[@theneverwatcher9860]

it was a glitch but now it's a scheme

[@TheLastestYo-KaiWatchFan9001]

Or maybe, since it has a negative entertainment value, weird rich folks came in, passed out because of how boring this place was, and got robbed by the staff.

[@TinyDeskEngineer]

truly the most terrifying of all fnaf characters is the IRS

[@herbivorethecarnivore8447]

I love how Scott made all these games and never realised his software keeps integer overflowing

[@passtheyaoi]

i mean whats he gonna do? run a million simulations to see if the events occur at teh expected probability? in the end it was still prettttty rare if not as comically rare as anticipated

[@Cinkodacs]

@@passtheyaoi Check the documentation for the methods? Random is one of those methods I always check, it varies a lot between languages, it's nice to know the range they work in sometimes.

[@itztaken]

@@Cinkodacs Clickteam does not have good documentation; searching for documentation on the RNG function only gives you the right thing 7 results down. And all it'll really tell you is the "maximum value" is 65535, but not much more (though the max is still more than I would expect from their documentation lol). It's not weird to just assume a game engine can generate random numbers larger than 5 digits.

[@samloredescarregado1755]

that's true, but i think as he playtested and saw these overflow werent affecting much at all, theres nothing else to do beside cracking the head to see if some rare png will appear in the game, either way would make it rare, not worth it

[@Hexagonaldonut]

Absolutely love the "you could use char, but... please don't" note at 7:38. For context, char (short for "character" as you might guess) as a data format is an unsigned 16-bit data type which *can* technically be used to store numbers, but what it's *actually* intended for is storing *text,* or more specifically individual letters, as the name implies. Programming languages tend to treat textual data types very differently from numerical ones, so actually *using* a type like "char" for storing numbers would require some type-conversion gymnastics, hence the "please don't". (It's interesting to point out regarding this subject that Java actually doesn't support unsigned *anything* for numbers. Every numerical data type is signed!)

[@1e1001]

16-bit char? i've only ever seen 8 (ascii/iso-8859-1) or 32 (full unicode)

[@Hexagonaldonut]

@@1e1001 At the very least it's what Java and a couple other languages use! I obviously can't speak to what char is in every language (some don't even *have* char as a data format), but I can't say I'm surprised some follow different standards. Hell, what's considered an "integer" can vary between languages.

[@Mimiyan_or_Pikapikafan]

I wasn't expecting to understand how overflow glitches work in detail, thanks!

[@itztaken]

yay an opportunity for pedantry 🎉 It's actually still called an overflow even if it's going in the negative direction. Underflow is a term that specifically applies to floating-point numbers that are too small to be represented, and turn into 0.

[@Mimiyan_or_Pikapikafan]

@@itztaken wow thanks for that! Man this was a massive learning experience lmao. I first heard of overflow glitches because of Pokemon RBY's EXP overflow glitch with level 1 Pokemon in the Medium Fast EXP Group (you normally can't get level 1 Pokemon in RBY, but glitches lmao.). If a level 1 Pokemon in that EXP group gained less then (I think it was) 64 or 63, it'd overflow and level the Pokemon up to level 100 instantly

[@camwoodstock]

this is kind of just a guess, but the reason that the games are capped to 16-bit numbers MIGHT be related to how Clickteam originated from a VERY, VERY OLD LINEAGE of game creator software, all the way back to Klik & Play. Klik & Play was literally built to run on *Windows 3.x*, which was indeed 16-bit! given that the underlying engine hasn't really been updated TOO much nowadays (clickteam has been surprisingly keen on backwards compatibility with project files), it makes sense that the standard variables are still in 16-bit.

[@itztaken]

That does seem to be potentially why, as the limitation dates back to at least 2006, though the implementation does use 32-bit integers. Overall, I don't see why the RNG wasn't updated to 32-bit in the decades since, as the only way that'd break backwards compatibility is if the program relied on undefined behaviour.

[@iansarmiento23]

For those who still don't understand, even though the code says it's 1/100000, the software it runs on cannot do that number so the chances are lower because it loops over.

[@TospikKing]

Why did u write this comment lmao

[@seymourkrelborn4780]

*chances are better

[@Josue_S_6411]

*The chances are higher.

[@miwr]

@@TospikKing For those who still dont understand

[@ragerasse7616]

Thanks! I have a really hard time understanding any computer or programming terms. I tried to make sense of the video but just found myself getting lost due to most of the things having a context to them that I had no clue about. Like what is the difference between a 32 bit integer and a 16 bit integer?

[@SomebodywithaYouTubeaccount]

This reminds me of the nuclear Gandhi glitch from the Civilization series. In one of the earlier games, the exact same negative number glitch happened, and when the programming generated a similarly massive number, Gandhi always went nuclear despite his commitment to peace. It’s become such a meme now that even in later games that avoided the glitch, the team still references “Nuclear Gandhi.” Civ 6 even gives him an 80% chance of having the secret agenda “Nuke Happy,” where the AI seeks to develop nuclear weapons and likes/dislikes others who do/don’t

[@itztaken]

Unfortunately, that's just a rumour and wasn't real, India just focused on science and usually ended up researching nukes first.

[@SomebodywithaYouTubeaccount]

@@itztaken The meme still pretty funny though, you gotta admit

[@swiftfated]

My brain didn't immediately make the leap to "every random number in all Fnaf games is an unsigned short as well" so you might have wanted to mention that when transitioning from pizza simulator to golden freddy. But awesome video still! I love learning video game math.

[@username5155]

1:47 I figured the whole “Clickteam doesn’t support other types of numbers” thing already kinda implied that

[@mrperson9877]

Love these breakdowns, they help clear up common misconceptions and are amazingly edited

[@fortheloveofgodlaugh2981]

Honestly without Markiplier's encounter, I think genuinely FNAF as a whole wouldn't have taken off so hard.

[@explodingbearz]

Videos like this are so fantastic. Excellent reminder of why I'm subscribed to you. A level of depth you don't really see with other content creators. Keep it up. Really love the videos

[@blacklight683]

Lol this is the hardest golden freddy and shadow bonny roast i've ever seen "you're so rare that you are no longer rare" they've been quiet like a body since this video dropped

[@kubeczek1019]

As someone who is really interested in technical fnaf I find this video really interesting! I'm really hyped for the fnaf glitches video!

[@slorty4494]

Kube Jumpscare??

[@Speed-TV]

My mind would be blown seeing shadow bonnie on people's playthroughs knowing it was a 1 in a million chance, but now I know it was more common than golden freddy

[@TheRandomWolf]

1/16,000 is still quite rare, being struck by lightning in a lifetime is 1/15,300 But yes makes way more sense than 1,000,000 because I’ve seen him and I KNOW I’m not lucky enough to see a 1,000,000 chance

[@Speed-TV]

@@TheRandomWolf only difference is that shadow bonnie has a chance every time you lift down the camera, which is bound to happen a lot. Although golden freddy is every second so he may be more common after all

[@davidnailgun4817]

@@Speed-TV i think it still runs the RNG every second and then waits to spawn it when the camera drops

[@Speed-TV]

@@davidnailgun4817 i know

[@lpfan4491]

@@davidnailgun4817 Does it just set the result when it happens or does it only spawn when the RN is right when you flip down?

[@Jackwald001]

Updating the wikia pages as I type this, sourcing this video. I was not aware it could work this way, well done!

[@FoxxNoodles]

The editing. The smooth keyframing. The assets to make with what game is being talked about. 3:17 Bro spelt out his name, and every letter had a sound effect. It's amazing.

[@l9m241]

Interesting how the biased odds go to 32768 which is close to the 32767 which is the 16 bit signed integer limit. I was curious about how big the RNG function was especially when dealing with overflow. Clickteam Fusions' engine is surprising! Funny how the Android port was just broken due to the Java/C++ Differences.

[@itztaken]

It goes to 32768 because it's 2/65536.

[@zandbergx9747]

A really genuinely excellent video. Love how the practical/technical physical implementations actually changed odds by a large amount. The explanation with FFPS's bug was great and your content isn't over-fluffed particularly, love it. Also big fan of the muted, telegraphed jumpscares, lol

[@PanMan47]

Your content is great! I love how this part at 3:11 also acts as a document for random events in the fnaf games, nice to see it all detailed

[@snerttt]

This is outstandingly comprehensive! The description is amazing lol.

[@birbies]

I remember like, 7-8 years ago I was trying to find the decompiled source of fnaf and managed to find it *eventually* after a ton of searching (i was like, 10). super cool to see this kind of analysis! well done

[@colekiesler6218]

Let’s all agree that Golden Freddy’s jumpscare is the definition of unexpectedness but it’s at peak.

[@kagu2811]

In FNAF 1, there’s actually a one in three gorrilion chance that Freddy does the Gangnam style (real)

[@za_vishmoongarr]

I saw that happen once

[@a-bombmori7393]

This happened to Markiplier but he had to take the video down for copyright 😔

[@AIEchidna]

this is such an awesome video, you've done it yet again, Taken!

[@-JJB-]

I love your editing! It’s so well done!

[@crystaljiang5976]

you can really tell that it improved a bunch over his past few videos!

[@Frizou]

Wow that was actually much more interesting than I first expected, I'm not really into fnaf but I'll always take a video talking about video games statistics, I really appreciate the quality of this video and I thank you for putting a jumpscare warning, way too few channels do this kind of stuff

[@wolfcl0ck]

Wow! Love videos like this that look into the deeper parts of how the games function. Well done!

[@foor]

It's insane how this answer was right under everyone's noses as well lmao Great video :)

[@partypug123]

Love your visual style. Editing was top notch

[@erichdavis2897]

Really well made vid man! You've got the editing skills of a master, and it really made this a good watch!

[@cheesebooba]

i really liked this video! i like these "education" typa vids about fnaf. just kinda makes me comfy

[@TacticaLLR]

Great vid! I love your content!

[@Faze_Reverse]

Like your style of fnaf videos, I wish this one was longer. I waited months to the point I forgot tho KZbin recommends came in clutch ill be subscribing and I hope to see more from you.

[@27ricky04]

It's the first time I see a video on this topic after starting to learn c++ in uni and it took me this long to finally understand fnaf rng, worth it tbh

[@masontirant4996]

The editing is REALLY GOOD

[@iminumst7827]

Honestly, it would be 5000 IQ move as a game designer / programmer to code a "rare" event to happen more often than stated even in the code. That way a lot more people playing the game get the excitement of experiencing a super rare event.

[@archerelms]

I love that idea lol You'd have to do it carefully though if you wanted it to be reasonably rare and not way too common or something wild

[@sagacious03]

Neat analysis video! Thanks for uploading!

[@DimentiosLoyalest]

I will never stop being entertained by the catastrophic results of games receiving numbers slightly outside of expected ranges.

[@joshistitic]

Your videos are amazing keep doing more! I wanna see AI breakdown of sister location some of the nights in that game are so confusing

[@a1k021]

I just did the pizzeria sim glitch, night 1 was fine. Don’t try it in nights 2 and onwards. I’ve been sitting here for 5 minutes as my returning visitors skyrockets past 30K 😂

[@JarheadHME]

Very very interesting, I love seeing the technical aspects of stuff like this gone in-depth! Out of curiosity, do the fnaf world lower tier reapers and unscrew 2 have different odds?

[@itztaken]

Lower tier reapers don't have lower odds, they just shoot less often. Unscrew odds are 1/30 and 1/600 for enemies and bosses, while Unscrew 2 is 1/20 and 1/500.

[@blueyandicy]

I love the alternative description LOL, like so true

[@yellowpowr8455]

It was creepier to see Golden Freddy NOT jumpscare people.

[@AskBlommaert]

That is very interesting and funny to see how such a little technical limitation can change so much!

[@GGGames._]

This video is a damn revolution!! Could I use the data and information from the video to make a Spanish version for my channel? I would love to make something like this possible, I don't think I can reach your level of editing tho

[@itztaken]

Yeah, sure! I think it's some pretty interesting stuff, but I should say though, I'm always a bit wary of information degradation in the FNaF community. Especially for the technical information, FNaF KZbin has felt like a game of broken telephone. For this video in particular, it was very important to me that the script and visual presentation would make it as easy as possible to grasp the concepts and reasoning behind it (it's a rather large claim after all). So, if you need any clarification or additional details, I'd be glad to help.

[@bryanmulcare9083]

I’m glad we got an answers how the games work and thanks for the showing good quality of math. Sometimes doing the math helps out to understand the game better. In steam to this day is still 1.35, I could be wrong but I did remember playing FNAF 1 recently.

[@crystaljiang5976]

loving that new profile picture! finally got your own brand :)

[@CataIana]

Always look forward to another one of your videos

[@DavidXNewton]

I love deep explanations of code like this - thanks for your curiosity :)

[@GameExploiters]

It's cool to see the game break by the engine and not Scott himself😭. Great vid tho

[@username5155]

At least there wasn’t some Mario 64 pendulum manipulation type stuff

[@itztaken]

it felt like it in the moment

[@shradibop]

i would LOVE a video on how UCN works! i used your video on fnaf 6 for my recreation, and it's super useful and interesting.

[@TheBadDeadFox]

Another banger, another banger from ItzTaken

[@horpuscorpus8299]

I wonder if Scott actually knew about this limitation of Clickteam, or if he knew but didn't bother with the specific value of things occuring so he just put in a really large number and assumed it'll be rare from there

[@itztaken]

It's obvious he didn't know of this limitation, though he did know of the 999,999,999 limit that counters normally have in Clickteam, as evidenced by the Random(999999999) in SL's Breaker Room.

[@myralyra2404]

Thanks for explaining this very clearly so someone who has zero coding experience (like me hehe) could understand it fairly well :D

[@Redd2D]

As someone who's super jumpy over jumpscares, thank you so much for the countdown, very appreciated

[@zanzaroon6961]

I had a mini panic attack when the mario 64 music came on following the "let me explain"

[@testerwulf3357]

I thought it was odd for some eastereggs to supposedly be suuuper rare..yet be ungodly common to see 😅 I guess now I know why! Thanks for the video

[@jacksnipe2441]

Great analysis. Why they would make these implementation choices is an amusing mystery.

[@chanlennox8133]

I really love these kind of videos, are there any other games/series you have in mind to do this kind of digging with?

[@itztaken]

Not particularly. FNaF is a pretty unique niche that has a pretty big fanbase, somewhat mysterious inner workings, very easily decompiled games, and a relatively few people that actually examine them. If I could easily decompile At Dead Of Night, I would love to look into it, but from what I know, it'd be pretty difficult.

[@chanlennox8133]

@@itztakenwell your pizzeria simulator game got me to finally play fnaf after all these years because it made it seem way more interesting from a gameplay perspective. So thanks!

[@keemotherapyman]

@@itztaken There's this wildly obscure game you could try digging into, maybe you've heard of it before? I believe it goes by "Smuper Smorsh Florsh 2" for the Nintendon't PC

[@JaredThe]

I love giggling to myself about these kinds of under/overflow glitches in games they're always the funniest

[@Sivta]

Great vid!

[@pacblox5417]

Never stop these types of videos

[@crystaljiang5976]

true

[@astralwyzard]

I could binge hours of your videos. With that being said, please make more videos so i can binge them for hours!!!!!

[@derpycats8072]

This video was extremely insightful

[@KiiBon]

Growing up is realizing early FNAF was held together by duct tape

[@Youhadabadday2021]

Here's another fun fact: There's actually terms for this exact type of bug. We call them interger underflow and interger overflow respectively. Both result in the same outcome, but underflow occurs when the number is less than the minimum interger limit, and overflow occurs when it's more than the maximum integer limit.

[@itztaken]

It's actually still called an overflow even if it's going in the negative direction. Underflow is a term that specifically applies to floating-point numbers that are too small to be represented, and turn into 0.

[@HollowLeif]

I have never played a fnaf game but I still find this type of content very interesting.

[@ghb323]

@5:50 65535 MOD 100 = 35, so in this case, a shorten RNG range will have an output of 0-35 occurring more often than 35-100. Range 0-35 will each integer be 1/656 chance while 36-100 will each integer be 1/655 chance. Calculated by dividing 65535 by 100, which gives out 655.35, or 655 remainder 35. We have 655 complete ranges of 100-size, with an incomplete range size of 36. Because Modulo outputs a number strictly lower than the denominator, the numbers in the lower part of the range happens more often.

[@GodlessAnti]

Greatest Five Nights at Freddy’s channel on the platform.

[@meteorolite]

I feel like this deserves a place in the Fnaf Iceberg. Well done.

[@commando_fada]

Love your content

[@Winasaurus]

While negative revenue provides seemingly unlimited revenue, I like to imagine that negative health and safety similarly provides AMAZING health and safety, curing the attendees of pre-existing conditions they had before arriving, and causing legal teams to pay your fees to fight their own case against themselves, because with health that good, there's no way you're liable for anything.

[@azadanzans5359]

Here I was thinking seeing shadow bonnie was one of the rarest things that’s happened to me… ever, for years. Well that’s some pretty smart research, and is incredible interesting, especially for games that I was 100% sure we had completely cracked

[@the_cheese_cultist]

so clickteam uses modulo for generating an integer in a range instead of a proper uniform distribution? cool

[@gloomyArietine]

Integer underflow causes so many cool glitches

[@MudakTheMultiplier]

I don't really care about FNaF, but hot dog I love a good breakdown of broken video game code.

[@rubiksmath7938]

Wow. One final thing I’ve noticed and I’m sure you and others have is that the final calculated odds (32,768 and 16,384) are exact powers of two, is this related?

[@itztaken]

Yeah, they're simplified fractions of 4/65536 and 2/65536, as the way it's setup, the probabilities of all the random stuff in Clickteam Fusion are out of 65536.

[@NoppyOwO]

GREAT HEAVENS MAN, YOU'RE GREAT

[@PebsBeans]

absolutely great video

[@esotericpince]

great video :) was easy to follow as a guy who knows 0 coding and failed highschool math

[@namelesschannel4103]

Barely know anything about fnaf, never played it, gud vid 👍

[@Dagu427]

I have no idea how you stumbled upon that android fix information but that's incredibly interesting

[@kalele60]

Man I can’t believe that Cassidy and the other one show up more often just because they both skipped math

[@hword5notsoofficial803]

This altered RNG makes the odds of seeing golden freddy in a run go from 2.63% or whatever it is, all the way up to 7.84%. Almost triple toe odds

[@pacblox5417]

I have a question, do you think the pizza sim glitch will ever be patched? I don’t know if these games get frequent patches im just wondering

[@itztaken]

Not on Steam, Scott has not patched anything in a very long time, and the only recent patches were to bugs with the application as a whole. I don't know about the ports, maybe? Most glitches are still present in the ports, but I do know they have patched a couple.

[@VasakonTvLikesSurvivor]

Good video and explanation! Even that the odds arent so small, i still cant believe that he apperead in my offie! Anyway bout fnaf 2 i heared that on you screen are numbers that you cant see and theese are ais of the animatronics. There is one that belongs to paper plates ( i think). Like ai of the paper plates and when one of them apears in office. If you know something about it, could you explain it?

[@itztaken]

The Paperpal acts just like every other animatronic and has its own AI, uses movement opportunities, and is even affected by camera stalling/flashlight stun. Every night the Paperpal has a 1% chance of having 1 AI. If they succeed a movement opportunity, they'll move into your room, appearing when you next flip the cams. Because of a mistake, his AI gets reset at the beginning of every non-custom night, so he'll only appear on Night 7/the Custom Night.

[@VasakonTvLikesSurvivor]

Ok thanks man

[@ag2023en]

1 ÷ 100.000 = 0.00001 1 - 0,00001 = 0.99999 The chances of NOT getting Golden Freddy (in FNAF 1) are that of 99.999% every second. You're only getting the chances of surviving the following seconds AFTER the already limited chance of having survived the previous second. A night in FNAF 1 lasts for 8 minutes and 55 seconds, or 535 seconds. You need to elevate 0,99999 to the power of 535, and then multiply the result by 100 to get the final probability (in percentage) of NOT getting Golden Freddy across the whole night, which are a little over ~99.46%. Then you just subtract the result from 100% and you get the probability of getting Golden Freddy across a whole night, which is little under ~0,54%, or 1 in a little over 187. (100÷(100-((0,99999^535)×100)) = ~187,4153991048) When you take into account how many people played FNAF, as well as the fact that each player can play multiple nights, Golden Freddy appearing isn't as unlikely as it might initially seem by the "1 in 100000" figure. And remember, that is just the chances of him appearing for each ONE individual night. Most players will play multiple nights. When you account for playing 5 whole nights in a row, the odds of Golden Freddy appearing increase to a little under ~2.64%, or 1 in a little under 38. 5 nights × 535 seconds per night = 2675 seconds. (100÷(100-((0,99999^2.675)×100)) = ~ 37,8852198051) So, around 1 in every 38 players who play the time equivalent of 5 whole nights will encounter Golden Freddy. Meaning Golden Freddy is a LOT more likely to appear than people might initially think.

[@koekje00005]

Fantastic video :) you should also be aware though that this is pseudo-random and not true random, so it will always be biased in some way

[@-C4-]

Thank you Pannenkoek

[@itztaken]

no problem will

[@Spagoooterman]

This guy has the ultimate flex: He can spell Shadow Bonnie's real name without hesitation and very fast

[@yellowpowr8455]

Thanks you, FNAF Pannenkoek.

[@joseihandlenotavailable]

man the shadow bonnie one is so goofy literally everyone thinking they are getting stupidly lucky and it's just so goofy

[@nopeleader2137]

looks like Scott didn't do his research the first one looks like something quite easy to fix, just make it generate random numbers more times

[@itztaken]

That would be easy if you were working with a _True_ Random Number Generator, but computers usually have to settle for _Pseudo_ Random Number Generators, which are entirely determined by the seed. If you generated 2 random numbers one after the other, the second number would always be determined by the first, and you still wouldn't be able to represent any probabilities more precise or smaller than 1/65,536. It would be possible to reseed the RNG with some user input or the user's system time, but that'd be needlessly complex. The best solution would to just use an extension like MT Random that uses a better algorithm and outputs a larger range of values. Anyways, it makes no sense for Clickteam to have limited the RNG to 16-bit values, as all other numbers in Clickteam Fusion use 32-bit, and the only way you'd know that the Random() function was different, would be to dig into the documentation and see the small note at the end.

[@nopeleader2137]

​@@itztaken oh right I forgor you still could space out the generation between the sub rngs since it's checking a generated number every 1 second