Startside Underholdning How lucky is too lucky?: The Minecraft Speedrunning Dream Controversy Explained

2 måneder siden

How lucky is too lucky?: The Minecraft Speedrunning Dream Controversy Explained

Ganger 3,100,756

30 704 304

Buy my book Humble Pi now!
www.penguinrandomhouse.com/books/610964/humble-pi-by-matt-parker/
UPDATE: Harvard Book Store have totally sold out.
My talk on 4 February 2021 at Harvard Book Store:
www.harvard.com/event/virtual_event_matt_parker/
Here is the original accusation against Dream.
Video: nomines.info/clip/video/XoO_rJ6Cl691lmQ.html&ab_channel=Geosquare
Paper: mcspeedrun.com/dream.pdf
And here is Dream's reply.
Video: nomines.info/clip/video/Yp_XpbiogrmcjYE.html&ab_channel=DreamXD
Paper: drive.google.com/file/d/1yfLURFdDhMfrvI2cFMdYM8f_M_IRoAlM/view
"Matt flips a coin 100 times."
nomines.info/clip/video/hWi-qc6qZZqlonU.html
"Holy Craps! How a Gambling Grandma Broke the Record"
content.time.com/time/nation/article/0,8599,1901663,00.html
Number of casinos in the world in 2011: 3,547
www.statista.com/statistics/221031/total-worldwide-casinos-by-region/
Roulette records.
www.roulette17.com/stories/record-reds-blacks-in-a-row/
CORRECTIONS
- At 09:08 I say “1 in 110 trillion” when I meant to say “1 in 110 billion”. The number on the screen is correct, it was just a verbal slip-up.
- At 25:27 I showed the 118 craps record as “1 in 1.2 × 10^9” when it should be “1 in 2.2 × 10^9”. The voiceover says the correct number.
- I slip and “more likely” instead of “less likely” at 33:47 (I think I may have even been going for “more unlikely”). But everything in the screen is correct.
- Let me know if you spot any more mistakes!
Thanks to my Patreon supporters who mean I can spend [[REDACTED]] hours filming myself trying to achieve improbably things. If you support me, you can get access to all [[REDACTED]] hours of bonus footage from this episode.
www.patreon.com/standupmaths
As always: thanks to Jane Street who support my channel. They're amazing. And I believe they have no opinion on Dream.
www.janestreet.com/
Endless filming by Matt Parker
Editing by Alex Genn-Bash
Some graphics by Ben Sparks
Minecraft consultancy by Oliver Dunk
Music by Howard Carter
Design by Simon Wright and Adam Robinson
MATT PARKER: Stand-up Mathematician
Website: standupmaths.com/
US book: www.penguinrandomhouse.com/books/610964/humble-pi-by-matt-parker/
UK book: mathsgear.co.uk/collections/books/products/humble-pi-signed-paperback

Nedlasting

Kommentarer

Stand-up Maths
2 måneder siden
I’m not commenting on how many takes that took. But feel free to guess! (And if you must know: the complete footage of every attempt will be uploaded to Patreon. patreon.com/standupmaths )
T Lo
8 dager siden
did you feel a little weird saying the same sentence over and over again?^^
Duck
11 dager siden
jkm
shamanahaboolist
Måned siden
Sneaky magnet behind the board?
Ellery Johnson
Måned siden
I'm super in the idle Heroes and a lot of it is based on RNG and a lot of people in like the Facebook groups will talk about people cheating and hacks is stupid stuff like that because they're too stupid to realize that people are lucky and people are smart and can figure out how to use luck to their advantage and quantify that but whatever.
Charske Vlogs
Måned siden
I wonder how long the bball thing took XD
Occasionalnerd
5 timer siden
Th ten billion human second century concept is amazing, and I'm saddened there's not another video on this but should it be in a future video I will be delighted. An excellent way of showing just how unlikely something is.
Iron Warrior
6 timer siden
That one guy defending dream:"but its possible to occure therefore dream is innocent!"
nate williamson
8 timer siden
Wonder how many times he filmed the darts and basketball parts and had to redo it lol
james pond
12 timer siden
How lucky is too lucky?? How many tries did that book throw take??
Anvilshock
13 timer siden
Minecraft: We're big in the news! EVE Online, KSP: First time?
Leon Filmalter
15 timer siden
26:09 dude.... That's a lot of craps
Steven Stokes
16 timer siden
Did anyone else start again from 1:08 to see if he started with a beard
levi c
23 timer siden
i like how there was 5050 runs for flipping a coin 100 times
Dreamhunter
Dag siden
TLDR; Dream hacked, get over it kid
LuxTenebrae
Dag siden
This is probably the best delivered explanation of this whole situation. It’s also explained very simply which is really nice
Anakin Ligman
Dag siden
My question is, aren't the other streams available? Couldn't they look at the stats that occurred over all 11 streams?
Athenri
Dag siden
Doing that is just going to obfuscate the relevant numbers. There's really no reason to want to do that. For an analogy, imagine that you're looking at me flipping a coin. I get 30 flips that are all just about normal. Then suddenly after I leave and come back, I get 30 heads in a row. If you're wondering whether I changed my coin, you wouldn't add in results that I got with something we're pretty sure is a real coin.
LS Aria
Dag siden
Only thing I can think of that might have some kind of impact is if he found a Looting sword which could massively affect the Blaze Rod drops, but even then the bartering is pretty damning...
Дмитрий Бондаренко
Dag siden
i hate dream
Tech Guy
Dag siden
He is too lucky. Let him be lucky. Kids follow the trend and such things. Kids would never follow real games where it require skills *cough * cough CSGO
crash
13 timer siden
Needing luck doesn't mean you don't need skill. I fail to see the logic.
fanny fanny high lights
Dag siden
How long did it take to make that dart shot???
Super 100
Dag siden
Uh
Matías Pascual
Dag siden
Great video, subscribed!
MrFeanaro9
2 dager siden
As a maths tutor with lots of Minecraft playing students, I just want to say a very heartfelt thank you for this video!
Gabriel Saunero
2 dager siden
I freaking loved your video. Totally agree, mixing Minecraft with Mathematics education for a lovely purpose such as discovering CHEATING :D A PERFECT 10.
SandalphonCPU • 210 years ago
2 dager siden
Obviously it’s because he’s the main character. But of course, he traded his home for it.
Kate vlroux
2 dager siden
Hello sir, if I wish to buy the book, where can I buy the extended version? Does Harvard Book Store still sell the original and or the ext? I'd like to know, thank you in advance.
Guard #3
2 dager siden
I literally just learned how to put numbers in and take numbers out of scientific notation in algebra class why am I here :’)
hatty 90
2 dager siden
And he only got what would have been 4th place, what is 1st place odds!?!?!?!
Null Pointer
Dag siden
@Shaltinanwenor that's correct, he was on world record pace until he was in the portal room
Shaltinanwenor
Dag siden
If memory serves, he was lacking one pearl when he went to the portal so he had to get another one. If he hadn't miscounted the pearls (or had a portal with one extra pearl in the portal) he was on pace for world first.
morbideddie
2 dager siden
As mentioned in the video this is an analysis of the full 6 streams. The law of large numbers states that as more trials are completed the odds will then to converge on the true value. Therefore over the 6 streams we should be seeing drop rates in line with the theoretical value. To give you an example, say you are in a dice rolling competition with 100 other people and your goal is to roll three dice and get all sixes in as fewer rolls as possible. Normally the odds of getting three sixes is 0.5% but you modify the dice to give you a 6 40% of the time meaning your odds of getting the three sixes is 6.4% per trial. You have 10x better odds of winning the contest than anyone else but you still will likely lose because there are 100 other people playing. In this case Dream modified the dice to favour him but that doesn’t mean that a few of the thousand other people he is competing with won’t get individual runs that are luckier.
prowo
2 dager siden
This wasn't the analysis of his odds in 1 run - it was an analysis of his odds over a week of streaming and hundreds of runs. His average luck over that long period of time was too high. You can get really lucky for just a small instant and it's not nearly as significant - it's sample size. Additionally, Dream is not the best speedrunner in the game and people can outperform him based on skill/strategy, no need to attribute the success of those on the leaderboard above him completely to luck.
Vasileios Lempesis
2 dager siden
But what if we had an enormous amount of bots playing minecraft... how many bot hours would it take to make these results likely... and is that even possible with the currently available processing power?
crash
2 dager siden
@Vasileios Lempesis No, just because a computer has done any amount of simulations, even if he did 10^25^25 amoutn of simulations, it still wouldn't change how likely it is that Dream did it. The only way to change it is if an actual run would be do-able in less time and thus be done more. Just because a simulation can simulate something within a second what a speedrunner has to do within 20 minutes or so doesn't change anything. It should not be taken into account. But if the run had been done after 10^22 times and there are runs that are almost equally as likely in there, then it becomes probable that a result like that has happened. Just like grinding in Doom or Goldeneye. If a newcomer to the speedrunning community gets a world record with incredible RNG then there are gonna be some eyebrows raised. Even though many other people have done those runs and gotten equivalent RNG it is still very unlikely for this one person to get this RNG on his first submitted run.
Athenri
2 dager siden
@Vasileios Lempesis Right, obviously we don't have to *actually* run runs, we can simulate them instead, and then we can probably get a lot more done. The question is how much fidelity to actual minecraft play we lose by doing so. Like in reality what we want to do is to try to get 42 ender pearl trades and 211 blaze rod drops and see how often we need less than 262 barters and 305 blaze kills - skipping out on the entire surrounding framework of Minecraft.
Vasileios Lempesis
2 dager siden
@Athenri that's an excellent answer. However we would not have to run the simulation in real time, it would be merely a question of processing power for the simulations to be run in just a few seconds. With enough processing power you could start a livestream, run all the simulations the first second of the stream and just playback the one that fits your purpose. What is missing from the equation of course is how much processing power it would take to do that. But I agree we would probably end up with a huge amount of time required. Some sources are claiming a global computing power of 2 x 10^20-1.5x10^21 FLOPS in 2015 others project that we will be in the 10^21 range in 2030...
Vasileios Lempesis
2 dager siden
@crash an outcome with odds in the 10^22 range would be almost inevitable after 10^25 attempts. For the sake of discussion, one could claim that if an AI on a supercomputer is able to run that many simulations within a reasonable amount of time, then in theory, that should be also accounted for in the "10 billion human second century" argument. If that many simulations were occuring somewhere then one could claim that the run was merely a chance event (if we can agree that human runs and simulations are to be counted together). I surely don't mean that you would have to believe them, but that it would have been a mathematically valid claim.
Athenri
2 dager siden
@Vasileios Lempesis You can do about 4 runs in an hour, so to match Dream's final run, around one in 8000, you'd need around 2000 bot hours. Totally not impossible. To match Dream's overall runs, we'll need 20 sextillion sets of...let's just say 32 runs. At 15 minutes per run, each set takes 8 hours. So we're, in the end, going to need around 160 sextillion bot hours before we expect a set to look as good as Dream's did. That's 1.6*10^23 bot hours. Let's assume we can run a billion bots concurrently. So we're getting 10^9 bot hours each hour. So we're getting 2.4*10^10 bot hours each day. We're doing it every day, around the clock, so in a year we get 8.76*10^12 bot hours out... So at one billion bots, we're going to need them to run for about 20 billion years before one of the sets is expected to be as good as Dream's. So if you had a billion bots running Minecraft non-stop trying to mimic Dream's runs for the entire existence of the universe to date, we still wouldn't get there. We'd be in the ballpark though!
alejandro.javier
2 dager siden
Don't you love probabilities 😘
S L
2 dager siden
the human second century is slightly misleading in this case because a minecraft speedrun lasts over 10 minutes not 1 second
Null Pointer
Dag siden
that's the point, and it isn't even just a single run but his sequence of runs over multiple streams. What he's doing is describing an upper bound to show how unlikely that sequence of runs actually was.
prowo
2 dager siden
It's not misleading, you just fail to understand what he means. The human second century is just to give the odds a scale. It's odd to think about, but just imagine that people do a set of 6 minecraft streams just like he did and try to get his odds, but shrunken down into one second, and they do that every second for a century. It's weird that that much stuff is supposed to fit into a second, but seconds in a century are just meant to show how much it happens.
SpyTUC
2 dager siden
Do you have a methlab?
On yo feet maggot
3 dager siden
This guy looks like Dorian Yates if he decided to become a mathematician instead of a bodybuilder
Christian Sitzman
3 dager siden
17:17 ... Nice
Götarp
3 dager siden
We want bloopers
Christo du Plessis
3 dager siden
Patricia breaking the world record in roulette and craps directly after each other really sunk any chance Dream had to convince the world. Btw, what is minecraft?
king frozen
3 dager siden
The beard is 10/10
HOLLSON
3 dager siden
Hello. Nice video. But nowdays, there are more speedruns like that and Dream is not the fastest anymore. How lucky they must be.
ThatKyle
Time siden
the strats in minecraft speedrunning has changed since dream's cheating controversy. it is now even more skill-based than luck. instead of getting lucky with pearls, you just need to find a bastion which is easy with strats like eray
prowo
2 dager siden
They have faster times because they are better and use more optimized strategies. Dream is not the best minecraft speedrunner in the slightest. You do not need to be luckier than him to beat him. Anyways, people who use modern strategies trade so much gold with bastions that they don't actually need to get lucky to complete runs quickly.
morbideddie
3 dager siden
Dreams luck in his submitted run are not suspicious, it’s suspicious that over the 6 streams he got consistently high rates which is incredibly unlikely. It’s like comparing getting 7/10 heads and comparing it to 70/100 heads. Same rate, vastly different probabilities. Also new runs use loads of new meta with ocean strays, bastions to minimise luck required in trading, quadrants in the nether etc. It’s basically a new run since Dream was doing it.
Athenri
3 dager siden
Not nearly as lucky, honestly.
Silvertarian
3 dager siden
Bell curves don't lie.
me me
3 dager siden
10:31 Dani 10th very cool.
mustafa
3 dager siden
Guys, he just had a good gaming chair
The Swiss Guy
3 dager siden
Why do people always just unanimously decide that they like someone... who then turn out that they’re in some massive drama.
Gauss Niwre
3 dager siden
Astrophysicist getting the statistics wrong... no surprise here... astronomical ladder, Betelgeuse distance "correction", etc..
Llamma Pajama
2 dager siden
The Astrophysicist doesn't exist. This is legitimately highschool level math.
The Uncle
3 dager siden
Glad that Scott Stapp became an educational youtuber. Didn't know he was British.
Harshita Joshi
3 dager siden
This is my first time here. I've never enjoyed math so much.😀
Gus Ziliotto
3 dager siden
I love how seriously you learned the ins and outs of the minecraft speedrun. Shows a lot of respect. ♡(◡‿◡✿)
C I R C U S
3 dager siden
I don't want to know how many attempts he needed for the intro.
Denis M
4 dager siden
Curious why the puzzle on his book shelf during the bowling pins in background was not turned to full completion (maybe just me)
Tomasz Rasolomampionona
4 dager siden
Timestamps for all lucky shots: 0:12 Dart 6:39 Basket 1.1 6:45 Basket 1.2 9:42 Basket 2.1 9:47 Basket 2.2 9:53 Basket 2.3 10:10 Basket 2.4 31:51 Pins
Hazel
3 dager siden
Thanks, also u forgot the book one at 17:16
Samuel Rosenpenis
4 dager siden
i like watching people finishing games AFAP as well, especially when they need more skill than luck, like mario kart.
Unusual Comment
4 dager siden
I don't quite get it having insane odds means that you probably should'nt attempt it but when you got it you can't say that it didn't happened. I mean patricia getting 154 runs thus nobody having ever done more than 118 doesn't mean it's wrong cause someone should've get 119 to 153 runs before jumping to 154
Д
4 dager siden
Probabilities of what that streamer did are wrong here. As I understand from the video, a player makes barters until they have necessary amount of pearls. So it is not correct to calculate the probability of X or more pearls for Y barters, it should be instead the probability of Y or less barters for X pearls. Simple example of what I mean: a coin was tossed twice. It came up tails once. What is a probability for that? It depends of what the process was. First option: a coin was tossed twice. There are 4 possible outcomes, in 2 of them the coin come up tails once. So, the probability = 1/2. Second option: a coin was tossed until it came up tails exactly once. It means for the first toss the coin came up heads, and for the second toss it came up tails. So, the probability = 1/4.
Д
Dag siden
@crash I did not say that probabilities change depending on what happens before or after, I said that probabilities change depending on what the process is. If a process is "toss a coin 2 times", then there are 4 equally probable outcomes, in 2 of them the coin comes up tails exactly once. Therefore the probability of "2 tosses, 1 tails" is 1/2. If a process is "toss a coin until it comes up tails", then there are infinitely many possible outcomes, in all of them the coin comes up tails exactly once. But all these outcomes have different number of coin tosses. Probability for an outcome with N tosses is 1/2^N. Therefore the probability of "2 tosses, 1 tails" is 1/4.
Athenri
2 dager siden
@crash nah, you're not understanding his comment. Probabilities model situations and it is, I agree, much more reasonable to ask the question of "how likely is it to get this many pearls this quickly" than "how likely is it to get more than this amount of pearls in this amount of attempts", because the pearls is what you're fishing for.
crash
3 dager siden
No. Probabilities don't change depending on what happens before or after. You can account for any bias though which this video goes into but getting a heads or tails is always a 50/50 no matter what. I can toss a coin 20 times and get 20 heads, but whether my 21st flip will be a head or tails is still 50/50. There is nothing physically changing about the coin to change those odds. It doesn't care what happened before.
Snow the Mega Absol
4 dager siden
Why do people still not realize this means nothing? Whether it was astronomically lucky or not doesn't tell you whether or not the run is impossible, just how likely it is. If you can imagine it, and the laws of physics don't prevent it, then you have to be prepared for the possibility. All anyone has been able to do is prove that the run is just really unlikely. Want to know what else is unlikely? Astatine, the rarest element on the planet. If you want a chance of finding enough atoms together to make 1 gram of the stuff, you're gonna need to check an average of 30 quintillion tons of matter on our planet. So if you go out and find it all on your first 1 thousand, that's not evidence of cheating, that's just another instance of possible things that can happen when you make your system probabilistic. If anyone ever does it again, then your intuition about how unlikely it is is wrong. If no one ever does it again, that means your math is right. But so what if your math is right? Again, all you've shown it's improbable. Improbability is not a synonym with impossibility, and the use of improbability to judge something as impossible is literally called the appeal to improbability fallacy. Everyone has tackled this investigation from the wrong angle. What you all should have been doing is trying to recreate the initial conditions under which the run took place, and see if an unmodified version of the game would produce the same results. Either it would, or it wouldn't, and then you'd have a clear yes or no answer.
crash
2 timer siden
@Snow the Mega Absol but what if CCTV footage shows the face of the doppelganger? Is that empirical? What about if he wore the same clothes as the convicted in court? They could've shopped at the same store. Technically all possible. Plenty of times has someone been convicted using CCTV footage. So according to your logic, these people should be freed from prison, right? Idk why you have such a problem with 'the odds are too small to be considered likely enough to have happened'. Everyone else seems to accept this, yet you don't. If someone flips a coin 20 times and gets 11 heads and 9 tails and then leaves the room. When he comes back the flips it another 30 times and gets 30 tails in a row. No rational person would simply believe he is still using a normal coin. And using this to devalidate a speedrun is not arbitrary. It's using likelyhood and statistics to come to a conclusion. This is why courts have judges. You can't just put down rules in a book and have everything going smoothly. Lines aren't always black and white. Saying this is arbitrary is saying any judgement system using a jury is arbitrary as well.
crash
2 timer siden
@Snow the Mega Absol do you need to define a point to pass judgement? You probably use the term moon. But it has no real definition. Do the thousands of tiny rocks orbiting a planet count as a moon? Does dust? Yet you still use the definition so trying to get the highground like this is not gonna work. What if a streamer has 24 hours of footage of him getting a blaze rod every kill and a pearl drop every barter? Do you think we should say those runs are valid? Should we free all prisoners who didn't confess on the off chance they have a doppelganger who committed the crime? Proven beyond any reasonable doubt. That's more than enough for this. And for your case of 'those big numbers are shaky and fluid at best' is unsubstantial. Unless you can show me slight changes in the droprates that make the number actually that much different. Millions of simulations couldn't get the drops Dream did, so obviously there's some validity to it. Even the number his own statistician came up with made him conclude that Dream likely altered the game.
Snow the Mega Absol
3 timer siden
"They literally never say that it is impossible for someone to roll a 1 in 20 sextillion chance." Whether or not you say it out loud doesn't mean you don't believe it. And clearly people do, which is objectively wrong. "This is not the appeal to improbability fallacy because they don't use the fact that it is improbable to say it is impossible" See above. If you believe it's impossible, then that is fallacious. If you don't, but still don't think the run is legit, then your judgment is fundamentally arbitrary, so you're in the wrong either way. "They say that it is improbable beyond any reasonable doubt and it is so much exponentially more likely that he is cheating" So where's the cutoff point? At what point do you decide to accept a run just because it is "too lucky"? You can't put a bar on that anywhere, to do so would be a completely arbitrary decision and a disgusting way to practice moderation. "which is the only rational conclusion." Whether you call it rational or not doesn't mean it's the only conclusion at all. "but I have no idea who you're arguing against in that case" I'm not arguing at all, I'm pointing out to no one in particular that all the shaky fluid statistics don't, have not, and never call tell you anything as to the objective truth. Any yes, the mods have decided that the run is, in fact, not legit as backed by their decision to reject it. The fact that they made a decision at all without actually knowing is highly presumptuous at best. If they actually didn't know, don't make a decision. If they did, share the results from an empirical analysis rather than a shaky courtroom style shout fest where people throw around big numbers in an attempt to create a narrative, when those big numbers are shaky and fluid at best, and straight up wrong at worst.
Snow the Mega Absol
3 timer siden
the prisoner analogy is horrible. Just because eyewitness testimony is strong evidence in a courtroom doesn't mean it is epistemologically, in fact it doesn't even count as evidence. Whether or not your footage is good or not depends on where the camera is, how good it is, and whether or not there's any other explanation for what it shows you. Games, which are just compilations of code that can be dissected down to their last bits, can give you so much more. And what have such analyses revealed? There is nothing about the run that is impossible, and no one has been able to show otherwise. "Beyond a reasonable doubt" doesn't apply to empirical analysis. The fact that you would employ a part of a legal system that doesn't even function right as a counterpoint is beyond me. If it's possible, there's nothing that can prevent it's existence, then you always have to be prepared for the possibility. To say otherwise is to admit that the validity of any speedrun in any game is just an arbitrary judgment, and thus that leaderboards are not arranged meritocratically. The run should be put back up on the table and actually looked at without hiding behind probability, it wasn't designed to be used like this. If there's straight up no way to prove its illegitimacy, not only beyond doubt but independent of it, then it shouldn't be treated as illegitimate.
prowo
2 dager siden
This does absolutely not mean nothing. You do not need 100% certainty to prove something beyond a reasonable doubt. They literally never say that it is impossible for someone to roll a 1 in 20 sextillion chance. This is not the appeal to improbability fallacy because they don't use the fact that it is improbable to say it is impossible. They say that it is improbable beyond any reasonable doubt and it is so much exponentially more likely that he is cheating that they're just going to take his runs off the leaderboard, which is the only rational conclusion. Your whole post might make sense if you're trying to argue against someone who says it's impossible to get this lucky, but I have no idea who you're arguing against in that case. The speedrun mods didn't say that, and neither did this video poster.
Ryan de Kwaadsteniet
4 dager siden
Genius.
TheR3dMage
4 dager siden
The ten billion human second century is an incredibly smart shortcut to look at odds and it feels like a concept I am going to use when thinking about rare events.
secretyzeroo TM
4 dager siden
12:02 : bruh, the video will continue 30 minutes and you don’t habe time?!
Jonathon Cowley-Thom
4 dager siden
Lotteries regularly go unwon, so the odds of someone winning it is not 1. It's nearly 1. But it isn't 1.
morbideddie
4 dager siden
True but sometimes multiple people win, so I guess Matt was trying to say is that on average 1 person will win per draw? His wording could have been better though.
kèhbab
4 dager siden
And he hasnt even taken world and nether spawns, gold drops and chest loot
Joseph Normyle
4 dager siden
This video was poggers
Strat
4 dager siden
Dream: Sees this video. *Sweats profusely*
83athom
5 dager siden
Don't disagree with anything you said, but as someone who dabbles with programming (especially java) I just want to quicky pip in. Random number generation in Java (and other codes) is not really random or fair. There are methods to seed the RNG functions to produce the same results multiple times over, which is why usually those RNG functions are fed a seed based on your system time so you actually get a "random" output (this is the default constructor of the object when you just call the random function). Similarly the same code ran through the same engine behaves differently based on what system or environment is running it, Java is especially notorious for this and doubly so with its random generation functions. This means theoretically it's also possible to fudge the game RNG one way or another based on what hardware is running it. I remember years ago testing this with a very simple code to write out an array of numbers between 0 and 10 and found some environments would overwhelmingly favor between 4 and 6 with few 1s or 9s while another would pretty much near evenly output all of the numbers.
Athenri
5 dager siden
This isn't *wrong* per se - but there are a couple of issues with it as applying to Minecraft: 1. Since this is across multiple runs, not just one, you'd want to know how often you hit a "degenerate" seed like that. 2. The seed is called thousands of times per second in the Nether, of which the RNG calls we're interested in constitute a low two-digit number here. This *is* in fact, a random sample of that generated sample, so unless you have reason to believe that the entire maybe 100k large sample is off and not just local regions inside it, that's a no-go. 3. In support of 2. you don't see results like this from any other speedrunners. 4. If he had, deliberately, somehow, found a way to set a seed in that version of Minecraft that would let him control the RNG for drops, for example, that would disqualify his run.
8BitXatu
5 dager siden
The chance of finding a 1/100 wild encounter rate shiny Pokemon (1/4096) with 6 perfect IVs (1/32^6), a favorable nature (1/25) and ability (1/2), it being the uncommon gender in a species with the most uncommon gender ratios (1/7), and holding an item with a 1/100 chance to be held... 1 in 1.54 x 10^19. This is about the rarest possible encounter possible in Pokemon without getting too much into the weeds, but according to Matt this occurrence is totally feasible. It demonstrates the astronomical luck that this streamer would have needed for his run to be legit. This happening in a community with many known cheaters on a game known for its seasoned modding community? I'm gonna bet that he's a cheater.
smol kookie
8 timer siden
It would be rarer if it had pokerus as well, and shiny rate was 1/8192 until gen 6
Frst Rspndr
5 dager siden
So if someone could only get 4th fastest with dubious means how did someone get the fastest time
prowo
2 dager siden
What Athenri said - this analysis is about Dream's luck over a massive sample size of a whole week of playing. Being moderately lucky over a very long period of time can have a lower p-value than being extremely lucky for a moment. For example, getting 7/7 blaze rods and 2/2 pearls (perfect at the time of his run) is way more likely than getting his total drops over the week, even though he did not have an individual run getting that lucky.
Athenri
5 dager siden
What cheating like this gives you is not so much impossible results - what it gives you is impossible *consistency* in your results. None of Dream's runs were anywhere *near* impossible. His final run was about a one in 8000 occurrence. That's not *that* rare. The issue is that the consistency of those results across 24 hours worth of streaming was impossible.
Quinn Johnson
5 dager siden
5 takes and how many takes would it take to get that kind of luck in minecraft
Sean Avery
5 dager siden
"I'm not saying he used a hacked client with 100% certainty. I'm just saying that if every human being throughout the planet's entire lifetime realistically played Minecraft nonstop for 100 years of their lives until they pass away the chances of getting those odds would still be off." He's basically saying that because there's TECHNICALLY a chance going by the laws and rules of Math and Science to be professional here, but it's obvious that he's silently telling us and to avoid the dislikes by Dream's irrational fans that he 100% used a hacked client in his run.
Owen Penner
2 dager siden
@warron24 Dream is a developer too.
Sean Avery
2 dager siden
@warron24 That's possible, but not in this situation anymore for countless reasons later on that proves that couldn't have been the case with some of the stuff Dream shared unintentionally giving himself the final nail in the coffin while still denying everything. The game also couldn't have just made a buggy mistake as Java is very simple and can't just have some sort of glitch like that happen where it improves his odds as the moderators and Karl Jobst's video discusses.
warron24
2 dager siden
I think he was also leaving open the possibility that Dream he wasn't cheating but had his chances improved by something he wasn't aware of. I don't know enough about Minecraft to know if that's possible.
Owen Penner
2 dager siden
Thanks you literally saved my life.
Bubbler Guy Super
6 dager siden
Also fun fact, he is wrong (maybe) with the humanly plausible thing of 3 x 10 to the power of 19 because Sullivan was struck 7 times by lightning which is 4 x 10 to the power of 32: much higher than his idea which makes dream's speedruns look like they were common.
lolong Donaire
4 dager siden
@Athenri godamn tell em Athenri that will teach em
Athenri
5 dager siden
You're assuming independence in who is "chosen" to be struck by lightning. I don't think that's a fair assumption to make. Not all people are equally likely to be outside in weather where lightning is remotely possible.
Bubbler Guy Super
6 dager siden
me getting 250 out of 400 blaze rods : thats only like 44% chance right? him: its 0.00000000000000000000000008%
Athenri
5 dager siden
for the intuition on this - when you *double* the amount of, say, blaze kills, the *average* drop amount doubles. But the deviation, how much we expect it to vary, is only multiplied by the square root of 2. So if, for example, you normally would get (bear in mind, I'm going to make up random numbers now for illustration) you kill 10 blazes, and are expected to get between 3 and 7 blaze rods. If you kill 40 blazes, you're expected to get between 16 and 24. If you kill 160 blazes you're expected to get between 72 and 88. So where for 10 blazes, we believe the result until we have more than 70% drops in the first case, for 160 blazes 55% is actually all we'll believe. In reality the *specific* numbers are different, but the gist, that quadrupling the number of kills quadruples the average but only doubles the deviation holds.
Daunte Love
6 dager siden
Minecraft jury duty
Caro
6 dager siden
Imma give you like and comment just for the introduction even tho I haven't yet seen 2 minutes of video
→ to the knee
6 dager siden
Lol, I just paused the video at 29:03 to make a calculation regarding 10^22, and what do I see in the background...
→ to the knee
6 dager siden
27:23 - Why is it not just 1 in 2^32? (Assuming it could be 32 black runs OR 32 red runs)
morbideddie
6 dager siden
Chances of getting red or black is 0.474. For 32 tries on either black or red the probability is 2*(0.474^32) = 8.4e^-11. Converting that the odds are 1/1.18e^11.
Violetsuperstitions
6 dager siden
Thanks for getting me interested in statistics! What a well made video!
→ to the knee
6 dager siden
15:21 - When P(THAT) rolled onto the screen, I lolled. More like chuckled out loud, but whatever.
Matthew Saints
6 dager siden
I thought it was Jim Carrey in the thumbnail.
CODE PRODUCTION
6 dager siden
i hate cheaters
Bernardo Pereira
6 dager siden
Great beard! Hope you stick to it
Hanan G
6 dager siden
First, I enjoyed your presentation. Your approach reminds me of the block box concept in Engineering - where you ignore what the apparatus does and just look at output and input out of it - ie not really caring about the validity of the compensation mechanism. Second, I wonder how unlikely are the current records that are acknowledged? Are they within the Human century benchmark or are they more likely than Dream's but still unlikely?
prowo
2 dager siden
Note that this analysis was using a sample of a week of Dream's speedruns, it wasn't about an individual run of his. Dream gets moderately lucky over a large sample size, which is why his lucky is attracting way more attention than some runners getting extreme luck in just 1 run that places them at the top of the leaderboard. This is why assuming that people above him on the leaderboard are more lucky than him is wrong. Also, that kind of assumes it's all up to luck how high you place. There are people significantly better than him at the game, and the current records actually use more innovate strategies that allow you to trade so much gold that you are likely to get enough pearls each run.
crash
6 dager siden
At the time the mods included a graph of Dreams' drops Vs another speedrunner and it was clear that dreams' was way higher. The reason Dream's run was disqualified was because he livestreamed for 24 hours. If he hadn't speedran and just uploaded the single run then he would've gotten away with it because having such a small sample size makes the luck more believable.
LoLrand0mness
7 dager siden
or those algorithm pleasing things... we are on it. don't cha worry.
Gabe Keeter
7 dager siden
This is a good video, I think Dream's argument should have been centered on the fact that java.util.Random has some serious issues... mostly not being random. It actually uses a seed to generate it's values *pseudo* randomly, which is why there are so many interesting tricks in the game like finding diamonds using clay patches, and while world seed doesn't affect drops and spawn rates, the theoretical value of those things likely doesn't tell the whole story. This kinda makes me think that it's more likely that the game isn't what we thought rather than Dream intentionally changing the values
Athenri
7 dager siden
Then that'd fail to explain why every single other streamer gets drops largely in line with the expected amount. Even the luckiest one they could find other than Dream was far far less lucky.
Sir Amoras
7 dager siden
I don´t even play Minecraft but I enjoyed this video very much, thank you Sir!
Patrick Sarama
7 dager siden
He just has a really good gaming chair guys
Sir Rounded
7 dager siden
Nice video, thanks a lot! I think it would have also been interesting to analyze the luck of just one of the videos. To see how plausible that one is. Yours is the 1. video I watched on this topic, but purely relying on my instinct I'd think having such a luck in not just one, but 6 takes, like Dream had, is mindboggingly improbable.
Athenri
7 dager siden
The probability of his eventual best run was like, one in 8000. Rare, not exceedingly so. Definitely entirely plausible. But six livestreams, altogether like 30-odd runs with this kind of bonkers outliers? No way.
Thaddeus Cosse
7 dager siden
What surprises me about this, is the amount of people that defended the paper. I'm glad I wasn't the only one that felt it wasn't well written.
Harshita Joshi
3 dager siden
I think, judging by this video, I could start writing papers like this. I hated mathematics cz I wasn't good at it.😂😂😂
Samuel Feder
7 dager siden
+
Mrbuck832 Idk
7 dager siden
new topic: when is a game too dependent on RNG to be speedrun?
Stacy McCabe
23 timer siden
The new straits are faster and are almost impossible to not get all he pearls
morbideddie
7 dager siden
Eh, it’s rng dependent but so are loads of popular speed runs to a large degree. Ocarina of time, goldeneye etc. Additionally, removing RNG means that the runs are only about execution, part of the appeal of RSG (as opposed to SSG) is the fact your playing a new world each time and the skill is in reacting to the world and routing a new route on the fly. And if you don’t like the lottery aspect you can play FSG or do live events etc. Plenty of options and no-one is forcing runners to do RSG.
T Lo
8 dager siden
destroyed 100
john bill
8 dager siden
As fast as possible? You mean... Afap
_ Sqrrl
8 dager siden
Still more likely to happen than the next Shrek movie
J-Swag
5 dager siden
:-(
Draculathescary
6 dager siden
True lol
Felutia's Fans
8 dager siden
Honestly, I never like anyone who overly introduce themselves as expert and explaining what expertise they have and how much they experienced, also how "important" they are. If I ever read or hear that statement, I never buy their words, it will be a pretty challenging situation to me to try to judge objectively by that moment
Fluffy White Llama
4 dager siden
PhD Harvard astrophysicist intensifies
Ctrl Z
8 dager siden
If dream was beating Minecraft every second for 3.4 quadrillion years, then and only then would he of been able to do that, he used hacks. Definitively.
krucial
8 dager siden
Anyone who's ever played Runescape knows this is entirely possible. RNG works in mysterious ways.
Stacy McCabe
23 timer siden
@Draculathescary 7,500,000,000,000*
Draculathescary
5 dager siden
@krucial 1 in 750000000 is possible but it’s really not
krucial
5 dager siden
@Draculathescary I guess I'm about half serious. If I was killing 240 fire giants that had a 1/20 (~5%) drop rate for a rune scimitar, and they ended up dropping 40, that would be pretty normal. I'm sure that just means I'm missing something but I've seen some crazy good and crazy bad RNG in that game. In the end, it's all just numbers right?
Draculathescary
6 dager siden
I really hope ur joking
morbideddie
8 dager siden
Nah, trillions of simulations and this luck was not recreated. Two entirely different games with two entirely different RNG systems are not necessarily similar.
JaffNo898
8 dager siden
I love this guys channel. I don't even like math. Saw you on numberphile first but this is great.
Dill Pickle
8 dager siden
Also. I have a VERY good challenge for you if you'd be up to it. GACHA games or LOOT BOXES and what the actual math of those are when you have the base % rates for each character and how often you should see that character OR a character of the same RARITY... IF you are interested let me know and I'll send you the data. I'd like to see if mathematics wise, they are on the up and up. 👍
crash
8 dager siden
Aren't those things handled by servers though? You can never be sure of the actual chances. But if you've got the data you can simply get a rough estimate and see how close that resembles the actual drops.
Dill Pickle
8 dager siden
I'm just here to see how many craps players came here to correct his gambling knowledge. On your very FIRST roll... your BEST outcome is to ACTUALLY roll 7 or 11. If you roll a 2, 3, or 12 you lose first roll. If neither a 7 nor 11 is rolled, and none of the auto loss numbers (2 3 12) are rolled, the number then carries on to whatever number is showing. Let's say it landed on 8. NOW the goal is to roll AGAIN until you either roll an 8 and WIN or you roll a 7 and lose. (Crap out) there is 0 need to explain the others parts like how you bet. Hard way or any of that. So yea. First roll. You WANT a 7 or 11 for best outcome. Worst is 2 3 12. But after the initial roll. If it ISNT 7 11 or any of those, you DONT want to see another 7 until you've seen whatever number you've rolled. 👍
KillerKitten753
8 dager siden
This made me realize that if multiple universes actually exist, there exists a reality where a completely legitimate speed runner has gotten disqualified for genuinely being that astronomically lucky, and people rightly assumed he was cheating.
KillerKitten753
5 dager siden
@RichConnerGMN thanks!
RichConnerGMN
5 dager siden
nice pfp
Magnus Kramnik
8 dager siden
Me: (throws 1000 dices, got some random outcome) How likely it is to get this outcome? Mathematician: 1 in 2^1000, it's nearly zero. This should not have happened. Me: but it happened, didn't it? Mathematician: my maths says otherwise.
Magnus Kramnik
8 dager siden
@morbideddie Yeah. Many people I know get probability completely wrong. They didn't even understand 50% probability of coin toss.
morbideddie
8 dager siden
@Magnus Kramnik oh, I didn’t think you were joking. Given the number of people who say these sorts of things and are completely serious, the Poe’s law is real. As for your question Athenri has already done a good job answering so I have nothing more to add.
Magnus Kramnik
8 dager siden
@Athenri Yes. That makes sense.
Athenri
8 dager siden
@Magnus Kramnik depends on what you're asking for, in particular, and the assumptions we make. It's going to boil down to there being N different outcomes and you counting a success as getting any of them, so P=1.
Magnus Kramnik
8 dager siden
@morbideddie I know what he's saying and I also know what I was referring to. I'm just making a joke. BTW how do you formulate the case in my scenario?
Jaronomo
8 dager siden
If it falls within the realm of possibility, and can be conceived to some extent as possible within the human realm, then in Dream's case lucky isn't too lucky. Highly improbable, yes, 100% agreed. But until there is concrete evidence of Dream ACTUALLY cheating in the game he is innocent. Just like our society is supposed to run on "innocent until proven guilty." Though people disagree at the chances something like this can occur, special phenomena does occur from time to time and Dream has as much right to remain innocent until proven guilty.
Andres Garcia
3 dager siden
@Jaronomo then why did he delete his videos and tweets defending himself?
Jaronomo
4 dager siden
@Andres Garcia But it is still technically possible. That's the point I'm making. Regardless of what you all are saying it is still statistically possible. Case in point.
Andres Garcia
6 dager siden
@Jaronomo basically, it is technically possible but not really. The numbers were still incredibly small after accounting for lots of variables in dreams favor. After that, think that it happened to be livestreamed, and all of it happened in a week. That just makes the probability even less realistic. It is more likely for someone to spontaneously combust without a reason than for dream to not be cheating.
Jaronomo
8 dager siden
@Scott I'm fine where I'm at Scott. You see, unlike you I am not completely skeptical of luck. Dream has always denied the accusations but has any other party, or person, done to LEGITIMATELY prove him guilty? If he was PROVEN guilty, then wouldn't all his records be stripped along with possibly being banned from both NOmines and Twitch? But, no, none of those things happened. Even NOmines released statistics in which Dot Esports could only muster saying it did not exonerate him but at most suggest it was not impossible he was lucky. So... Try again.
Scott
8 dager siden
@Jaronomo Read the research paper that the mod team worked on and you may actually understand the Topic of conversation.
Shawn Shurtz
8 dager siden
Maths .
roi’s cleo
9 dager siden
The dislikes are from Dream stans
$\*/ IcEbERg \*/$
\*/ IcEbERg \*/
9 dager siden
So, basically, a mathematician, who is also a stand-up man, does a video about Minecraft in a dude perfect style. What a world I live in...
Andrew Pye
9 dager siden
I’d like to point out that he dismissed the comparison to the lottery because “people win all the time”, but to use an analogy later referenced, “that’s like throwing a dart out of a plan and drawing a target around where it lands”. The odds of ME winning the lottery is very low. The odds of SOMEONE winning is 1 because they draw until someone wins. But.... obviously this Minecraft run is way less probable than winning the lottery, so maybe reasonable to throw that out as an argument.
Athenri
6 dager siden
@Andrew Pye Yes, there's an obvious imposition of meaning in the lottery. That was the point, and the distinction between the lottery and the darts and airplanes analogy. Gold star!
Andrew Pye
6 dager siden
@Athenri there's an obvious imposition of meaning in the lottery. You have to have the correct sequence of balls. Not sure why you're throwing it out as a question of probability. Like I said, they draw until somebody wins, so somebody will win every time. But we would be talking about the odds of a certain person winning, which is imposing a stricter/meaningful boundary when we are talking about the lottery. So idk I feel like you just argued my point for me...
Athenri
8 dager siden
Nah, those are different. One is about unlikely things being likely if you try enough times. That's the lottery. Unlikely for one person, exceedingly likely for a couple million. This is legitimate statistics - if you want to know if a result is likely to happen at all, you have to account for the many people who try. Another is about being able to assign objectively meaningless probabilities by pretending any result is rare in a meaningful sense. "Oh man, he threw three dice in a row, and the first was 2, the second was 4 and the third was 3, you only get a result this rare once in 216 attempts" - when really, any combination of three ordered dice is that rare. This is bad statistics. Meanwhile if you have some kind of imposition of meaning - maybe you're only interested in sums, you can talk about probability.
Filipe Paixão
9 dager siden
To be fair... the minimal amount of time is: 5.39 × 10−44 s so: 5.75 × 1062 is the actual probability of a human in 100 years do a specific thing exactly like he did.
JIALE
9 dager siden
How many takes did the intro take
Rocco Spezzaferri
9 dager siden
fully understandable video. but one thing - maybe a little bit off topic - is tickling my mind. how can it be estimated the impact of the fact that numbers in the game are pseudo-random generated and not really random? is it known what is the algorithm used by minecraft? firs time watching your videos, love them. ciao
Rocco Spezzaferri
9 dager siden
@crash got it ty
crash
9 dager siden
The mods went into this in their document. Basically, it's defined by the world/world seed which is also randomly generated. That number is being called hundreds of times a second, there is no viable way to manipulate this. The number that is used for generating the loot for the pearls and the blazes is separate from each other as well. The odds of the RNG numbers aligning this much is as likely as it being true randomness.