A beard? How long did that take to film?

February 4, 2021 10:46 AM

Noted internet maths person Matt Parker explains the Minecraft speed running controversy: How lucky is too lucky?
posted by Pendragon (36 comments total)

Matt Parker explains the Minecraft speed running controversy

I think we should reserve judgment until we hear from Trey Stone.
posted by The Tensor at 12:54 PM on February 4, 2021 [9 favorites]

I've played Minecraft off and on, but needed some background on the controversy to understand what was going on. This Polygon article helped me out.
posted by They sucked his brains out! at 1:28 PM on February 4, 2021 [6 favorites]

My son and his friends were chatting about it during a recent Minecraft session (mostly noting the breakdown in a popular multiplayer server, they observed players wearing the same skin Dream uses and were assuming these players were Dream 'stans there to support their idol). Their takeaway was that the people in Dream skins were to be shunned, because they were clearly fomenting unrest in the community.

I was aware there were speedrunning controversies but never read into it enough to figure out what player was at the root of it - the Dream comments make more sense to me now. I really did like the explanation of the odds involved here. It seems the evidence is just a wee bit damning!

Personally I primarily play on a private server with limited whitelisted players, and am in no hurry at all to "finish" the game. I'm just happy building bridges and roads and tunnels in the silly world my son and I built together over the past 5-6 years. The closest thing to a controversy in our world is that I feel he abuses his teleport privileges, and we had to talk sternly to a new player who got full netherite gear SUSPICIOUSLY fast ... (Scandal erupted when a long-time player accused him of cheating via many many signs planted near his base. He apologized profusely and promised to never again use an X-ray plugin on our server. I kinda sorta believe him, but what are you going to do, he's 11. It's the risk of deciding it was OK to let my son invite more of his friends to the world.)

(Building multi-thousand block long roads between far-flung bases is surprisingly pleasing to me. Plus after digging that giant cylindrical hole down to bedrock a while back, I have lots and lots of rock and cobblestone to build said roads. I might be utterly awful at PVP, and regularly get my butt kicked by tweens in Bed Wars, but I can build a really nice road.)
posted by caution live frogs at 1:51 PM on February 4, 2021 [2 favorites]

According to the video, the likelihood of each drop from a given encounter is determined by a loot table. We're looking at two different sorts of encounters (piglins and blazes), so: two tables.

If it was just one suspiciously-lucky set of drops I would have thought that he had found some glitch that overwrote the values in the relevant table. That sort of glitch – overwriting – seems to be not uncommon. According to comments in the Polygon it was supposed to be a no-glitch run, but hey, he might have stumbled on it accidentally. Having it happen twice basically eliminates that possibility: you don't accidentally stumble across a lucky glitch twice in one run.

On review, I suppose it's possible that all the loot tables are stored continuously, and a single glitch could lead to them both of the relevant ones being overwritten simultaneously, but in that case you'd expect lots of other things to be overwritten too, and the game would look really weird.
posted by Joe in Australia at 2:07 PM on February 4, 2021 [1 favorite]

I supposed it's not impossible, but a "glitch" where a table gets overwritten sounds far more like the kind of bug that might happen on an old console game (where the software was super-optimized and close the bare metal) than in a PC game written in Java.
posted by The Tensor at 2:27 PM on February 4, 2021 [1 favorite]

Unless it's fetching the tables from disk every time anything is possible. Here's a fun list of bugs relating to loot tables. https://minecraft.gamepedia.com/Loot_table
posted by tmt at 2:36 PM on February 4, 2021 [1 favorite]

The original investigation paper reviews the Java code and doesn't find a way to explain the successful runs with glitches. It also points out that they're two unrelated parts of the codebase (e.g. using different random number generators) that would both have to fail in precisely the way that helped this particular speedrunning task, out of many ways to fail, and only for this particular speedrunner. It's plausible in the abstract that a loot table could be glitched by player input, but in this particular case it sounds like "this speedrunner who likes writing Minecraft mods wrote a Minecraft mod" is a lot more likely.

This whole episode is fascinating as far as how consensus-truth is formed in a community where there's no established authority. Like it's unlikely that an old world authority (court, investigative journalist, peer reviewed science paper, etc.) will ever care about this thing enough to nail it down, so people who do care have to convince each other, and it's interesting to see how they engage when they're trying to do that.

(I'm also working on an alternate theory where having enough people all stare at the same random number generator and will it to produce the same random number at the same time can break the simulation and cause it to happen. Behind cheating, that seems like maybe the most plausible explanation. And you have to feel for Dream who could very possibly be the most likely person in the world to surface that bug.)
posted by john hadron collider at 2:44 PM on February 4, 2021 [8 favorites]

I don't know minecraft, but one of the YT comments pointed out that the speed runs involve a lot of "straight down digging" which miraculously hit the resource needed, which apparently isn't really supposed to be a thing. Dunno if that actually means anything.
posted by maxwelton at 3:03 PM on February 4, 2021

Anthropic counter-hypothesis: Across all the multiverses where Dream speedran, ours is the only one where life survived the subsequent events. So, the probability of the speed run is 1, because we're here. How the existence of a speedrun saved the world to be determined; I look forward to the subsequent papers.
posted by joeyh at 3:18 PM on February 4, 2021 [11 favorites]

World champion of flipping most heads in a row embroiled in cheating scandal for statistically implausible number of heads flipped in a row. "I think the math is just really convincing that thirty-six heads in a row is so vanishingly unlikely that we have to presume this is cheating" says previous record holder of thirty-two heads in a row.
posted by Pyry at 3:32 PM on February 4, 2021 [3 favorites]

People online (not here btw) seem to be spending quite a lot of effort to say that someone with an incentive to cheat, the ability to cheat, and the opportunity to cheat, did not in fact cheat, despite the truly astronomical odds against this sequence of events occurring in an honest manner.

...I mean, if the odds are truly that insane, when the event(s) occurred Dream probably would have stopped the speedrun to just stand in awe of what just happened. That would have suddenly become the entire point of the video, speedrun discarded immediately, "dudes! did you just see that? holy shit! holy shit!"
posted by aramaic at 3:48 PM on February 4, 2021 [4 favorites]

per the Polygon article "why fake a fifth-place speedrun"?
posted by GuyZero at 3:52 PM on February 4, 2021 [1 favorite]

I don't know minecraft, but one of the YT comments pointed out that the speed runs involve a lot of "straight down digging" which miraculously hit the resource needed, which apparently isn't really supposed to be a thing. Dunno if that actually means anything.

Anecdotally speaking, your odds of hitting lava are much greater than your odds of hitting any resource you actually need. If you're wearing diamond armor, your odds of hitting lava first seem to go to as close to 1-in-1 as makes no odds. It's why the 1st rule of Minecraft is: "Never dig down." (No, really: https://lmgtfy.app/?q=1st+rule+of+minecraft)
posted by erikred at 3:55 PM on February 4, 2021

I have a barely passing interest in Minecraft, speedruns, and statistics yet I found the combination surprisingly gripping.

Very clever to have the interstitial gags be a red herring.
posted by ChurchHatesTucker at 4:09 PM on February 4, 2021 [1 favorite]

I don't know minecraft, but one of the YT comments pointed out that the speed runs involve a lot of "straight down digging" which miraculously hit the resource needed, which apparently isn't really supposed to be a thing. Dunno if that actually means anything.

I believe that these speedruns are done with a specific world seed, which means locations like ore deposits and the end portal are in the same place every time. They're digging straight down because they've memorized where everything is. The only part that's still left to chance is the locations of mobs and their drop rates, which happens to be the exact thing this controversy is about.
posted by Mr.Encyclopedia at 4:17 PM on February 4, 2021 [3 favorites]

Dream's response to people complaining about scripting.
posted by The corpse in the library at 5:42 PM on February 4, 2021

Sorry, sloppy commenting. That video is about a different discussion, but I think it's interesting insight into this world. The first few minutes are just waiting for notifications to go out -- jump ahead a bit.
posted by The corpse in the library at 5:45 PM on February 4, 2021

per the Polygon article "why fake a fifth-place speedrun"?

Someone might use a cheat that merely increased their odds of winning because a cheat that assured victory would be too obvious. With a speed run competition you might have to turn in your best time before you know how good it needs to be to win.
posted by straight at 6:05 PM on February 4, 2021 [2 favorites]

"No one who didn't wind up in first place could have or would have been cheating" is not a compelling argument.
posted by Earthtopus at 6:10 PM on February 4, 2021 [1 favorite]

I believe that these speedruns are done with a specific world seed

Its the flip of that, actually - speedruns are done with random generated world seeds, and if you watch speedrun streams on twitch they make a point of showing themselves getting a new random seed each time.

I spend a lot of time in Minecraft community twitter. I don't have a position on whether or not the run was a cheat. It was, at best, a 5th place run. Dream has been a breakout YouTube star - going from 100 subscribers on YouTube to over 17M and counting in about 18 months - and will continue to be highly successful even if he never speedruns again -- its where he started, but his YouTube career is not at all related to speedrunning. He is well known (and freely admits) to be highly highly competitive in everything he does.

All of which to say -- I can't imagine what his motivation would have been to cheat. He wasn't going to make any money doing it, and has a lot to lose by being caught. There is a small army of people online who seem to really, really hate him, and who have gone out of their way to try and create controversy around him. He's achieved a lot of his success by being genuinely kind, open, and supportive with his extensive fan community.

I don't know the answer. But bear in mind that he would have zero to gain by cheating on a run, and a tremendous amount to lose.
posted by anastasiav at 7:47 PM on February 4, 2021 [1 favorite]

The guy was talking 10^22; that’s vastly more than there are grains of sand on Earth (10^18). It’s an astoundingly huge number. The odds are unimaginably against this being an honest speedrun — an honest player would have a heart attack when getting those results. I can barely wrap my head around how huge these numbers are, it’s defeating my ability to come up with analogies that mean anything.
posted by aramaic at 8:55 PM on February 4, 2021 [3 favorites]

I could imagine a situation where a speedrunner decides to tweak the drop table for what they consider to be the most tedious part of a speedrun, not realizing how noticeable the end result would be.
posted by Mr.Encyclopedia at 9:00 PM on February 4, 2021 [6 favorites]

OMG my 12 year old has been yammering about this for a while. Wait till I tell her it made Metafilter. I don't know if I'll be briefly cool in her eyes or if she'll never follow Dream again.
posted by gaspode at 10:01 PM on February 4, 2021 [2 favorites]

re: "why cheat a 5th-place run", someone in the comments of Matt Parker's video, who had watched the run in question, pointed out that Dream's run was in fact on pace to be the world record, but it was wrecked near the end by an unrelated instance of bad RNG (fewer than average pearls already placed in the Endergate).
posted by rifflesby at 10:04 PM on February 4, 2021 [1 favorite]

Another point against the "why cheat" crowd: which is more likely, a human makes an inexplicably poor decision, or they roll a 1 on a 10²²-sided die?
posted by Panthalassa at 10:44 PM on February 4, 2021 [6 favorites]

It sounds like the statistics were based on the loot tables and an analysis of the game code. Has anyone collected empirical data on the unaltered game's behaviour in this situation?
posted by polytope subirb enby-of-piano-dice at 11:42 PM on February 4, 2021 [1 favorite]

per the Polygon article "why fake a fifth-place speedrun"?

Maybe he plays with improved drop rates all the time? I mean, nobody likes grinding: improved drop rates are an obvious minor hack that can make things less tedious while still being enjoyably random. And without the mathematical analysis, it would be easy to believe that it's a small enough thing that it isn't really cheating, not like running a teleport script or whatever.
posted by Joe in Australia at 12:53 AM on February 5, 2021 [2 favorites]

I mean, nobody likes grinding

Also— this is all within a competitive media environment. They're all chasing viewers and subscribers— the more exciting you can make the game is a large factor in determining how much money you can make. "One of the top 10 fastest Minecraft speedrunners in the world" is pretty good branding.

So, if you could tweak the game subtlety to mean that whenever a viewer decides to watch your stream, they're more likely to see something exceptional, that's seems pretty darn tempting.
posted by Static Vagabond at 8:43 AM on February 5, 2021 [2 favorites]

I supposed it's not impossible, but a "glitch" where a table gets overwritten sounds far more like the kind of bug that might happen on an old console game (where the software was super-optimized and close the bare metal) than in a PC game written in Java

The entire reason Minecraft took off (mods, pluging, textures, and generally the community adding stuff to the game that could be shared with other community members) was early enthusiasts just hacking shit into the game to make it different. Some of those early people ended up working at Mojang itself, but overriding the probability tables is kinda what has always happened.

As an aside: while Java prevents some attacks (accessing an array out of bounds isn't just going to run whatever happens to be in memory at that bad index), there is still plenty of room find exploits by decompiling everything and seeing how it works.
posted by sideshow at 8:44 AM on February 5, 2021

Also, for "why cheat": every year a few Cat 4/5 cyclists (4 is "I actually practice and have some sort of training plan, but I'm not super dedicated", 5 is "I'm right off the street") get lifetime doping bans after testing positive for PEDs after races that had between $100 to $0 in prize money. They don't even have the YT subscribers this Minecraft dude has.

If there a competition, people are going to want to twin, and some people are going to be willing to cheat to win. Doesn't matter the stakes, the reward, or anything.
posted by sideshow at 8:50 AM on February 5, 2021 [4 favorites]

re: "why cheat a 5th-place run", someone in the comments of Matt Parker's video, who had watched the run in question, pointed out that Dream's run was in fact on pace to be the world record,

Which raises the question, have other players had their runs subjected to like analysis? If Dream thought others were cheating that would be a powerful incentive. C.f. performance enhancing drugs.
posted by ChurchHatesTucker at 10:26 AM on February 5, 2021 [1 favorite]

I'd say the most damning evidence is that Dream paid (?) for that defense document which, while not quite at Giuliani levels of "competence", is clearly not all that credible.
posted by maxwelton at 11:06 AM on February 5, 2021 [4 favorites]

Probably the most common form of speed run cheating is splicing together segments from different recordings to make them appear like a single run. For example, in a game with ten levels, a cheater might use video editing to combine their best play-through of each individual level into a single video.

This is basically the same as the dart-throwing hijinks in the Matt Parker video: Each individual clip is real, but the video still “lies” by omitting all the failed attempts between the successful ones. (I find this interesting given that Dream did not apparently use this form of cheating.)
posted by mbrubeck at 12:53 PM on February 5, 2021

Another note on "why cheat a 5th-place run" that sat well with me was given by Karl Jobst, a YouTuber whose channel is filled with speed running related content (not specifically speedruns). The Biggest Cheating Scandal In Speedrunning History (quote from the "why a good player may tweet" section of the video, starting ~15:52):

Players don't cheat to get a faster time. They cheat to get a time, faster.

That video is a nice primer to the whole situation. It explains what a Minecraft speedrun entails, who spotted Dream's luck first, the back and forth reports, etc. It also mentions something else people have been doing to try and understand the numbers from the mathematical reports: simulations. Have a computer try millions of times to get the same luck as Dream showed. Would have really liked to see Matt Parker bring out his Excel spreadsheets to show us how to simulate the barters!
posted by Martijn at 4:09 PM on February 6, 2021

I don't speed run competitively, which is to say that I don't do it at all and only have a vague idea that it's about more than pressing buttons really quickly. But this cheat looks both elaborate and easy to detect. Isn't it very likely that more players are cheating by breaking Minecraft's system for randomised world creation, to give them a particular carefully-chosen seed that has (e.g.) an End portal already loaded with eight or nine Eyes of Ender right near their spawn point?
posted by Joe in Australia at 8:34 PM on February 6, 2021

Matt Parker on Twitter: "First I invent the new unit: “ten-billion human-second century” and then PewDiePie mentions “Dream out here getting ten billion of a human century of whatever”. Close enough!"
posted by Pendragon at 2:00 PM on February 12, 2021 [1 favorite]