Hacker News Reader

Comment by fracus | Hacker News Reader

It is kind of hilarious the fraud was uncovered by some esoteric technical investigation when there was definitive evidence in plain sight the whole time.

"For instance, the "1996–2001" copyright date seen on the title screen in Groobo's video is inconsistent with the v1.00 shown on the initial menu screen, suggesting Groobo's run was spliced together from runs on multiple different versions of the game. Items acquired early in the run also disappear from the inventory later on with no apparent explanation."

It reminded me of a joke in Northern Exposure where Ed uncovers a thief by listing circumstantial evidence and then adding one piece of evidence that he almost forgot.. he witnessed the thief stealing a radio.

View on HN

the whole thing is so chaotic. I don't get the impression that Guinness World Records or anyone in the Youtube comments is aware it's a segmented run, but it is clearly stated in the original submission:

https://web.archive.org/web/20091121211238/https://speeddemo...

it's cheating for other reasons too, but it's confusing as fuck trying to figure who originally thought the record was segmented or not.

Being a segmented run is not a problem, the issue is that the segments are not from the same save meaning that the RNG could never produce the given levels in a single run, which is a requirement for segmented runs.

But yeah a lot of the "it's obviously cheating" comments seems to be from people not realizing that the player is allowed to re run the level and no required to go in blind, some of them may only have played it multiplayer where there isn't a save state for the game.

Hmm... I'm not familiar with the speedrunning community. However, if I check the rules it cites, while it does say this today[1]:

> There obviously needs to be continuity between segments in terms of inventory, experience points or whatever is applicable for the individual game.

> manually editing/adding/removing game files is generally not allowed

If I check the Wayback Machine, it doesn't go all the way back to 2009 when the run was done. Earliest is 2012, and there it doesn't say anything about continuity or editing/adding/removing game files.[2]

[1] https://kb.speeddemosarchive.com/Rules

[2] https://web.archive.org/web/20120701000000*/https://kb.speed...

Those sound like rules you'd file under "too obvious to even mention". It's not a speedrun if you're editing files at will.

Very specific changes are sometimes allowed by specific game communities.

Like, what would the alternative be? People would just hook the intro to the credits and win after one second.

If you don't mention the rules that are too obvious to mention, someone will think they can put together a segmented speed run with their best times from each dungeon level smooshed together into one run. After all, that's a lot of good segments.

And a little bit of 'rng forcing' using outside tools to get drops they want that they can't get and then tweak your fireball damage to make things work for the final boss.

Thats just not what a segmented run is. Just like any game that allows save-states, you could abuse save states to retry any part of 'the run' changing the entire save file to a different run and 'splicing' them together is splicing, which is not allowed. Tweaking fireball damage is literally hacking the game to get the desired result. Unless you find a way to perform ACE with your inputs. Which is not the case here.

Which is probably why they have clarified the rules since then. And why they have always had judges evaluate the submissions to check that the comply with the spirit of the rules.

This happens pretty often. Karl Jobst has a lot of videos about cheaters, and it’s always interesting to see how many were ridiculously obvious in hindsight. Frequently, people won’t expect a good player to cheat, because “they don’t need to.” But cheating, at thay level, tends to be about saving time waiting for the perfect RNG.

It reminds me of the blatant aim hacks that for example Flusha used in Counter Strike, yet people online still defend him.

There's observation bias there - the cheaters that we see are the obvious ones, we never know who gets away with it when they got everything to appear perfectly consistent without any telltales.

Like even for RNG, it'd be possible to fake that on a real console, with extra hardware writing to the bus. We'd never detect that, all we'd have would be the statistical arguments of like "one in 10^10 tries" or whatever.

Reminds me of the argument that rich people, given political power, won't act to harm the government/society for their own enrichment "because they're already rich!".

Everyone knew it was probably faked, but for the longest time there was no smoking gun that refuted the run's legitimacy without any plausible deniability. You have to remember that the speed runner gave plausible answers to many of the concerns brought up, which took a lot of concrete work to disprove (for example, proving that the map seeds they used required different runs, which required special tooling to bruteforce check against).

Were the two inconsistencies I quoted not smoking guns?

The problem is that the menu is not part of the timed portion of the video and is just there for the viewer. In his response the runner said he simply used an old random intro he had when he combined the segments. The disappearing item might not have been enough either since it would not have had any consequence and could have been a simple continuity issue when redoing segments. If it was the only thinks we had pointed out it likely would not have been removed based on the response from SDA.

The premise of being a segmented run uploaded by an amateur a decade ago provides a _lot_ of leeway.

Nobody knew it was fake before the investigation started.

The investigation didn't even start because they were suspicious. The team wanted to create a TAS and decided to recreate Groobo as a starting point, then optimise from there. First step of making a TAS was to bruteforce the map generation seed.

And they quickly encountered issues, and it quickly devolved into an investigation of just how "cheated" it was. Partly because for their TAS usecase, they didn't really care if the run was played on multiple versions, as long as it was still a single save file with a single map seed. But I assume it was mostly curiosity at that point.

Groobo's defence was didn't come until near the end of the investigation, and was more or less: "Yes, I did that. But it was all disclosed to the SDA judges, and they allowed it under the rules of that time".

What's funny is that I immediately noticed the inventory inconsistency but since I'm not a Diablo speedrunner I assumed it was some kind of intended glitch that's known and allowed. Then I continued reading and saw they just noticed it.

The famous Van Meegeren forgery "Christ and the Adulteress", which was sold to Goering, used a more modern blue instead of the original Vermeer's ultramarine - and for a good reason, because Van Meegeren could not, during wartime, obtain ultramarine from London, which is where the only contemporary vendor was located.

It could be detected by contemporary means, but no one thought about it in time, and there was no wisdom-of-the-crowds yet.

Eh these kinds of runs pop up all the time in the speedrunning scene. Busting fake runners is an easy way to make views on YT.

First signal is when it's from a person who doesn't have a history of consistent achievements on the public leaderboards. It's not a rule - sometimes there are extremely talented up-and-coming legit runners who go from pretty bad times to WRs in a matter of months (e.g. Derek MacIntyre in Factorio 100%), but their past runs often show steady improvements. Many fake runners pop out of nowhere, or have few/inconsistent top times/PBs.

Second, every serious speedrunning community has people who check the video recording frame-by-frame; other rules usually include a whitelist of mods, posting the save, (for games that have RNG) the seed, and (for games that have replays) the replay. Breaking the rules just outright disqualifies a run. The frame-by-frame videos usually are enough to uncover cheating. E.g. in Minecraft any%, stronghold navigation relies a lot on luck and intuition; but there are visual cues for which room might lead to the end portal, and if a runner doesn't take the split second to look in that direction it's a tell-tale sign they've already scouted the seed.

But yeah, some formerly-respected runners actually put a lot of effort into cheating, and it takes another expert runner (or even the whole community) to bust them. More in-depth for Minecraft: https://www.youtube.com/watch?v=XoEQ8wtawPI

Groobo has records in several other games, he also held previous records for Diablo, the previous runs looks like they have less issues so it seems to be something that escalated from just replacing one or two levels in the video in to making changes to the game and splicing everything together from unrelated runs.

That's why you post the seed. If Diablo seeds the RNG from the current time, the rules should at least require posting the approximate time range (e.g. down to a minute). NTP also isn't hard.

Minecraft allows the use of an external tool that extracts the player coordinates to help triangulate the stronghold (there are runners who can do the triangulation in their heads in 10-20s (check out Couriway) but that will easily be a huge difference for a WR).

Timers need to hook into the game to trigger start/stop (WR attemps are often retimed from VOD, sometimes a close tie in a 4h+ run is resolved to single frames). If Diablo runs are to be taken seriously, a tool should exist to extract/inject the seed for verification.

Kinda sounds like Parallel Construction

Reply on news.ycombinator.com