Kalindu Sachintha Wijesundara19.67 FMC MEAN -

[somecuber23]

I know how it feels to be accused of cheating and nobody ever deserves to feel that way; I just don’t want the cubing community to become a toxic place where every record is accused of being fake.

There's no smoke without fire.

If a good chunk of top FMC solvers are suspicious of this due to the technicalities of the solutions (that go beyond most of our heads), then this record should rightfully fall under scrutiny.

 

[NezyDaBest]

congrats on kalindu for breaking the sub 20 barrier!

 

[JDAW6]

There's no smoke without fire.

If a good chunk of top FMC solvers are suspicious of this due to the technicalities of the solutions (that go beyond most of our heads), then this record should rightfully fall under scrutiny.

How about someone share the solutions? So we can have a gander at it. If the solution is super suspicious then yeah I’ll agree it’s fake but just claiming it’s fake because his average has gone down a lot is quite ludicrous.

 

[Kit Clement]

How about someone share the solutions? So we can have a gander at it. If the solution is super suspicious then yeah I’ll agree it’s fake but just claiming it’s fake because his average has gone down a lot is quite ludicrous.

These were shared to the FMC Discord about 18 hours ago. The evidence posted in this thread is mostly based off that, but the main highlights from what I found in the discord server:

• Many solves don't look at a lot of EOs. Some have a lot of EOs listed, but all of solves only really look at a few of them. Most high level FMC competitors look at many more possible EOs for possible continuations.
• The DRs found are unusually long for someone who claims to average around 22 moves. A few solutions use 12-13 move DRs, which wouldn't even be typically considered by someone who has an average that low, as you'd typically expect a 24 on a DR like that, assuming you even find the optimal finish.
• The solves sometimes only find/explore about 2 DRs (Daniel did an analysis of luck based on exploring 7 DRs, see below). Most high level FMC competitors find about 10-15 different DRs and then pick the best few to find finishes.
• The finishes on DR averaged 9 moves across all of the solves. 99%+ of DRs don't even have a finish this short.
  • Daniel Egdal ran 1 million simulations of 6 attempts where 7 DRs are found across each attempt, and the lowest average DR finishes came out to 9.66. This suggests that a streak of six solves like this is less likely than 1 in a million. This doesn't consider possible cancellations beyond one move, which is quite rare anyway. so it still highlights how impossibly short these DR finishes were. (edit)
• One of the solves using NISS had continuations for both L/L' options listed for a given EO, but given the context of when this move happens, L or L' is completely irrelevant for the continuations in that given context. This is something that anyone experienced with DR would understand.
• All six solves are the optimal solutions after a given EO start, and all match exactly what Cube Explorer gives for that finish, up to some cancellations/trivial substitutions.

It's a nearly unanimous consensus among those who have experience in the event and who have reviewed these notes that this is a clear cut case of cheating.

EDIT: Because this post is getting shared quite a bit, I figure I should mention that while I have good FMC results in the past, I do not use DR myself. I follow FMC methods at a high level so I understand the general approach of DR, but this post is meant to just be me summarizing what others have said - I am trying my best to not insert my own opinion into this post and just collectively observe what has been discussed by those who know better than me.

 

[JDAW6]

These were shared to the FMC Discord about 18 hours ago. The evidence posted in this thread is mostly based off that, but the main highlights from what I found in the discord server:

• Many solves don't look at a lot of EOs. Some have a lot of EOs listed, but all of solves only really look at a few of them. Most high level FMC competitors look at many more possible EOs for possible continuations.
• The DRs found are unusually long for someone who claims to average around 22 moves. A few solutions use 12-13 move DRs, which wouldn't even be typically considered by someone who has an average that low, as you'd typically expect a 24 on a DR like that, assuming you even find the optimal finish.
• The solves typically only find about 2 DRs. Most high level FMC competitors find about 10-15 different DRs and then pick the best few to find finishes.
• The finishes on DR averaged 9 moves across all of the solves. 99%+ of DRs don't even have a finish this short.
  • Daniel Egdal ran 1 million simulations of 6 DR solves, and the lowest average DR finishes came out to 9.66. This suggests that a streak of six solves like this is less likely than 1 in a million. This doesn't consider possible cancellations, but it still highlights how impossibly short these DR finishes were.
• One of the solves using NISS had continuations for both L/L' options listed for a given EO, but given the context of when this move happens, L or L' is completely irrelevant for the continuations in that given context. This is something that anyone experienced with DR would understand.
• All six solves are the optimal solutions after a given EO start, and all match exactly what Cube Explorer gives for that finish, up to some cancellations/trivial substitutions.

It's a nearly unanimous consensus among those who have experience in the event and who have reviewed these notes that this is a clear cut case of cheating.

Well it pains me to say but I do think that his solves were fabricated.

 

[Kumato]

These were shared to the FMC Discord about 18 hours ago. The evidence posted in this thread is mostly based off that, but the main highlights from what I found in the discord server:

• Many solves don't look at a lot of EOs. Some have a lot of EOs listed, but all of solves only really look at a few of them. Most high level FMC competitors look at many more possible EOs for possible continuations.
• The DRs found are unusually long for someone who claims to average around 22 moves. A few solutions use 12-13 move DRs, which wouldn't even be typically considered by someone who has an average that low, as you'd typically expect a 24 on a DR like that, assuming you even find the optimal finish.
• The solves typically only find about 2 DRs. Most high level FMC competitors find about 10-15 different DRs and then pick the best few to find finishes.
• The finishes on DR averaged 9 moves across all of the solves. 99%+ of DRs don't even have a finish this short.
  • Daniel Egdal ran 1 million simulations of 6 DR solves, and the lowest average DR finishes came out to 9.66. This suggests that a streak of six solves like this is less likely than 1 in a million. This doesn't consider possible cancellations, but it still highlights how impossibly short these DR finishes were.
• One of the solves using NISS had continuations for both L/L' options listed for a given EO, but given the context of when this move happens, L or L' is completely irrelevant for the continuations in that given context. This is something that anyone experienced with DR would understand.
• All six solves are the optimal solutions after a given EO start, and all match exactly what Cube Explorer gives for that finish, up to some cancellations/trivial substitutions.

It's a nearly unanimous consensus among those who have experience in the event and who have reviewed these notes that this is a clear cut case of cheating.

thank you for adding relevant and important information to this thread. i feel like a lot of people are forming their opinions solely off of hunches and that's incredibly silly to me, so all this info is very useful :3

 

[kubesolver]

Two important factors:
- FMC is the event where you're guaranteed WR result if you are cheating. More like using engine in chess rather than other speed events where you still need some skill and preform well under pressure
- the person had access to scrambles before the competition

These two in my opinion should be enough to cancel a record even without additional evidence of cheating.

However I think the evidence presented already is enough to conclude it was unfair result. It looks like person not only isn't skilled enough to get this result but even to comprehend if the solutions could be plausibly figured out by the best competitors.

Having followed a lot of cheating instances on other games it looks similar to the cases where someone would open an engine during a chess game, saw a suggested solution and thought "Oh. Surely I could have thought of that myself!" and then finding a rationalization why would they play it themselves anyway.

 

[tseitsei]

Anyone who is capable of doing DR and HTR solves on a reasonable level (let's say around 25 moves or lower average) would find those notes extremely suspicious. He just doesn't seem to check almost anything compared to other good or even semi-good solvers. So either he has some new completely groundbreaking way to instantly see if there is a short finish for a given DR or he is cheating.

Also on top of that the sheer luck needed for such short finishes from these relatively mediocre DRs is insane. One guy on the discord server ran a computer simulation of 1 million solves and his means from that comp were significantly luckier than ANY mean during those 1 million simulated solves. I mean you can get really lucky on a 1 or even 2 solves in a row but this guy did it 6 times in a row.

 

[abunickabhi]

I know how it feels to be accused of cheating and nobody ever deserves to feel that way; I just don’t want the cubing community to become a toxic place where every record is accused of being fake.

The cubing community is already toxic.

anyway looking at Kalindu's FMC notes, it looks like he did not pursue too many EO options. I have seen Chong Wen go insane in checking a lot of possibilities. I do not know DR well enough to comment on the number of moves he took to reach DR state and the less moves finish he got.

 

[An Animating Cuber]

I bet this is going to go the same as Uladzislau’s 19, a clearly cheated result that the WCA lets stand.

 

[bldcuber314]

I know how it feels to be accused of cheating and nobody ever deserves to feel that way; I just don’t want the cubing community to become a toxic place where every record is accused of being fake.

If somebody does cheat, then they deserve to be accused. I agree that we shouldn't be jumping to conclusions, but this does seem suspicious.

 

[EthanMCuber]

I know how it feels to be accused of cheating and nobody ever deserves to feel that way; I just don’t want the cubing community to become a toxic place where every record is accused of being fake.

Not every record is being accused, or even half of them. This is the second one in 2024 iirc, and both were accused for good reasons. I understand defending someone because they're nice, but if there's evidence, you have to get at least a little suspicious.

"accused of cheating" and "cheating" are different things. I understand if someone is being wrongly accused of cheating, but if someone has cheated before or is most likely cheating, they do deserve to feel that way, as they did what they're being accused of. (You were also one of these people and admitted it, which is why lots of people didn't trust you.)

Also, I saw you said on the 8.09 OH WR thread that you didn't know which cubers to trust, which is exactly what you said you didn't want. If you don't want it to be toxic, don't be toxic.

Sorry if this sounds harsh, and I personally think he cheated, but we'll have to see how the WDC responds.

Not exactly true, God’s (in this context of the word not exactly a god from religion and stuff, but more meaning perfect optimal solver) number is if you were to give any scramble to the perfect solver, it would take at most 20 moves.

yeah this is what I meant, I was just trying to get away from religion lol

 

[UKCubeFan]

Well it pains me to say but I do think that his solves were fabricated.

I haven't looked at the guy's record on WCA, but I will. Suspicion is okay, but everyone has a right to a defence.
Sudden big increase in FMC ability
WCA delegate - with access to whatever
FMC people saying it looks

No one mentioned this (I think) - see how the guy does at the next event, whenever it is.
If his results are similar - what's the problem?

 

[OtaMota]

If his results are similar - what's the problem?

The possibility of that being cheated as well?

 

[Clock_Enthusiast]

I haven't looked at the guy's record on WCA, but I will. Suspicion is okay, but everyone has a right to a defence.
Sudden big increase in FMC ability
WCA delegate - with access to whatever
FMC people saying it looks

No one mentioned this (I think) - see how the guy does at the next event, whenever it is.
If his results are similar - what's the problem?

I don't really do FMC, but I think it's certainly worth investigating if the statistical likelihood of this happening is very small with his previous results. That being said, he also may have performed extremely well. I think the investigations are of greater interest right now given that this mean is a world record (and a significant one, as it's the first sub-20).

 

[Shah]

Sorry, just want to add this. His PR average on regular 3x3 event is 19.xx seconds which makes it even slower/harder for him to check many solution possibilities to get the best one (because in an FMC attempt you would do scrambling and solving cube many many time).

I said this because I experienced this my self, I use DR-HTR in FMC and on regular 3x3 I'm averaging sub 18-19 seconds (my PR avg is 15.xx tho) and really find difficulties to check many EOs + many possibilities of continuation from each EOs due to my lack of speed on regular 3x3. So, I guess this is very suspicious.

I'm the guy that finished second in that comp with 24.33 Mo3. I don't think speed is that relevant when you're half decent at FMC. In an actual FMC solve, I barely solve/rescramble the cube more than a few times. I can just go back and forth comfortably. Things like NISS prediction, doing a linear finish on a solved cube to switch to inverse (for HTR/DR) and so on really help once you get a hang of them. I've not done speed solves in a long while and I probably average closer to you (or worse) in 3x3.

 

[Chris Choi]

Sorry, just want to add this. His PR average on regular 3x3 event is 19.xx seconds which makes it even slower/harder for him to check many solution possibilities to get the best one (because in an FMC attempt you would do scrambling and solving cube many many time).

I said this because I experienced this my self, I use DR-HTR in FMC and on regular 3x3 I'm averaging sub 18-19 seconds (my PR avg is 15.xx tho) and really find difficulties to check many EOs + many possibilities of continuation from each EOs due to my lack of speed on regular 3x3. So, I guess this is very suspicious.

Honestly, you dont need to be that fast that 3x3 to check a lot of EOs. Firstly, writing down EOs doesn't require any movement of the cube. it is easy to write down all the eos from just the original scrambled cube. secondly, you can often check multiple EOs from the same continuation (maybe they differ by a final move at the end and stuff). I dont think his 3x3 speed is necessarily indicative of anything suspicious. I concur with the things Shah says as well. Unless you're going to the inverse which u might do a total of 2-3 times per going through EO, you really dont need to be solving the cube for every EO you check. And by the time you reduce the EOs to a set of satisfactory DRs to continue from , that's like less than 10 at that point. Hence, I dont think we need to be giving his 3x3 results any more attention.

 

[gruuby]

I know how it feels to be accused of cheating and nobody ever deserves to feel that way; I just don’t want the cubing community to become a toxic place where every record is accused of being fake.

It's accused of being fake because of the evidence leading to that conclusion. It's not an accusation we just got from nowhere. We aren't questioning Max Park or Eryk Kasperek because their progressions are well documented and didn't happen in a week.

 

[UKCubeFan]

He got 22, 24, progressing each competition.

From what I can see on WCA Kalindu would go straight into top 100 (FMC Av) at No. 1.

How fast do they update records on WCA website, if it happens?

 

[qwr]

Having followed a lot of cheating instances on other games it looks similar to the cases where someone would open an engine during a chess game, saw a suggested solution and thought "Oh. Surely I could have thought of that myself!" and then finding a rationalization why would they play it themselves anyway.

It's probably worse than that. In chess the top level players only consider among a few candidate lines to play, and their skill comes down to evaluating how good each line is. Here it seems like these solutions are like random sequences just pulled out of thin air.