• Welcome to the Speedsolving.com, home of the web's largest puzzle community!
    You are currently viewing our forum as a guest which gives you limited access to join discussions and access our other features.

    Registration is fast, simple and absolutely free so please, join our community of 35,000+ people from around the world today!

    If you are already a member, simply login to hide this message and begin participating in the community!

Will 3x3 reach it’s limit?

Do you think a sub-3 single is possible?

  • Totally

    Votes: 33 70.2%
  • Maybe

    Votes: 10 21.3%
  • Absolutely not

    Votes: 0 0.0%
  • It’s impossible to tell right now

    Votes: 4 8.5%

  • Total voters


Jan 2, 2019
Louisiana aint saying my city
With the amount of 3x3 world record singles that have been broken in the last 3-4 years, it has me thinking about the possibility of a world record solve that is unbeatable, and or impossible to surpass. I feel the 3.47 single shook the cubing world, and many didn’t expect it to be possible, especially from a small cuber in China that had an in-competition average of 7-seconds, plus I think I can speak for the whole cubing community when I say that the fact it wasn’t Max or Feliks was pretty shocking as well. So with all this in mind, we have to stop and think, will the 3x3 single get to a point where it is literally impossible to beat, and will remain unbeaten for the entirety of the cubing world to go on. Now I can see someone getting 3 seconds and possibly sub-3 seconds. However, if that were to happen I feel it would not be possible to be beaten again, considering how lucky and rare a scramble would have to be for it to happen, also you have to include the fact anyone could get, “said” scramble in comp, and it could be someone who averages 30 seconds who would screw it up or it could be someone who averages 5-7 seconds and could actually achieve such a feat. So with all this concluded, in my personal opinion I feel sub-3 single is 100% possible, and I believe you could say anything past is impossible for a human to do, because you would have to have cross and f2l practically done and have to get an OLL skip.

One Wheel

Feb 24, 2016
First off, if a 3.47 single shook your world then your world is pretty shaky. Cool? Heck yeah. Surprising? Sure, a little. Not world-shaking.

As far as the limit, it will be interesting to see. We’re already getting near the limit of hardware: a lot of the gains that have been made in the last 40 years have been as a result of improved hardware, but I would be surprised if there is as much as another 0.1 seconds left to gain there.

I was trying to figure out maximum TPS and thought of Flight of the Bumblebee. I found a claim of a guitar performance as fast as 1300 BPM, or 86.6667 notes per second. The problem with that is that those notes are largely played with tension in tendons of the wrist and forearm rather than muscles for each movement. It may be possible to use more tendons and fewer muscles with very strong magnets, but that would have to be very strong magnets and would have some trade offs. I suspect the limit is around 12-14 TPS using hardware very similar to our current hardware.

The most significant advances will have to be in efficiency. Right now the best cubers average roughly 45-50 moves, IIRC. I’m just guessing that it is entirely possible that with a few decades more development a method of two-looking an ergonomic 35-move solution may be developed. 12 TPS for 35 moves, plus timer start and stop, is about 3.15-3.2 seconds for an average. A lucky 20-move solution is very plausible at some point, which would mean that with optimal execution sub-2 is hypothetically possible.


Jun 27, 2019
my house, cubing.
It just needs a decently fast cuber and lots of luck (Du got an Edge OLL and PLL skip and XXCross.) And yes, solutions just be improved. We’re capping TPS to be around 15 so efficiency needs to be improved. As Feliks said in his 2018 interview, it just takes a lot of time practising and studying solutions, as there are so many ways to do a Cross and F2L, not even considering trying to plan XCrosses, forcing edge OLL skips, etc.


Dec 24, 2015
If there's a limit, we're not there yet.

The theoretical limit would be a two-move scramble (as in, M') in the hands of someone who's very good at timer starts/stops; maybe somewhere around 0.2 sec is reasonable. (I don't have a Stackmat and I don't know what reasonable start/stop delays are, but 0.1 sec seems about right.) Anything above that can't really be a hard limit.

(Warning: speculation ahead; I'm not trained in statistics.) In fact, thinking about the limit in terms of scrambles or solutions isn't even the right thing to do. If you just abstract out the cubers and their methods, and just look at the times directly, this becomes a question of how often a sampled random variable beats the previous record: something very well modelled with the harmonic numbers / logarithm function, independent of the underlying distribution, under the assumptions that the underlying distribution (i) doesn't change over time and (ii) is continuous. (*) Assumption (ii) is not completely accurate, since in actual comps the times are quantised to centisecond precision, but it is at least "approximately" true with what our current record is like—we can start worrying about quantisation artifacts when we get finally get below 2 seconds, say. Assumption {i) is also not completely accurate, but the inaccuracy is biased in favour of my argument: as One Wheel said above, methods will get better, which makes records happens faster.

(*) I'm guessing there's a name for this result. Basically, under the two stated assumptions, the probability of the \( n \)th result being a record is \( 1/n \), so we expect to have \( H_n=1/1+1/2+\cdots+1/n\approx\log n+0.577 \) records in \( n \) results. This grows slowly, but without bound; the probability of getting a record decreases over time, but it never hits zero. Now, if we actually look at the distribution of times, I think it's fair to say that the solve distribution is not Gaussian-like if you look at the tails: for one, Gaussians have infinite support, while negative times are impossible. For CFOP, last layer skips and such play a huge part in lucky singles, so times don't necessarily follow a unimodal distribution. One might say something like "3.47 seconds is such a huge drop from 4.22", which is true if you're looking at the times alone while ignoring the underlying distribution, even though CDF-wise, they might not be that different.

Also singles are dumb, caring about singles is dumb, averages are the only thing that matter.
Jun 29, 2019
And this is definitely possible because Feliks got the sub wr unofficial record and max did to. I definitely think that this record will probably be cut in half before it stops. And just think maybe one day people will be able to see a fmc solve in inspection and do that as there solve.

Tony Fisher

Dec 6, 2009
Visit Channel
Talking about limits suggests there is a point where improvement must stop. That is not the case though. Improvements will be increasingly smaller so they can continue forever. Sure there will be barriers that can never be passed but that doesn't stop continued improvement (in theory). Several things could ultimately stop a certain time being broken. Probability for one. If a record relied mostly on a 1 in a quadrillion lucky case then the world might simply run out of time and be swallowed by the sun before there's any reasonable chance of it being broken. Evolution might stop it too. There may be a turning point where man simply does not get faster at cubing and we become decreasingly unlikely to beat a certain time until there's virtually no chance at all. These things mentioned are over thousands and more years so it's a pretty safe bet that the record will steadily improve for many years to come unless the popularity dies away.

Want to hide this ad and support the community?