[WR] Feliks Zemdegs 4.73 3x3 single

[efattah]

Is there any data on this anywhere?

Well I am an electronics engineer, mechanical engineer and physicist, and I developed scuba diving computers and timing devices and ultrasonic communication systems for many years. The two main sources of error in the stackmat will be:
1. Manufacturing deviation in the crystal oscillator used to keep time (error in ppm)
2. Unit-to-unit variance in the capacitance of the finger-touch sensors (this means that some stackmats will 'start' and 'stop' faster or slower than others when you touch your fingers and/or remove them)

Of these the second will be the greater source of error. The first would give an error of around 0.0001 seconds on a 5 second solve, depending on the quality of the crystals. The second source will be much greater and could be as high as 0.01 seconds and this error will be fixed and not related to the length of the solve. Without knowing the exact construction of the finger touch sensors and the various quality checks in their manufacturing process it is hard to say exactly how big the error is.
For the less technically inclined this means that if Felix had done the exact solve on a different stackmat the recorded time could have been as much as +/- 0.01 seconds higher or lower.

Mats did 4.744 +/- 0.01 = 4.734 to 4.754
Felix did 4.737 +/- 0.01 = 4.727 to 4.747

 

[1973486]

Felix

4 times now pls

 

[Christopher Lai]

Slow Motion Footage of 3x3 Single World Record
Feliks Zemdegs recently broke the 3x3 Single World Record with a time of 4.73 seconds, just barely beating Mats by 0.01 seconds. Congrats to Feliks Zemdegs!!

 

[Dom]

Wow! Nice job, Faz!

I knew this was coming soon when I watched his Ao100 video. He broke Mats's record at least once in that video, with I think 4.60 (someone correct me if I'm wrong).
And yes I'm not surprised that Mats took it so well. He and Feliks both practice good sportsmanship.

Edit: fixed auto correct

 

[1973486]

Felix

okay what's going on

 

[Dom]

So this slowed down version really shows the initial reaction from Mats. So upset! But then he smiles and congratulates Feliks.

Also notice that guy with the long hair in the back going nuts? Very appropriate reaction.

Edit: fixed auto correct

 

[Dash Lambda]

Mats was using a Valk3-M ??

As mentioned earlier in the thread, I was actually more impressed that Mats beat Felix in the overall competition. The record is great too of course (congrats Felix)! Although WCA will need to start implementing new rules about the minimum amount you need to break the record by in order for it to count as a new record. The stackmat timer itself has a certain error margin in its timing. In fact it is not entirely clear to me that this new record beat the old record by more than the intrinsic error of the timer itself. Like in the 100m dash in track and field, less than a certain margin is considered a tie.

Yep, he was using a Valk-M. Chris Tran got rather poetic about it.

As @Meep said, measuring out to hundredths can be seen as a way of accounting for error. The error from the trigger circuit and the quartz resonator (especially within the span of 5 seconds) is negligible, probably lower than +/-1/10,000 of a second.
The time it takes for the solver to trigger the timer is, of course, counted as part of the time.

I honestly wouldn't contest the sub-1 2x2 solves on the grounds of hardware limitations, let alone sub-5 3x3 solves.

EDIT:

2. Unit-to-unit variance in the capacitance of the finger-touch sensors (this means that some stackmats will 'start' and 'stop' faster or slower than others when you touch your fingers and/or remove them)

Hold on, they use capacitive sensors?
...
Well that kind'a throws off my numbers. I don't know too much about those.

 

[YouCubing]

Felix

stopppp

 

[SolveThatCube]

This is epic! Well done Faz. But Mats' reaction was like "OMFG! This can't be actually happening! .01 seconds! Really?! @#*%!$&*!#*$%!" He recovered well though.

Congrats again Feliks!

 

[Dash Lambda]

So, now that this record is up on the WCA, let's take a look at all the nxn records...
Singles:
2x2: Maciej
3x3: Feliks
4x4: Feliks
5x5: Feliks
6x6: Kevin
7x7: Feliks
Averages:
2x2: Lucas
3x3: Feliks
4x4: Sebastian
5x5: Feliks
6x6: Feliks
7x7: Feliks

****ing christ, 2/3...

 

[tx789]

I can think of a better way for him to of gotten the WR back.

Betting it by .01 betting the person who took the record off you while there're sitting next to you.

 

[Loiloiloi]

Well, you can think of truncating the thousandths digit as an implementation of this already. For example, 4.739 and 4.730 is a tie.

However, 4.739 and 4.740 are not a tie, even there is only a difference of 1 ms between them.

 

[Dom]

okay what's going on

stopppp

OKAY guys no reason to be alarmed. Everything's going to be okay. I fixed my auto correct. Sheesh.:rolleyes:

...As @Meep said, measuring out to hundredths can be seen as a way of accounting for error. ...

...I honestly wouldn't contest the sub-1 2x2 solves on the grounds of hardware limitations, let alone sub-5 3x3 solves....

I was actually thinking that maybe we should go to the thousandths decimal place for times like these. Do you think they'll ever make timers that go out 4 decimal places for events like 2x2?

 

[pjk]

Congrats Feliks, impressive as always.

Maybe it was obvious to Feliks, but I think it's pretty awesome how he chose to do break up the obvious already made pair on the green cross to do a different x cross with a better continuation. Its also awesome how he didn't do the obvious 2nd pair first. I love it when fast solves are fast because the solver did something smart.

Agreed - nice solve.

 

[Weston]

Well I am an electronics engineer, mechanical engineer and physicist

Wat?

 

[obelisk477]

Wat?

https://ca.linkedin.com/in/eric-fattah-67902740

http://www.bloomberg.com/research/s...01614&previousTitle=Liquivision Products, Inc.

 

[RhysC]

good grief what have i done

 

[Dom]

good grief what have i done

Somebody get that poor kid another SD card!

 

[efattah]

So Feliks beats Mats by 0.007 seconds, within the error of the capacitive touch sensor. Wanna know who really solved faster? Cameras are the most accurate timing devices, if run at high frame rate. This is why the IAAF will resolve 'dead-heats' in track and field with the actual video of the race and the electronic time of the race is corrected every race by the timing extracted manually by the video. So, someone with good video editing skills should just create a video with both solves side by side starting from the moment they remove their fingers from the stackmat. However the frame rate needs to be faster than the victory margin. Margin is 0.007 seconds, 1/0.007 = 142, so technically you would need 142 frames per second to distinguish the victor. This is even beyond the 120fps limit of most fast cameras. Which only further shows that the victory margin is even less than 1 frame at 120fps and questionable. Still, a side by side video would be awesome.

 

[Dash Lambda]

So Feliks beats Mats by 0.007 seconds, within the error of the capacitive touch sensor. Wanna know who really solved faster? Cameras are the most accurate timing devices, if run at high frame rate. This is why the IAAF will resolve 'dead-heats' in track and field with the actual video of the race and the electronic time of the race is corrected every race by the timing extracted manually by the video. So, someone with good video editing skills should just create a video with both solves side by side starting from the moment they remove their fingers from the stackmat. However the frame rate needs to be faster than the victory margin. Margin is 0.007 seconds, 1/0.007 = 142, so technically you would need 142 frames per second to distinguish the victor. This is even beyond the 120fps limit of most fast cameras. Which only further shows that the victory margin is even less than 1 frame at 120fps and questionable. Still, a side by side video would be awesome.

Cameras aren't that reliable for absolute measurement, only relative measurement. I guess you could film the solve and the timer and stop it when the solver's hands hit the timer, but that would still be difficult to discern (just imagine trying to decide which frame his hands hit the timer on a record this close) and it would be rather impractical for comps.
Might also have to take into account the latency in the timer's display at that point...

Though, we still don't actually know if that's within the trigger's error or not... That is, unless you know what specific component SpeedStacks uses in that timer. I'd gladly take one apart and see what I can dig up from part numbers, but I don't have one...

I was actually thinking that maybe we should go to the thousandths decimal place for times like these. Do you think they'll ever make timers that go out 4 decimal places for events like 2x2?

If an event takes so little time that 4 places becomes meaningful (as would be the case for 2x2), then it's basically just a contest of who can flap their hands up and down the fastest, and I don't think that's really in the spirit of speedcubing. For close records like this one... Eventually, sure, I can see someone making a timer accurate to four places. I can't see that happening any time soon, though, just too expensive.