No, it's not what you think. I'm eyeing the world record for MOST RUBIK'S CUBES SOLVED WHILST RUNNING A MARATHON. It's currently 100 (in a 4:45 marathon).

I'm a faster marathoner than the guy who set the record. But I'm a slower cuber. I taught myself to solve the cube in 1980, worked hard, and got my average down to about 40 sec. These days, out of practice, I probably average about a minute. With a little effort I could probably get back to 45-50 sec.

Uli Kilian says he averages about 30 sec. (not while running!). So... I have some work to do. I'm looking for the fastest way to get from my current solving technique down to 30ish sec. Now, in principle a minute average is already fast enough (100 x 1 minute = 1:40:00), but I'm sure there is significant overhead doing this while running; also the less time spent solving the less it will impact my running form. (Also I'd like to run the marathon in under 4 hours. I don't think there's any world-record benefit to that; Guinness gives you 5 hours. But my self respect demands I go for 4. I've never run that slow anyway!)

This is what I do now:

1. Top corners (always starting with blue)
2. Top edges
3. Bottom corners
4. Bottom edges
5. Middle edges

Yes, I am being lazy by asking here, but my goal is not to get super fast. I just want to get a bit faster. Does anyone have any recommendations on the best (= takes the least time commitment) way to go about it? Will I have to learn a completely new technique, or can I perhaps just optimize what I do now?

Thanks!

If you slightly adjusted several things, learned F2L (an easy feat for logically minded cubers) and learned perhaps 3-4 new algorithms you could probably get your average below 30 seconds in a matter of a month.

I would be glad to teach you some techniques via some video chat venue, and I'm sure there's many other more talented speedcubers who could do the same.

I'll be happy to help you out. Unfortunately I think it would be better for you to learn a new method. Tonight I'll try to produce some algorithms that may help you on your way to Fridrich. It's not necessary to learn all the algorithms for Fridrich at all (I use very few) but you can still get faster times. I reckon we can get you down to sub 30.

Feel free to join the Daily 3x3x3 Average & Single 2012.

My best single solve was 35 seconds with a corners first method but my average was higher, closer to where you used to be.

Switching to F2L I was able to get an average down below 30 seconds, so perhaps that is worth a shot.

Dave

I promise you I'll try beating your record within 5 years ^^ I'm also aiming for two world records, not-speed wise but quantity-wise I currently use a 32 sec average friedrich, but I think the best option at this point is F2L which can help get 10 seconds down on the average time.

Seems silly that you could just run to the finish line, stop 10 ft before the end to break the cubing record before completing the race.

F2L will definitely help, but at first it will be much much slower than your current method. It took me a long time for F2L to beat my old method. If the marathon is soon, you might not have enough time to practice to make it that good.

Can you elaborate steps 3/4? I wouldn't recommend learning a new system there. I'm just curious.

First of all, It is really a nice idea to brake this world record!

I think it would be the best if you could find a speedcuber who is willing to tech you.
From the technique you use, the best "upgrade" would be the Gilles Roux method. Unfortunately it is used by only a few speedcubers. So maybe it is easier to learn the Fridrich method from a speedcuber.

You can find many videos, and solution descriptions, but the eye to eye learning is always the best!

I know it's been mentioned several times, but F2L. The whole solve revolves around the first two layers. 5-second F2L means a 10-15 second solve for many. A 15 second F2L is about 25 to 30 seconds. But if your F2L is not efficient, you can find 30 seconds to be a tough (but not impossible) feat. You can learn 1-look LL which not even the best of the best have done, but if you don't get to that step fast enough, you can kiss that time good-bye.

Although F2L is the most important thing, OLL can also be quite important. It's not necessary to use 1-look OLL by any means (I've never bothered to learn it), however a couple of tricks can significantly speed up a solve to someone who isn't accustom to the Fridrich method.

The first thing you have to do is make a cross on the top face. There are three different situations, each of them can be solved using relatively similar algorithms to one another. Below the images are the algorithms.



The Horizontal Bar: F R U R' U' F'.



The "L" Shape: f R U R' U' f'.



The Dot: F R U R' U' [F' f] R U R' U' f'. The bit in brackets can be abbreviated by doing the slice move in between the front and back faces in the direction of F.


Now that the orientation of the cross is covered, now it's time to do the orientation of the corners. This is the second look in your '2-look' OLL method. I developed these algorithms so that they really only rely on two OLL algorithms, the Sune and the Anti Sune.


The Sune: R U R' U R U2 R'.

The Anti Sune: R U2 R' U' R U' R'. (The Anti Sune is the normal Sune algorithm reversed.)

Each of the following cases is made up of a combination of those two algorithms. I have put them in brackets to indicate where they are.


Front and Back: [R U2 R' U' R U' R'][R U2 R' U' R U' R']. This algorithm can then be abbreviated to R U2 R' U' R U R' U' R U' R'.


Headlights: [R U2 R' U' R U' R'] U' [R U2 R' U' R U' R'].


Opposing Corners: [R U R' U R U2 R'] U2 [R U2 R' U' R U' R'].


The Weird One: [R U2 R' U' R U' R'] U [R U R' U R U2 R'].


The Other Weird One: [R U2 R' U' R U' R'] U' [R U R' U R U2 R'].

Wow, I wake up in the morning and have tons of answers. Thanks everybody! Looks like most of you say F2L.

One thing that goes without saying, I guess, but I should say anyway, is that I'm interested in minimizing solve time from first seeing the cube to solve, rather than time after an inspection period. I see this in Wikipedia:

"Besides the first step, which can be planned during the customary 15-second inspection time, the entire solve of the cube consists of executing predefined algorithms based on the state of the cube."

Those 15 seconds are effectively going to count as part of my solve time. Is this an issue?

Well honestly, I think the whole idea of the record is silly, but it's there, and as the two communities I'm most active in are marathon running and puzzling, it seems made for me. I'd imagine Guinness would have an issue with running to the end, stopping, and doing all the solves -- that's not solving them "while running a marathon". But there would seem to be a big gray area. Can you walk? I don't see why not. Stop for each solve, and run in between? Hmmmm.


Bottom corners: first position, then orient. Positioning is zero or one transform, pretty fast. Orienting is performing a baryon (twist three corners) or a meson (twist two corners) or sometimes both. I could optimize this. (I'm using Hofstadter's terminology from 1980; I'm sure it's outdated...)

Bottom edges I solve by what's probably a rather unusual method, that could probably be sped up if I did something more standard. I cycle opposite pairs of bottom edge slots into the middle layer, and using two degrees of freedom (left side and middle layer) position and orient the correct edges. Then cycle the slots back to the bottom layer, and do the other pair. Occasionally an edge from each pair will be in a slot of the other pair, which makes it slower.

I don't think you should worry about the 15 seconds. Most speedcubers only use a fraction of that time to completely figure out the cross and possibly an F2L slot or two, but for someone who is running a marathon it would be easier just to start solving straight away.

Would a better cube shave off some of your time? Probably, if you haven't been into speed solving in years.

Also, how would scrambles be done? Would you have to have a running partner scramble the cubes while you solve? I would think that Guiness would have an issue with self scrambled cubes.

-d

So I tried out solving while running for the first time this morning. Solve times:

1:56, 1:24, 1:36, 1:47, 1:08, 1:30, 1:31, 1:10, 1:56, 1:33, 1:55, 1:36, 1:37, 1:16.

As a first effort, I'm happy with that, I guess. Stiff cube, and I was running on trails. It affects my running form less than I expected, but it takes so much attention that running speed is compromised. I will need to run faster while solving, or spend less time solving.

Thanks! I think you have "horizontal bar" and "L-shape" switched, though.

The corner orientation sequences are basically what I do now. ("Sune" = my "baryon"; two corners = "meson".) Actually my sequences for the meson cases are not identical, but they are the same length.

No, the algorithms are right. You have to make sure you hold the "L" shape in the same way I showed in the image.

Oh, my bad, I was starting from solved and expecting to flip the edges that were colored. But because they also permute, that just doesn't work. Got it.

These algs for the Weird Ones I found to be more efficient than the provided still using only Sunes/Antisunes:

The Other Weird One: (Headlights)

[R U R' U R] U2 R2 [U' R U' R' U2 R]

Weird One: (Chameleon) (Reversed the other)

[R' U2 R U R' U] R2 U2 [R' U' R U' R']

I use these variations as they are faster for me than the more direct OLL algs. However, we're all different people so Luke's just may be faster for you.

OK, I'm working on getting F2L faster. Will just take some practice, I think.

But now, how about permuting the last layer? Should I learn Fridrich's 13 transforms, many of which also have to be learned backward, and/or reflected? What simpler set of transforms would suffice?

Thanks!

Many of the permutations are useful, but the G-Perms take a considerably larger amount of effort to get right. This is the site I learned off of. On that link, the G-Perms are numbered 8, 9, 10, and 11. These are the ones that are more difficult to learn. I did know them at one time but a 1-month break was enough to lose the recognition. My times are not at all affected by just using two-look PLL for these cases since even when I knew them the recognition then alg took as long as just 2 algs.

One tip I used for learning F2L, was to take off all of the stickers on the last layer. That way, only the first 2 layers are showing, and it helped me recognize what to "look for" for solving the first two at the same time. It made it easier to recognize how pieces were moving around, instead of just applying blind algorithms.

Hopefully that made sense.